More ramblings written in 1998
This is the second installment of the Manley Labs point of view. The first one dealt with tube sound and a few common tubular misconceptions.
This one attempts to explain some practical details about distortion, with a bit more on tubes and transformers. If you read the pro magazines you might get the impression that all it takes is distortion to get that "warmth" everybody must be asking for. Distortion - You don't have to be a genius or expert to know that there are many, many flavors of distortion. There may well be two to ten characteristic types of distortion for every active audio toy out there. Clipping is one aspect, but long before clipping there is usually low levels of THD and there may be particular distortions in the lows and highs (before you turn any knobs). If you are convinced warmth = distortion, you may not want to read further, but you should. We attempt to describe the useful and not so useful kinds. For this discussion, lets define distortion as "a change in the shape of a sine wave" going through a device. These changes tend to add harmonics and non-harmonic frequencies to the pure sine wave. Why define it this way? There are tech heads out there who will say compressors and EQs by definition are distortion - but we usually use compressors to avoid distortion and we don't think most people would say EQ sounds distorted. Lets try a common sense approach.
1) Total Harmonic Distortion (THD). For our purposes - let this describe the intrinsic non-linearity of every simple gain stage, tube, transistor or FET at the nominal operating level. Usually more gain introduces more distortion. A tube or FET with 30 dB of gain typically produces some second and/or a little third harmonic distortion in the neighborhood of .1% or -60dB. Typical THD specs are convenient measurements but not particularly descriptive sound-wise. It lumps together even and odd harmonics, lower and upper harmonics, and even clipping can be a sub-set of THD. Three different devices with the same THD spec will very likely sound different. It is the shape or curve (deviation from a straight line) of the non-linearity that creates this difference. The curve means that different signal levels get different amounts of added harmonics. It would take a spectrum analysis of the distortion at several levels and frequencies to provide meaningful criteria on what to expect sound-wise but all this detailed info would be pretty hard to digest. "More than you wanted to know". There are ways to design for better linearity and ways to make it look good after the fact (negative feedback or cancellation techniques). Most modern designs aim to lower THD but it might be remembered that between .3% and 1% is what the old research indicates is the threshold of audibility. We think the number might be much lower, but very much depends on the harmonic number and other factors (above the 4th harmonic, distortion becomes significantly more audible). Obviously, the audibility very much depends on the listener the original signal and the quality of the monitor system. 2nd & 4th harmonics (thickness) are normal for class A, 3rd &5th harmonics (aggressiveness) for push/pull & transformers. Upper harmonics (harshness) are likely in high feedback circuits.
2) Clipping. The essential fact is that a device has both a THD characteristic and a clipping characteristic and in practical use they are different and creatively used in different ways. Practical experience shows us that the sound of clipping depends on the initial sound, the unit, and how hard it is driven. The sound of a cranked Marshall stack is a lot different from an overdriven mic or from an A to D converter & digital FS saturation. It is not as simple as "hard or soft" clipping either. The Marshall is a study in a whole chain of stages clipping and various EQ curves between stages and several other factors from power supply sag to cheapest possible parts count.
3) IM, TIM (Transient Inter Modulation) etc. Usually bad words. This is the non-harmonic distortions but they also always accompany THD and clipping. If there are good harmonic distortion and bad distortion then there has to be acceptable IM and awful IM. It is easy to measure and hear IM but usually it is so closely related to THD that it might as well be lumped into the vague THD spec. TIM seems to affect the highs when sudden wave form changes like transients or square waves are involved. Slew rate and feedback are the culprits. Slew rate distortion is real but practically non-existent in music.
4) Zero crossing or crossover distortion. Some circuits use separate circuits for the positive and negative halves of the wave form (for efficiency). This distortion is common in the field (with long cables) but inconspicuous on the test bench. Unfortunately crossover distortion is rarely appealing. A good reason for Class A circuits.
5) Complex distortions. We'll use this to describe where the THD, IM and clipping and crossover distortion interact with frequency, compression and the other distortions. Mics, transformers, guitar amps, speakers, analog tape, vintage gear, crunchy compressors, and most real world gear falls into this category. This is a particularly deep topic and where most of the confusion resides. The good news is that we can easily hear and categorize or describe the sound - the bad news is that spec sheets will probably never be adequate.
