Lab JC

lately, the discussion has been gently poking around in the sonic muck, for the purpose of a closer look at distortion as an electronic sound effect. electric guitar is an obvious and simple context to begin with, but i want to make it clear that the threads lead in and out in ALL directions. there are other places to start in terms of sound design. film (especially animation) is a good one and so different from popular music, for example. still, the electric guitar emerges as an instrument of change. in the 1960s, particularly in connection with psychedelia and the british invasion, new sounds and mass production transform music forever. but in terms of distortion, there are guitarists whose uncanny intuition have transformed the perspective of what it means, both as a vocabulary and a source of symbols.

the blues guitarists definitely relied on distortion to do what they did. but it wasn’t intentional design: little amps turned up loud, aggressive and cutting though. nothing wrong with that! when one thinks of distorted guitar, while it starts with the blues, it rapidly began to shape shift. link wray, jimmy hendrix (and in some crazy parallel way, sam andrew and james gurley at the same time or even a little earlier…), lou reed, robert fripp, and jody harris and bob quine (as a Gemini pair) are seminal figures. that short list could be expanded a bit, but it’s in the ball park. they transformed the electric guitar into both a new voice and into an expanded environment. what isn’t so obvious is the boxes they stomped on to do what they did… distortion effects by rodger mayer, bob myer, and others, are crucial to this evolution. my bloody valentine might seem a long way away from muddy waters, but not if you follow the sound design.

there is no doubt, to me, popular culture preceded academic exploration of sound design in the 20th century. electronics has everything to do with this. primarily because most of the advances in the technology appeared so fast, once they appeared, the uses lead the understanding by a progressive margin. it did make for a lot of bad music! still does! but it also changed music forever. sound design is now a permanent feature of most music making. this process did eventually slow down, in the analog context, and reversed itself almost entirely. now, people search for ways to use analog techniques to make new sounds… as they do with digital stuff. even “classical” acoustic music has taken on this idea in an utterly analog sense. lou reed’s “metal machine music” was performed by an all acoustic orchestra in germany not long ago, for example. but it didn’t start out that way. the sounds came first.

the digital era has kickstarted a new cycle which is ongoing… but the entire arc shouldn’t surprise anyone because it is a feature of the post modern. it’s intertwined with the economy, politics, and disaster. many will point to musique concrete, stockhausen, or even colombia university in the 50′s and 60′s as the root of the tree… i want to say out loud, this is bullshit. raymond scott, in one perspective, yes. more than anyone else, he created a system to compose with sound design (for scott, words are also sounds) in a popular, widely understood manner, before anyone had imagined such a thing. the others: important, yet only partially formed and mostly referential experiments (stockhausen’s “hymnen”). still, the bad apples led the way, and the rest of the power structure didn’t catch up for some time. they have done their best to see it doesn’t happen again. but it most likely will. maybe not in the USA? we americans could use an “arab spring” ourselves.

john cage fits into this, but not for the simplistic reasons most people imagine. i think he took for granted the similarity between words and the “language” of images and sound (it’s the reverse of raymond scott: sounds are words)… but it isn’t a perfect fit. it is uncontroversial to say academia has always had a problem with the human body. the physical senses, other than sight, aren’t far behind. the gestural powers of the body are different from words. it is a blind spot that still messes with the academic perspective. there is a long pattern of alternately completely denying or hysterically wallowing in “physicality” (that’s a word, not the meat). does it have to be so stark? as far as the current discussion is concerned, i will suggest there is an inexorable connection between distortion and the human ass. words cannot explain what these two know without them. and, there are things that cannot be explained well with words. a wall of sound needs no words. and the hands of the musician, together with the technologies that extend the reach, aren’t separable. for me, it’s another form of dance.

in the previous posts, i separated the most well known distortion effects for guitar into “fuzz” and “overdrive”, and linked them to two approaches: clippers and intentional “maladjustment” of amplifier operating points and headroom… i want to add a subtle but very interesting feature of older designs. most modern distortion effects rely on op amps to do these jobs. because the gain is so high and feedback is necessary to define the transfer characteristics, the slide from linear to non is very quick. when discrete devices are employed, the range of operation near cut off and saturation can be used much more sensitively than is possible with op amps. there is much more of a range of nonlinear distortion generated by operating a stage in this way, even though clipping may never occur. strange uncorrelated artifacts can be generated that have a wider and often less palatable sound, but still contain enough of the original signal to make out the source. this idea, in combination with other techniques, can be used to make truly original sounds, that cannot be made any other way. that is a problem for a manufacturer trying to make something repeatable… but there are ways to do it. let’s have a closer look…

a word about triodes: triodes were the first practical voltage amplifier, and also the simplest. the gain is low to modest… they are relatively linear without tricks, and the cutoff and saturation transitions are not very complex, especially at signal currents. compared with bipolar transistors, they aren’t as rich in terms of distortion generation because of their electrical simplicity. modern VHF triodes have more to offer here, than the older audio ones, because they have sharper transitions, are more sensitive to grid current, and can make a lot more higher order harmonics before clipping. pentodes offer even more interesting possibilities… especially in the transition from the “resistive” part of the characteristics (at low voltages) to the “constant current” part of the characteristics. in order to keep things simple and get the most from the discussion, i am going to use a hybrid circuit to demonstrate the ideas. it’s easy to build and listen to… but it is absolutely plausible to do this with tubes alone.

here is the idea… you can clearly see the connection to the previous circuits.

the first stage is simply a jig to make the various Q3 operating points adjustable. i have shown this stage with a typical low noise NPN bipolar transistor. but, what you need to get here is a germanium transistor. use a socket, try anything… the transition areas are much stranger than modern discrete silicon, and more interesting. the buffer is absolutely necessary so don’t waste your time trying to wiggle out of it. this is not hifi. it could also be a FET, and you could current source it for upscale. i made all three transistors the same… this is for demonstration purposes. use what you have.

the point of being able to adjust the bias and B+ (and the load of the tube stage) is that you now have control of the operating points. you can dial things in the middle for symmetrical clipping, or skew off towards cutoff or saturation (the diode line). because there are two stages, the flexibility is outrageous. in the example above i have leaned out the amplifiers and lowered B+ so as to plumb the areas at the very corners of conduction. this makes for a very asymmetrical waveform that creates an odd mix of harmonics… by varying the bias alone, you can get an effect similar to a ring modulator. the added overtones are related to the difference in the tops and bottoms… it can be awful! it can be genius. have a look:

here you see output compared to input. note both the asymmetry in the X axis as well as the Y. this setting bottoms out at -20 mV… but doesn’t top off until around 140 mV! note also the rising level… this would be whackier in reality. in practice, this is not as predictable from one transistor or tube to the next. especially at low B+. i have not talked at all about the coupling issues between the stages and why they are done as they are… suffice it to say for now, that slowing things down a bit and choking things off a bit (with grid or base current) is also a technique. one thing at a time… the range of adjustability of the “jig” is more than enough to account for differences from one gadget to the next. this is but one example of many possible. 1000 sounds and no presets!  next, i think it will be time to look at EQ for distortion… that is the main ingredient to the “special sauce”. more soon.

in the previous postings, i started rambling about rock and roll distortion, and even a little bit about how to make make some. but, now i want to get a little more into the subtle issues. again, this is introductory and in blog form… it’s a huge subject that a book could be devoted to. as always, we can just start somewhere, and build on it later.

i mentioned the possibility of both “soft” and “hard” clipping and put up some images of what it would look like on a scope. the addition of odd order harmonics (3rd) and frequency emphasis/de-emphasis are key elements to the sound, but limiting (clipping a waveform is unambiguously limiting it in amplitude) is also a crucial feature. why? given enough gain after the input, the dynamic range of the output is brutally reduced. it doesn’t make any difference if the input signal is small or large, one size signal comes out. this translates into an overall smoothing of variations and a smearing of all intent or accident into one contiguous sound.

a brilliant feature of this radical reduction in dynamic range, is another interesting effect: sustain. because small signals are boosted to the amplitude of the clipping threshold, and large signals are reduced to the same, any overdrive or decay is essentially transformed into one long (sometimes insanely so…), one sized event. think of carlos santana playing lead in “oye como va”… that was a big muff, the classic clipper fuzz.

sustain is an interesting transformation of the guitar. electric guitars are plucked instruments whose decay has mainly to do with the lossy qualities of the mechanical parts and the thermal absorption of the materials it is made out of. those instruments made with bolt on necks and light wood and metal parts (stamped steel bridge) have very little natural sustain and a signal envelope that rises fast with plucking the string and dies out soon after. it’s NOT a piano. the fender telecaster would be a good representative example of this kind of thing. the resonant qualities of the physical construction don’t help either in storing much energy in the bandwidth of the instrument (to drag it out longer). twang! by adding a fuzz box to the signal chain, this percussive dynamic sound is transformed to a smooth legato vocal or bowed like sound that pushes the guitar into roles other than comp or texture, which was the first role of the guitar. now, the guitar is a voice. for good or evil? hahhahahahahahahah! jimmy page used a distorted telecaster in “heartbreaker”, so that may hint at the answer to that question?

clippers are a good place to start with this stuff because they are simple. but their simplicity also makes them one trick ponies… all the way fucking out. that’s an important concept for the effects business. every box, a specialist toy. you can understand right away how idiotic hifi is as a business model today: a bunch of old smelly men, looking for status and distraction, in all the wrong places. the sense of exploitation is both closer to the surface and the herd more obviously delusional. for every piece of hifi gear sold, hundreds of musical instrument or pro audio gadgets are sold. it’s a really big business. i wander… if clippers are the logical far end of the distortion business, then what comes at the beginning?

“overdrive” is the industry term for small to considerable amounts of level sensitive progressive distortion. it can go as far as total clipping madness, just as fuzz boxes automatically do, but it can take a while to get there, and do some other things along the way. as i briefly mentioned before, exceeding the input range of a gain stage, deliberately, will also distort the output signal more or less, and depending on the arrangement, can wind up clipping or, even cutting off the amplifier entirely. clippers can’t do that in a progressive manner, because of the generally restricted dynamic range. not to be underestimated in the subtlety or heinous destruction possible, overdrive can be crudely described as having both more dynamic shifts in effect AND a wider range of possible harmonic and enharmonic sounds.

because there is a period of transition from “clean” to distorted, and this range can be related physically to the fingers of the musician, a “touch sensitive” character to the effect is often the important quality. with fuzz, there is no touch… it’s all or nothing (actually, it’s all or noisy). overdrive is ALL about the touch. but how does one build that into a design? good question? i was hoping you might ask… (talking to oneself is a sign of madness).

the transducer responsible for getting the movement of the guitar strings into the amplifier is generically called a “pickup”. it is a magnetic pickup related intimately to an alternator… the magnet(s) and coil are fixed and the steel strings move. by varying the field in proximity to the coil, an alternating current is induced in said coil. there are 2 common arrangements of this design: the single coil and the humbucker. the humbucker is essentially just two out of phase series connected single coils with one positioned to cancel common mode… (by reversing the magnetic field in one) it does this at the expense of some bandwidth, but gains output.

as you can imagine, it is not a particularly sensitive approach… the strings don’t have much mass or flux varying permeability and the magnets have to be powerful and the coil thousands of turns. just in order to get 50 – 300 mV of peak AC signal… all that copper and impedance invites stray pickup… especially of line voltage. shielding helps, but the long cables typically employed mean hum and noise pickup. i can imagine jeffrey jackson or dave slagle making field coil pickups, with pure iron or permendur poles! if they don’t, someone will sooner or later. but it was talked about here, first, you johnny come latelys… i don’t care. it won’t make or break the system. humbuckers can double the output and can have decent common mode rejection… but the sound is different. a bit less dynamic. more rolled off and midrangey. fortunately, musicians are WAY more pragmatic about this sort of thing than audiophiles generally are. the electronics that follow are heavily relied upon to fix, enhance, or modify shortcomings of the transducer/instrument front end.

okay, given that the input range will go from nothing to approximately 50mV (vintage single coil) and up to 1 V peak to peak (“super hot” humbucker), that gives one a place to start. the amplifiers typically used for guitar sound reinforcement are already designed with this input sensitivity in mind. exceeding it will overdrive the amp. that is a discussion for another day… because there is always some kind of input level adjust at the amp, differences in pickup type and output level can be taken to account. in designing an overdrive effect unit, we will want to match the output range to something similar.

in a previous post, i drew a test jig in which the bias, load and power supply range could all be adjusted. the purpose of doing this was to be able to practically experience what shifting the operating point deliberately into the “wrong” could do for effect. if we drive this stage with a good clean signal, and provide some guitar sensitive eq to the result, this will make an excellent test bed for overdrive. let’s have a closer look how this might be arranged…

the original concept has been developed here to include a safely adjustable B+. with the “bias” pot at 2.7K, the “load” pot at 220K, and the B+ pot at minimum setting, this is the operating point you might wind up with… lets look at the waveform.

ahhh, nicely clipped! with 1 VAC pp sine wave in, we have a very slightly asymmetrically clipped 70 VAC pp. with only 630 mV of bias on the 12AX7, even self biased, we have totally exceeded the input range.  lets mess up something else… just by turning the “load” pot down to 20K, and leaving the other pots as they were, we get:

and the waveform is now very asymmetrically clipped and the distortion is much more complex.

finally, i will leave you with the next logical development. here, we have a “clean” gain stage with some house cleaning added… the 68K grid resistor helps with the rfi and emi crap the cable and pickups find. the cap bypass, so often hated in hifi, and so truly beloved in musical electronics, is very important for leading edge “bite” and dynamics. some of the “touch” comes from doing it this way. now, the input range can be pickup sized: it’s shown as 50 mV. danelectro territory. the 1 meg input gain adjust pot is the grid leak for the overdriven stage… with the “bias”, “load”, and B+ pots all set as shown, here is what you get…

here are images of representative waveforms of this approach. soon, we can add some EQ and further housecleaning…

so, that last post got my mojo workin…. a quick note is all i’m going to drop here. if one was to add a mid cut filter and buffers, here would be a way to do it, this will be the lowest noise version. a simplification would be to dispense with the last buffer. then add a second “gain” pot and you could use the “saved” section of the 12AX7 as an extra gain stage and deal with the highZ output (bad with long cables). madness.  still, this is a distortion box and they are all, by definition, noisy. ever use a big muff? then you already know… as is, 100mV in should give you 100mV of death metal out. a one meg pot will back it off. the mid cut is real muff style and will mess things up appropriately. here are the waveforms at 1KHz and 200Hz…

ok, i should stop. but here is the added extra stage arrangement… not as well behaved as the first one. but it has adjustable output level. you may not care. these are some of the techniques one can begin with to make interesting distortion effects.  this is enough.

all this talk of linear circuits has turned the world a little grey. i need me some color. and not just primary color… chartreuse, cerulean, kaput mortum violet. it’s time to change channels… and not just to “pimp my ride”.

among the diy hifi 1%ers and the ultrafi diaspora, distortion has mainly received a seriously bum rap. yes, some like to give the 2nd order harmonic distortion artifacts an easier time of things, much to the mainstream tradition’s chagrin. most often, distortion measurements are used as a blunt weapon to humiliate, belittle and cajole the errant peers and maladroits of the electronic “sciences”…  and of course, as meaningless marketing. there is some older research to back up the lack of importance of even order harmonic distortion in terms of listening. but is it really all that bad. for playback, within reason, i say yes. the whole rationale for hifi hangs in the balance…! but is distortion just inherently bad? as it turns out, for music making, there is almost no amount of distortion too much… in certain quarters, it is the very foundation of musical flavor.

one interesting observation after many years of looking at practical linear circuits: most of the really troublesome analog problems come from switching behavior… interestingly, most of the problems facing digital systems are analog effects. turning on or off is not something any analog device does perfectly. all the devices we use to make digital systems are completely analog at some range. in terms of making electronic distortion, this dualism/factoid has much to offer.

ahh, distortion, oh glorious buzz, crystalline harmonics and angelic glitch! a head full of droned out cranked up amps and i see two opalescent visions of myself, and there’s one rising from each of my shiny boots… and if the melvins, swans (from back in the day, “nobody beats you like a cop, when you’re in jail…”), and cop shoot cop (shine on elizabeth!) don’t wake you tha fuck up to the world of right now, nothing will. maybe you are wondering what the hell i am ranting about?

fundamentally speaking, “distortion” is any deviation from the “original” signal that results from passing through a stage of amplification or other electronic function. obviously most functions create distortion under this definition. just think of a log amp, or an equalizer… the very concept of EQ actually puts passive speaker crossovers into a whole new light: distortion boxes! but most hifi people think only about linear amplifiers. enormous effort goes into making a “transfer system” (having a transfer characteristic) which is in all possible ways equivalent to the input multiplied by the gain (a fixed number). one overlooked issue among many engineers and hackers is that just by using an amplifier, we have given up on this idea, because the time it takes to traverse such a practical transfer system made with the devices we actually have to use. the inertia added makes it impossible. time, all by itself, has a say. the deviation can be admittedly small, and can be rationalized by the tradition as a reasonable “group delay” to be factored in. but i am a fan of honesty in electronics. the more stages, the more complex the demands on the supply, the greater the inertia and delay. but i stray…

in this case, we can do the typical western intellectual division of the universe into two opposite forces (one of them is always implied to be “bad”, without saying so. mind/body, mind/nature, male/female, etc.) and describe the types of distortion as “correlated” or “linear”, and “uncorrelated”. correlated distortions are largely going to be a matter of amplitude and phase (time) variations. simple harmonic distortion comes to mind… all the rest falls neatly into the other category as “non-linear” distortions. some of those are sort of “semi-correlated”, such as inter-modulation distortion (sum and difference) which has roots in the linear stuff. but some stuff is just plain whacko, extra evil gnarly shit and who wants to share space with that? (it’s all a matter of perspective…) however, as it turns out, there are subtle interconnections between the two sides that cannot be totally separated. as is nearly always the case!

an interesting analog question to raise is whether or not there are parallels in the material world to these electronic distortions? the answer is yes absolutely, and no. there are some very obvious comparisons. air, for example, is very compressible. but not linearly so; that is to say, non-linear at small and great magnitudes of compression. the sound of lightning, the thunder clap, is a perfect example of a complex acoustic pulse of large magnitude containing all sorts of linear and non-linear components… all created by rarefying and compressing (and ionizing) air, very suddenly. there are more eloquent people to talk about this than me. the amplitude and phase deviations possible, along with the more complex failure mechanisms, one observes with practical transfer systems can also occur in the natural world. not easy to compete with a thunder clap, though… are there particularly electronic distortions that have no equivalent in the material world? yes. and they are all non-linear. more about this later.

most physical musical instrument sounds are ultimately dependent upon distortion of the fundamental event that generates it, even if the spectral content can include mainly non-linear components as well! think of drums, or bells, and other mallet instruments. the sound they make is fundamentally 2nd harmonic: the strike is heard as a high odd order pulse (which could be very complex if it is a hard stick, or simpler if it is a felt ball) but compresses the air in the column or the material of the piston (might be skin, wood or steel, or…) and damps it. the dimensions of the piston and the speed of sound in the material it is made of, determine the root characteristics, but getting it moving requires energy, which in this case comes only from the mallet. the mallet “stops” the piston during the blow. the release can, and usually is, much greater in magnitude than the blow… this is out of phase with the strike and distorted in amplitude over time… etc. the resonant qualities of the “air column” behind the piston (if there is one) filter or store the energy further, enhancing or attenuating some frequencies at the expense of most others. just trying to describe this in such a simplified way gives one the sense of just how “un” linear the creation of physical sounds is. drums are fairly simple! think of a piano… distortion is very much at the heart of our musical taste and imagination.

what makes distortion “musical” or just plain nasty? let’s keep an open mind about this for the time being… this is a long conversation. however, i can give one important and simple starting point as a reference. the sound of rock and roll (and the blues!), is mainly the sound of 3rd harmonic distortion, in higher or lesser amounts. this is a good place to start because it isn’t as complex as a ring modulator or bit crusher… and many people know what rock and roll guitars, keyboards and vocals are supposed to sound like (?). 3rd harmonic distortion is created in abundance in the crappy output stages of push pull beam tetrodes amps (Fender), and in the overdriving preamplifiers of many other well known musical instrument amps and pedals. below is a graphic for 20% 3rd harmonic… a typical value in a blues guitar sound. this is minus the treble boost, mid cut and bass boost you would also typically want… one step at a time.

so, back to the original question, but now lets ask: how do i do that? well, i am going to describe some of the most basic variations of electronic distortion creation, and suggest gently, as is my wont, what the combinations might beckon… the goal here is primarily 3rd harmonic in character. but if we go a little further… cthulu!

one of the most obvious and brute force electronic distortions is clipping distortion. this is what happens when a signal is, by various means, hard limited in its amplitude. as signal amplitude increases beyond a certain threshold, the “top” and/or “bottom” of the waveform is “clipped off”. this can be done with analog means by using a switch, arranging the bias, the power supply, or the gain of the system to deliberately exceed the linear operating range with the available signal. for example, you can make any linear amp into a clipper simply by sending 100 times the input range into it! a further nuance of this approach is whether or not the onset of clipping is “soft” or “hard”: soft clipping is notable in its slower transition to “clipped”… a gradual rounding over that ultimately results in the limiting value. hard clipping is fast and unambiguous. soft clipping is much more complex sonically, even though hard clipping produces far more odd order harmonics and is brighter. the gooey slewing of soft clipping is an area ripe with variations and subtle differences…

in the image below you see a basic diode clipper… two shunt connected diodes, in reverse, will turn “on” at 600mV (the forward voltage of a 1N4148) or so, “shorting” the signal to the clipping threshold (the forward voltage). with 1 volt pp AC in, there is still some sense of the sinewave visible in the slope of the waveform between the clipped tops and bottoms. in the image below that, that is no longer obvious because the signal level has been increased to 20V pp, and the result is almost a square wave. almost because the top is not perfectly flat! diodes are not apparently perfect switches after all?

one of the simplest ways to improve upon the quality of the switch is to put it in a feedback loop (ahh, the dreaded nfb has already appeared). below, you can see how this would be implemented.

here it is… and below that an image of the waveform. here we don’t have the need to drive the diodes with a large signal (the 12AX7 does that) and the top is flatter, although still not perfectly flat! the small bypass cap and the input “impedance” are also necessary for symmetry. the 12AX7 does need to be loaded well, over the frequency range of interest. additional steps can be taken to “improve” the “fidelity” of the clipping (steeper sides and flatter tops).

this is one simple technique, and kind of a blunt tool at that. with this kind of technique, no matter what the size of the original signal, as long as it exceeds the threshold of the clipper, the output is limited and distorted in a predictable way. it is easy to see the similarity between the ideal 3rd harmonic distorted sine wave and what a clipper can do at modest and overdriven levels. this is classically known in the rock and roll business as “fuzz”

traditional fuzz boxes/diode clippers like the “big muff” or the “octavia” or even the “tube screamer” are all based on this approach. but what about more sensitive techniques? perhaps something that can straddle the linear into the hard clipped? for this, we can look first at deliberately choosing the “wrong” operating point for an active device, so as to create a range where the signal level can determine how clean or distorted the output might be… the bias, DC operating points can be chosen to make this possible. another technique could be to cascade specially designed stages so as to optimize just how “wrong” this can get.

next is a test jig you could make that uses pots to make the bias, loading and B+ all adjustable, simply so you could vary everything and really see what “wrong” can do. immediately you should notice that in the middle, the variations don’t make a huge difference… at the more extreme settings, things get more screwed up.

this can be tried with vocals, guitar or keyboard sources for some interesting effects. remember that more signal conditioning will be necessary to make these sounds useful and able to couple to your other equipment. i am trying to talk fundamentals here.

LT spice doesn’t do pots, so i have drawn them as series connected resistors with the wiper connection between…

a 12AX7 is a good choice for this job because it is high impedance and will only draw 5 mA or so if you turn low the bias, load and B+ pots… it will still cook if you leave it that way, but not much danger to anyone if it burns up. cheap or used 12AX7s are ideal for this treatment. russian 12AX7WAs are perfect. because the 12AX7 is a dual, it is a great opportunity to cascade two identical stages like this for extra sick weirdness. the work isn’t done yet because some housecleaning and eq will be necessary to really get the most out of what different kinds of distortion you can pull out of it. but this could get you started… you can also have another look at the completed distorting preamp i did earlier in the tremolo article, for tips with this.

cascading two FB clippers will also do wonders… here you have the core of a tubed “big Muff” pi, minus the tone control and buffer… i’ll save that for later… the waveform follows…

this was fun. more to come…

there are so many definitions of simplicity in the analog electronic world that the description is close to meaningless.

on the engineering side, it is generally used to satisfy certain design preferences or bias. rarely is the method “simple”. there are also varying kinds of electrical simplicity in a physical sense. i like the one that involves a simple transfer characteristic with a first order low pass well outside of the audio band. that is not so simple to accomplish, as it turns out. another one i like is integrated functional simplicity, where many jobs are accomplished at the same time with the fewest tricks. such as, using the natural characteristics of materials and devices to accomplish more complex functions… it looks so simple when it’s done and working! before that, anything but simple. how it fits together is absolutely not simple, even though the end result might conform to several different kinds of simple. there are more kinds of electrical simplicity than this. and many other opinions about it, too.

on the advertising side, “simplicity” most often refers to a commercially constructed ideal of elemental “purity” and “innocence” (both equally meaningless in real terms, but dangerous and seductive fantasy words) and is therefore valuable. you know you want it, you dirty humans! this is often portrayed, at least in the states, in a kind of creepy love story between absolute evil and simple innocence. i love that story… my favorite story! blue velvet and little red riding hood. “mulholland drive”, is one of my favorite movies. also. “let the right one in”, in the original swedish.

on the consumer side, “simplicity” is most often connected to physical appearance. things made from unpainted metal that have one switch and no knobs, for example. with a remote made from the same material with the minimum of buttons. for some reason silver or white, reflective colors, seem to have have become markers for “simple” as well. a mirror is simple? your face is simple? you ARE simple? hmmm. just as gloss black is “elegant”, and polished gold is “rich”…  you occasionally see gear that doesn’t even have an indicator that shows it’s on, but those things don’t generally sell well outside of a particular fetish tribe. gadgets with this kind of appearance infer an inner simplicity, that rarely exists. it’s the idea that matters here, not the reality. remember the stereo in “a clockwork orange”?

for the diy builder, and many consumers, simplicity is mainly a mythical quality that is hidden away inside the chassis and the careful choice of certain types of parts. there is an intense sub genre of parts fetishism associated with it. it’s an opaque “black box” kind of quality. “pure”, but also somehow “classical” in a newtonian steampunk sort of sense, unsullied by 20th century relativism or quantum mechanics. and even perhaps more than a little touch of “right”, as in the sense of “correct”, but with an older ozone flavored alchemical feeling around it. a secret “right”, that only a few enlightened minds are privy to… sounds complicated to me!

i suppose i am not exempt from all this, except i have nothing simple to market or consume… i don’t have anything simple to sell! especially not myself. no, not simple. it isn’t simple for me to design or make anything… and i can honestly say i have not put any effort into making it simpler, either. and anyone accusing me of being simple would probably win my heart over, and the right to walk all over me. that would be music for these ears! oh, to be simple for someone! but, sadly (?), it has NEVER happened. i am beginning to suspect it won’t. and, i AM usually right (my wife doesn’t think so).

for the diyer, electronic simplicity has mainly boiled down to a reduction of components. low parts count = “simplicity”. hahhahahahahahahahahhaah! what a ridiculous concept. it isn’t true. that kind of “simple” means simple to assemble! gear with a low parts count is very rarely simple. each component does many jobs and usually not optimally. the character and nature of the materials plays a much bigger part in the performance. the transfer characteristics of each function are much more fixed, and there is less flexibility in operation, even if the range is relatively “wide”. you usually can’t so easily edit or adjust a low parts count design for optimum, and the nonlinear aspects are features whether you want them, or not. often these sorts of gadgets can’t even be hooked up to just any load. that does make your systemic choices simpler as you may be limited in terms of application. this is not to say that simplicity is a myth. just that it’s more complex than a low parts count implies.

think about what happens at the edge of something… anything. it might only be moving wind around a corner, a jewel bearing in a watch, or a piece of rock skipping on a pond. a waterfall. a meeting between two different states. this is a naturally turbulent interaction. a very low parts count for sure! yet, one of the most complex interactions known… something that still challenges physics, math and modeling on super computers today. something that is still studied and the knowledge base is expanding. it isn’t “complete”, in other words, whatever that means…

there is also the simplicity of the fewest possible changes of “state”. many years ago, there was a man who would go to washington square park in nyc on summer weekends with an edison cylinder player. a big one! he dragged it around with an over sized vintage radio flyer, himself looking kinda like a tall wimpy poopdeck pappy. he would play caruso, and valentino accompaniments, flapper stuff, all sorts of fun stuff… i somehow remember “i’ll never see maggie alone” (“there was her mother, her father, her sister and her brother…”). it sounded fantastic! much better than i had imagined and actually within the limits, authentic. behind the scratchy bandwidth limited old mechanical recording shtick, there was a palpable and substantial experience that electronic recording doesn’t have. i am not alone in this judgment. it is an experience shared by sid smith and dick sequerra too. we had this conversation over pizza many years ago. eddy reichert and i have also talked about it. why is that so?

great sounding player and horn speaker

dick’s straight up answer was simple: fewer changes of state between the original energy and the reproduction. less loss, plain and simple. no, the bass and treble were gone… and surface noise was added. but the middle was “closer” to the original by virtue of the simpler chain of transformations. kinetic to storage media, and back again for playback. electronic media requires many more transformations than that! each one losing subtle info that cannot be retrieved. mechanical recording has a sense of audio verite that is built upon it’s simplicity. if you haven’t listened to a victrola or edison player recently, you owe yourself the experience. that is a kind of simplicity. but it isn’t so simple to describe… it also isn’t optimal.

for another example: input, interstage and output transformers are complex components electrically. they add levels of complexity to a circuit in excess of the reduction in parts they allow. this is not to say there is no place for such components! no. just that they are absolutely NOT simple. depending on the quality and application of a transformer, you minimally have added a 4th order filter with large values of parasitic energy storage (some large enough to qualify the trans in question to have as many as 4 more poles). you add time to the process at hand too, and distribute the changes over the audio spectrum in a varying way. this one component could actually be replaced with dozens of other components and still do the same job in the same way. yes, i know. this argument really catches in the craw of many. but it can be demonstrated.

vacuum tubes, because of the simpler physics that underlie their operation, all by themselves, have a “simpler” transfer characteristic than say, bipolar junction transistors. tubes work because of the physics of heated conductors. semiconductor physics, heated or not, rely on much more complex interactions between materials that are themselves modified to express their useful characteristics within carefully engineered ranges. this statement includes a comparison based on the added complexity of the use of coated cathodes for tubes, instead of pure or thoriated tungsten, for example, with P or N doped silicon. coating a cathode with oxides (calcium, barium and strontium are the common ones) employs electrical chemistry that is related closely to the science of semiconductors. especially when things go wrong! but before i go too far down a road in the opposite direction, i want to make a point. there is a big difference between building a machine for a particular job with easy to apply components (op amps, like transformers, are way easier to use than tubes, and have simple predictable results…) than it is to really consider all the relationships between the parts and weave them together in a way that appears to have been “meant” to happen. it still gives me a great deal of satisfaction to consider that the musical instrument business has had a terrible time trying to replace the push pull 6V6 guitar amplifier. they have had 40 years to do it… it just refuses to die and go away. culture is important, after all. that is easy to say, but not at all simple.

i sign off here with an idea for a really simple circuit, that isn’t simple at all. ever wonder what a triode “sounds” like all by itself? here is a way to do it… yes, all those resistors are needed to get the thing to be stable, and biased. and, the op-amp servo, keeps the DC output well under a mV. you could get rid of the servo and DC couple it to a tube amp with a large cathode resistor on the first stage, just to allow for the wandering output… but nothing else. no headphones – no DC amplifiers.

this would make a good non-inverting line stage. or maybe a headphone amp…? it needs a bigger power tube for typical modern phones actually. perhaps a 6550 triode wired? but totally ok for beyer dynamic DT880S (600 ohm) or similar. you could rearrange this with your favorite or maybe some old tube you wonder about… some safety provisions would have to be added to protect the load from offsets or failures… more about this later. actually, i will build this. interesting…

some will complain about it being a cathode follower and that you won’t “really” know what it “sounds” like… fuck you, moron!

here is the gain stage version, although it inverts phase and will have some gain, so you still won’t know what it “really” sounds like… since whatever goes in, will still be transformed into something else afterwards. so fuckin there!

the bypasses and compensation are probably absolutely necessary and tube dependent. yes, it’s so simple, you will have to treat every tube you put into it as a unique situation and opportunity. not a one size fits all kinda gadget. but fun… totally fun!

i have been in south korea working on silbatone stuff and i wanted to post a little something about a cool circuit i’ve been playing around with for phono. i’m going to do a mic pre front end for ribbon mics with it soon. but it’s very very quiet (just under 1nV per root Hz) and sounds lovely. this is not for beginners or probably those who have a moral problem with hybrid circuits… but it is interesting anyway.

cascode circuits offer a way to build your own pentode out of two triodes, JFETS or BJTs, and have some advantages over standard pentodes… lower noise and no screen current. many have tried JFET/triode cascodes in an attempt to get even lower noise but still hang on to a “tube character”. i would have included myself into that group some time ago but now i believe this has it’s own sound and i prefer it for phono and mic pres… blasphemy! but i have still had some reservations about the harmonic spectrum of the typical cascode. tube or N type JFET (i love the 2SK146 and 147) with high Gm type triode (WE417 or 437, etc.) on top. it is ok for blues, jazz and rock and roll… but i haven’t been totally happy about the classical music thing, or the dense noise thing yet. cecil taylor records or cop shoot cop just doesn’t doesn’t sound it’s best… it’s just a touch too hifi, and that is a four letter word in my book! i think pentode operation has a better timbre even if the distortion is higher. Gm operation of a pentode or cascode can simplify the distortion spectrum over standard operation. i have mentioned this before. but cascode, even single ended, has a slightly forward exciting sound typical of circuits with a touch more odd than even harmonic content… even more so balanced. the distortion is very very low! but the timbre isn’t perfect.

but i have found my new toy! hybrid folded cascode.

as far as i know, folded cascode was done first in audio by john curl (?) with JFETS. he does folded complimentary balanced circuits that are a study in symmetry and beauty. they also work as good as it gets. i think he is still the master of it, but the technique has expanded into chip design for memory and many many other jobs too. a folded cascode is a complementary amplifier in cascode which requires an added resistor to leverage the I/V conversion between the two halves. because the two amplifiers can be independently biased and loaded, it is possible to “tune” the response with more finesse than with a more standard cascode. using a P type JFET folded into the cathode of a triode is a way to have a complimentary amplifier with a pentode like transfer charateristic. the gain can be modest or high, depending on the configuration, and it can also be arranged as a Gm amplifier. i know of no one who has made anything commercial with this other than us… but now you can make something with it! have fun!

here is the basic idea as i have been using it. this is a Gm amplifier (loaded with a constant current source) and a resistor to ground. in this case, i am using a very large valued resistor (1 meg) and getting 60dB of gain… (the sim says 68dB). that is mainly due to the JFETs… solid state varies 1000 times more than tubes in it’s characteristics.

and here are some pix of the circuit in the flesh… the other tube you see is a bendix 6900 used only as a buffer for the folded cascode. the empty tube socket eventually gets a WE437 for make up gain after the RIAA filter. that goes beyond the scope of this discussion. today we are only looking at the first stage, which is a hybrid folded cascode circuit using 4 each 2SJ74 PJFETs and one D3a wired in triode. what you see is the transfer characteristic at 10KHz with 3mV input. the scale is expanded for detail (uncalibrated).

badass. i know it’s not good to brag but this is really very good. actually, it performs very nicely, but most importantly, it sounds just lovely! pix to follow of the entire phono preamp…

this next installment of the EH 12AY7 mod series is going to deal with the power situation. the original has a very clever (if i say so myself) and very economical power supply completely organized around the idea of compact size and market/regulatory acceptability… (much of the pedal world functions far far away from the high voltage world of vacuum tubes). it makes all of the various voltages needed from one 12 volt AC wall wart, with few parts and without any switchmode anything (still too noisy for cheap analog). while fulfilling the design requirements, it is not ideal in terms of the best performance. mainly, because of the limits of the wall wart and supply. also, the method for distributing all of the various currents interconnects them… a drag on any one of them, really affects them all. this is mainly a problem because of the phantom supply. certain modern mics draw a fair amount of current from the phantom. some of them, more than is available for the preamp…

the upgraded circuit improves the isolation and the drive capability of the buffer, but also requires even more current than the original. the solution for all this is a more optimized and dedicated set of supplies for the various jobs at hand. this way we can optimize the various functions. let’s start with the big one: the high voltage B+.

double regulated 220 VDC supply

ok, here you see a pretty straightforward solid state regulator. it has three basic sub-circuits. it is based in it’s entirety on National Semi’s© venerable applications propaganda for the LM317. i am a big fan (although they didn’t write this stuff, bob pease and jim williams are such heroes for me. because of guys like them, the quality of the writing and depth is routinely excellent. the national applications notes are deeply useful and well done. the guy who did the work for the LM317 was very matter of fact and sometimes really funny). an added mosfet current limiter ahead of the chip, and a transconductance regulator after, make it much more sophisticated and safer than originally proposed. some small refinements to reduce the likelihood of high frequency oscillation, and better short circuit protection have also been added. a few words about all of this are warranted.

the high voltage rail calls for 220 Volts DC. this was provided for with a “pi filter” following the rectifiers in the original. in order to reduce hum and variability of the rail, we can arrange for a separate supply and then regulate it. mic pres often deal with tiny signals and the balanced differential circuit used in the design of the mic pre only has so much power supply rejection, and it works less well the higher in frequency you go. a tightly regulated supply takes care of that! however, an interesting problem does arise. high feedback regulators, such as the LM317, or even any of it’s improved versions (such as the wonderful LT1084, 85, 85…), also make high frequency noise that isn’t so easy to compensate for. all amplifiers add noise, but especially those with gain, so it isn’t that remarkable. it’s analog! deal with it. those chips have a LOT of gain. the unregulated supply did have some hum and sag, but no high frequency crap at all. now you know why the DIY world is so dependent on passive solutions… it takes work to deal with this crap! the remedy is not that complex. following the chip with a “no gain” regulator, the stiff regulation is mostly retained, but the noise is reduced by a huge amount. we can also decouple the final regulator with a small film cap, which will respond to transients with aplomb.

let’s have a closer look at the sub-circuits involved. the first mosfet is your basic source follower, with the reference voltage derived from the regulated output of the chip regulator downstream. the 100k resistor and the 15 volt zener make sure the gate is never more than 15 volts away from the output of the chip regulator. the 20 ohm resistor is sized so that if there is a short circuit at the output of the reg., the increased voltage drop across it will begin to cut off the mosfet and protect the LM317. this could be bigger in this case because the total current draw is so small, but since it will be used for stereo, and i am lazy…  the mosfet will heat up a lot under shorted conditions (a decent heatsink is needed on all the chips), but normally doesn’t get hot. the LM317 has a maximum voltage rating of only 35 volts. by combining it with the mosfet in this way, it can handle 100s of volts.

how is that? the LM317 and it’s family of adjustable regulator chips are “floating” regulators. the built in error amp works to match the drop across the reference resistor (in this case, R6), which is connected to the output, with it’s internal voltage reference. in this arrangement, a better more accurate reference is used instead (the LM329). the point is, this drop can be set without a direct ground connection, hence the “floating” moniker. the LM317 can source up to 1.5 amps provided it has a good heatsink and is protected from shorts. the Linear Technology© versions have better performance all around, but do cost more too. i use them all the time for low and high voltage regulation.

the final sub-circuit is very important. sometimes called a “capacitance multiplier” or a “transconductance regulator”, it is essentially a source follower (cathode follower for the vacuum tube folk) with a high impedance voltage reference on the gate that is bypassed heavily with a large electrolytic. this is important. the time constant needs to be LONG. that will multiply the effective “capacitance” of the output by the Gm of the device. a small film cap at the output can be made to behave as if it is orders of magnitude larger. handy. and it has no gain! if arranged in a stable configuration, the output noise can be in the nanovolt range. in this case, the uVolt range is good enough! we are talking about a 200 volt supply! 220V/1uV is 10^9 noise reduction! this is an improvement over the 2mV hum level of the original.

in the case above, i have deliberately injected 100mV of 60Hz hum on the 250 volt “unregulated” input. the 0.8 mV PP output ripple across a 10K load (22mA) speaks for itself (you will need to download LT spice to sim this). mosfets have the Gm and low rON necessary to do this right, but tubes can do a fair job of it too! as long as the load is constant, the supply won’t sag. additionally, while the impedance of a 6V6 or 6L6 cathode follower can’t match the mosfet, it would  be fine for line level stuff, or a screen reg in a push pull amp.

some will complain that it isn’t a tube regulator. i don’t care. tube regulators can sound great! but they are not capable of this level of performance. at least not without a large increase in complexity and cost. and you will never reach the really low noise level. however, if you are willing to go hybrid, there are some interesting possibilities… even just following a tube reg with a mosfet “no gain” reg will do quite a job. that’s for another time.

the heater supply on the original pre amp PC board is regulated and can be kept as is. the prudent upgrade is to feed it from it’s own transformer and filter. perhaps, just removing the chip and tiny heatsink from the board and mounting it on the chassis or larger heatsink will be enough improvement there… we’ll see.

the phantom supply is next.

okay, in the last installment, i outlined 5 areas of improvement to the EH 12AY7 mic pre, to explore and an initial design for how to deal with some of them. these were, gain, isolation, a lower output impedance, distortion and noise, and improved power supply in support of the other issues. i intend to get into it, but first, i am going to inflict some more wandering brain on you… it’s my blog! go make your own.

you still do hear a LOT of tongue wagging and brouhaha over the “subjectivist/objectivist” dualism. i want to say something about this, because it has a bearing on what is happening here. i hope i don’t overdo it, but it was bound to require some clarification sooner or later. it is a work in progress, but the basic flavor is bitter, with a touch of bicarbonate of soda.

“subjective” critics of the traditional engineering approach, to solving the issues of electronically reproduced sound, tend (an understatement) to dismiss wholesale the significance of measurements and analysis based on research and testing. it makes no difference that all electronic gadgets have always involved the use of “high” technology, often based on new science, for the present day, or even in looking back. audio engineering is described by this camp almost as if it were alchemy or magic. for them, it IS! and, they are apparently afraid to know more of how that magic works. it is so prevalent and the group is large enough that this perspective is heavily reused in marketing many highly engineered commercial products today. ironic isn’t the right word to discuss this. schizophrenic is a better one. the difficult thing about this for me is that at the center of this perspective is one very important misunderstood and misrepresented truth.

“objectivist” or “rationalist” critics of the subjective camp self righteously point out the delusional aspects of this. but often have very little self examination of their own. in fact, the wall of outrage directed so intently at the herd on the other side of the river, seems to allow them to avoid this altogether. they are, as a rule, as dense and unpleasant as the others are, and even more defensive because they feel science makes them right. “it’s in writing, mathematics, it’s been measured, for crying out loud!” the status of their situation is seemingly more complicated than the others, because some of them have an awareness of the truth at the center of the other side’s weird universe, that they lack, and yet don’t want at all for themselves. even though it’s true. the less sophisticated ones are just assholes.

so what is this truth? it’s the thing they have in common. that they are obsessively fixed on the meaning of reproduced sounds… and desperately intent upon control of that. but, no amount of objectifying will limit, reduce or expand the range of what is possible. music is not a signal. it isn’t a bit depth. it is a cultural practice. it is subject to all of the messy motives people have. if the gear was really and truly “neutral” (i have a great deal of difficulty with the choice of words here) and music passed on through all the changes of state between microphone and loudspeaker, without the subtlest change… really transporting one from the present location to a site far away in perhaps both distance and time. well, no, it’s not possible in an “objective” sense. that is a ridiculous idea. that is the “subjectivists” truth. although they are generally clueless as to what it means.

reproduced music is not “real”. and whatever “reality” it may have once had, is so socially constructed these days that it can best be compared to fashion photography or advertising. “objectivists” seem terrified or angry that even reproduced, music could be so wildly representative and that the true measure might actually be in emotional or anthropological terms and not scientific ones at all. we live in the post modern time and the “facts” are that fake is every bit as important, or more so, as “real”, for millions and millions of people. the music business is built on that idea. the argument is a case for the emperor of clothes, or ice cream. not math. and resentment over it. what a miserable point of view. the “subjectivists” are at least more fun, even if they are full of shit. when it comes to reproduced music, fun counts for more than misery.

chefs and cooks mitigate technology, science and art day in and out. no one gives them a hard time about it. it is clear that the preparation of food is both rooted in cultural and technological history and innovation. it is also clear that a good chef needs good eaters to be appreciated. if you don’t like shellfish, who cares how well it’s made. yes, marketing and commerce confuse things here as everywhere. but a great plate of fish is very difficult to argue over, if you like fish. this clarity is utterly lacking in the playback world. marketing runs that world… it’s almost all lies, based on their own fear and resentment. and it doesn’t matter that much if it’s “objective/subjective”… they’re both missing the point.

enough. microphone preamps, even cheap ones, have a job to do. to get the signal from the transducer (and it can be a range of very different transducers) and get it into a size and form that lends itself to recording or post production. but it also has an aesthetic: to be “musical”. ahh! how infuriating…

the one “aesthetic choice” i make here, and it is a somewhat controversial one for the tradition, is that this cheap mic pre runs open loop. there is NO error correction applied. what is, is what is, as they say. now, that is not to say that i do not care about distortion or bandwidth… measured performance is very important to me. however, there is more than one way to skin a banana.

the first two areas of improvement were gain and isolation. fortunately, there is a component we can draw upon that fills the bill. a step up transformer. trannies get a bad rap in all but two places, diy hifi and recording studios! and that is funny, because it couldn’t be for more opposite reasons: in the diy world, its retro sex appeal and love of the object in itself. for the recording engineer, it is simply practical. transformers afford isolation from the hostile common mode environment. not just the canceling of it, which is the world of the instrumentation amplifier, but real honest to goodness isolation. accidentally connect the wall mains to it? who cares? flourescent lights? big schmear. 80 feet of mic cable next to a gas generator and a motor bong… no problem! that kind of practical is hard to ignore.

i bought some 1:3 SUT that appear to be tamura units from yamaha mixers from the 70′s. removed from equipment and sold on ebay for a reasonable price. let’s look closer at what they can do…

here is pix of a test set up. the signal comes from my marconi generator to the red and black primary, the scope is and shield ground are connected to the blue and white secondary.

my test setup is a generator, oscope set up in X/Y with the generator on the primary and X ch, and the trans secondary on the Y ch. the AC voltmeter is on the input to make sure the generator is at the same level when i switch frequencies. sorry for the blurry pix. they were all taken with an iphone.

here you see the 25mV RMS input level, and a 200mV  plus change PP output. 1:3.

1KHz @ 25mV RMS input, here's what you get...

25mV RMS

what you want to see is these next images which show the transfer characteristic of the transformer at 100Hz and 20KHz. note the phase shift (oval shape) at 20KHz… quite reasonable actually. and finally a 2V RMS input gives 8V and change PP out without compressing (you would see curvature or bending over). there is both cartesian and X/Y examples. good to know it will survive a closely placed SM57 on a snare…

25mV RMS in @ 100Hz

25mV RMS in @20KHz

now, the important thing here is that the trans steps up from a 600 ohm load nicely and the X/Y is a nice straight line to the right… and no wiggling or funny ground loop stuff visible. this is a good trans, especially for 20$ for the pair. you can expect to pay more than this generally, but good deals do come up.

now cinemag and jensen both make good brand new versions of this, as does lundahl and sowter and others. for vintage stuff, you can use whatever you can find, like i did. these worked out just fine. tamura is good stuff. adding this to the input of our micpre will add both isolation and gain. and they are linear and have enough headroom for the mics they would likely meet. all that is needed now, is to remove the input blocking caps and wire this in instead. it will be necessary to revisit the bass cut feature later, but that can be added simply by using an appropriately sized film cap switched in or out between the trans and the H pad.

2V RMS in @ 1KHz

2V RMS in @ 1KHz

the next change has to do with noise. in order to minimize distortion, i increased the voltage across the FETs by biasing the 12AY7s up, until i got the lowest 3rd harmonic. in this configuration, that turned out to be 18 volts. i would have guessed 7 or 9 would have done it. but the combination of these FETs with this tube worked out that way. this is NOT the lowest noise arrangement.

above 3 – 3.5 volts of drain source voltage, FETs produce an increasing amount of flicker noise. very similar to johnson noise but for different reasons. at or below this drain/source voltage, there is very little. so i have lowered the bias to 1.8 volts. this should put 3 volts at the cathodes of the 12AY7s with 2.5 mA between them. 1.8 V happens to be the breakdown voltage of a yellow LED, which is also very quiet. some kind of LED will make for a good voltage ref for the bias. the distortion is a little higher, but it is still very low (0.1% for 8V RMS out) and the pre gains some S/N (another 5dB). the step trans helps here too as it adds very little noise of its own and 9dB of “gain” ahead of the active stages. the increased current through the cascode helps a little too, and we are on the way.

the lower output impedance of the buffer stage can be accomplished 2 simple ways: increase the current through the tubes; and change to a beefier output tube. both work well and are easy to accomplish. we can change the cathode resistors to 10K each from 33K for a threefold increase in current… and we can use a 12BH7 instead of a 12AU7. the heater circuit will need to be adjusted to accommodate the extra current required by the 12BH7. this will make for a considerable improvement in output impedance without any fancy tricks.

ok, so the first four areas of improvement have had some solution proposed, and the way forward seems plausible. the power supply will need to be flushed out a bit to support these changes, but we can also improve the noise performance by regulating the B+rail and by making a better 48V phantom supply. that will be next.

i think this will become a running post.

ok, i left off “part 1″ with a brief explanation of how and why an “affordable” (let’s just say it: cheap) commercial tube mic pre came to be. some notion of the design issues, along with some of the issues of amplifying balanced signals, staying balanced (in order to reject unwanted environmental noise and interference sources), and above all economy. an effort to design “good bones” into the gadget also leaves us a chance to explore the possibility of building upon this design as an affordable way to get somewhere substantially better. this idea is something that can quite often be expanded to many older professional products, as well as some unusual currently produced things.

from the previous segment, if we simplify the circuit so we can see the “bones”, it is fairly obvious that we have a differential cascode gain stage, followed by a pair of cathode follower buffers. a very simple 2 stage circuit. a constant current source in the common cathode forces a high degree of balance between the two halves. the gain of the circuit using the original parts is about 56dB in practice. this is an “average” and was arrived at using the audio precision “portable one” owned by EH. the spice version predicts 60dB. i take this as a home run and proof that the models are pretty good! the gain is largely determined by the transconductance of the FETs, which is modest. this is enough gain for most mics, but definitely not for ribbon mics, or some older condensors in any job other than close placement. i am not a fan of close placement, but it is a lucky thing that almost every other recording engineer is. still, gain is an area of improvement.

small offsets and AC common mode stuff between the inputs of the diff can be passed along this circuit if one “side” of the output gets loaded down more than the other. as mentioned previously, the 12AU7 buffer does not have a low enough impedance to be “immune” from loading effect. this is an open loop design, so there is no error correction at all. a lower output impedance would be an improvement. it is the case that some “prosumer” gear, namely recording interfaces, that have economical methods of gain adjustment at their inputs, do not have completely symmetrical input impedances because the gain is adjusted with negative feedback, and the “balanced” inputs are subtley different at certain gain settings. this is typical where the input circuit uses an inverting opamp stage although why is beyond the scope of this installment. suffice it to say that as the amount of feedback goes up (the gain goes down), one input’s Z goes down (the inverting) and the other (non) stays the same. if no effort has been taken to account for this in the design (a sign of economy), offsets and common mode can be coupled through by the gain of whatever it is in front of the recording interface (the mic pre). this economy is acceptable generally because the impedances commonly encountered are isolated from the outside world with instrumentation amp circuits and other methods to extend the CMRR of typical op amp based mic pres. but this simple tube pre has none of that! it’s an open loop, “let it rip” kinda thing. it works fine as long as the output sees a balanced load, but if it doesn’t, you can get hum and ground born noise coupled through from input to output, as well as from the power supply. so another area of improvement would have to be isolation.

and finally, noise is a pet peeve of mine… way more than harmonic distortion. nonlinear and enharmonic distortion is rarely nice in pre/amps either. and it is so often the case in cheap gear the S/N ratio and distortion spectrum are roughly considered. it is “cheap” after all. as long as these specifications fall into a certain range, no further effort is spent and the “sound” becomes fixed by this limit, plus any nonlinear/harmonic attributes part of the system. most of the op amps used in the business have a class B output stage and are also run LEAN at the input. there is so much feedback, it is assumed that all forms of nonlinear distortion are “taken care of” and the noise floor is determined by the Gm of the input transistors (the bias) and the various resistive elements (all impedance creates noise) of the design. this is fantasy. of course useful, but still not what many consider to be the case. many of the most commonly used op amps add crossover distortion particularly to small signals (the output stage has to turn on…), such as those encountered in a mic pre. and of course the noise floor can be impacted negatively with small signals (noise) that just cross the threshold of “on”.

another more controversial issue concerns the classic op amp, which has a gain – bandwidth relationship determined roughly by the miller “compensation” between the first 2 stages. this means that if the gain is 100dB at 10Hz, it is often only 32dB at 20KHz open loop (6dB per octave drop). the feedback error correction is not uniform for the bandwidth of the particular arrangement. there is much less feedback at higher frequencies always… also more distortion and noise. the general opinion is that this is unimportant, as far as the audio bandwidth is concerned. and for cheap stuff, doubly so. it is a good idea to look at the noise and interference sensitivity of the design very closely so as to reduce these things… a wideband open loop design can equal or improve upon this, and need not involve the complexity required for having even more gain and then losing it with error correction. do it right from the beginning and you don’t need to fix it afterwards.

you can see an open loop op amp (no criticism of it… it is typical) above. compare with the open loop sweep of the EH12AY7 mic pre below…

the gain is MUCH less in the tube gadget, but the bandwidth isn’t bad at all. the S/N for a 4558 or TL074 diff with the same gain and a 6.8K input impedance isn’t much different! it’s still 70dB roughly, for a 3mV RMS input signal.

ok, so the fourth area of improvement would be the noise and distortion spectrum, which ain’t bad already… this does imply one important last improvement: the power supply. these five areas are where i will focus my energy. below is a simplified starting point. i have removed the phantom supply, bass cut, phase reverse and monitior sidechain temporarily, just for visual simplicity. next installment will explain what has been changed and why.