Even a "simple" passive component like a transformer falls into the category of "complex distortion". Saturation first. In a reasonable transformer the saturation is dependent on signal level (power) and frequency - The lower the freq, the easier it will saturate. The saturation itself is mostly a function of the magnetic core size, core material and number of turns on the primary windings as well as the circuitry around the transformer. We believe that the biggest part of reason people describe vintage gear as warm sounding is due to the transformer saturation. Interesting statement from a company building tube gear. Consider vintage discrete Neve, UREI and API products - all solid state - never heard anyone say they were lacking warmth. This low frequency saturation typically adds odd harmonics to the lowest octave(s). We perceive the pitch of a note when the fundamental frequency is entirely removed as long as there are a few harmonics. The low octave is often not reproduced by the speakers or rooms. If one inserts a good transformer in a circuit it will still measure "flat" down to 20 Hz but to our ears the bottom has become bigger, fatter, richer and yes, warmer. However, with a cheesy transformer or misapplication, the effect may be quite different and the saturation may not be subtle - it can sound like plain ol' ugly distortion. Some folks question other aspects of transformer distortion. Long before saturation there might be a low level of THD. There is some hysteresis involved with making the core change its magnetic domains from north to south. How low, again depends on a number of factors, but a well designed transformer is cleaner than a tube or transistor. It's the cheap and cheesy ones that are mostly to blame for the bad rap. Another problem is that the driving circuit may distort because of the (quality of the) transformer. In the highs, the transformer may present or reflect capacitance to the circuit at the square of the turns ratio. This can act like a low resistance in the highs, and for the circuit, its like driving a short circuit. A similar thing happens in the lows due to the basic inductance. And yet another distortion can occur due to reactive imbalances if it is driven push/pull. The bad news is that all of these are likely when lack of transformer knowledge and/or budget transformers are part of the design. The good news is that only the most pleasant aspects of saturation are likely to be audible in the best gear. It is mainly why analog tape, tube and vintage solid state are called warm and why budget and digital gear isn't.
There are frequency response concerns with transformers too. A real transformer has a number of basic resistances (the DCR of the windings), capacitances (distributed in the windings and winding to winding) and inductances (including leakage inductance). This makes for a (wide) band pass response. But just as important, the transformer is typically the connection point to the outside world and that means that other R's, C's and L's may be the wild card. An input transformer can be forced bad, if the wrong device is driving it. The source impedance can affect a high frequency resonance peak or roll-off. One interesting exercise is imagining a "perfect" theoretical input step-up transformer. If it had a 100:1 turns ratio (40 dB voltage gain), any capacitance on the secondary would reflect as 10,000 times that C back on the primary. This combined with the usual source impedance creates a LP filter. And this is a theoretically perfect lossless input transformer! This is the main reason why high step-up ratios are impractical. With an output transformer, the load directly affects the power through the transformer and thus, the onset of saturation. It is possible to have these effects minimized or far out of the audio band but it is not easy.
While we are on this topic there are some benefits of transformers virtually impossible with any other technology. As an connection to the outside world, you can use a transformer as a floating interface. This means it is not connecting grounds - no ground loops. It is equally happy driving balanced or unbalanced (unlike so-call transformer-like circuits which lose headroom and often stability). A typical transformer has significantly better common mode rejection across the whole audio band than almost all op-amps (better hum and noise rejection). It also becomes a safety device, RF protection, DC protection and "impedance matcher" or "distortion canceller". The biggest downside is the cost, weight and did we mention cost? This is the real reason transformers get dissed rather than used more often. At $50 to $500 per transformer one can find a lot of excuses for not using them.
How about that tube distortion versus transistor distortion thing? You might want to re-read the first "Tube Talk". Ya can't quite generalize it that easy. For example: What about most of these new tube products that claim to be "warm" (repeatedly), and certainly distort, but are virtually unusable for anything professional? Sometimes it seems the more often the "W" word is used in ads, the more likely it is meant to exploit those with less experience, knowledge and budget. Kinda sad really.
Unfortunately a real discussion is not as simple as tubes versus transistors - it mostly depends on how they are used and the overall circuit. Believe it or not, tubes are cleaner than transistors. It is accepted practice to use gobs of negative feedback to make a transistor circuit clean enough to pass audio. The negative feedback is like a correction circuit that attempts to remove any difference between the input and output. It improves all the basic printed specs from noise to output impedance. Three subtle problems though. First, the transition from "clean" to clipping becomes sudden and this sounds nasty and unnatural and definitely electronic. Second, feedback can decrease the second and third harmonic but increase the 6th and above (oops). Third, we suspect there are transient inaccuracies - imaging collapses, transients smear and esses become ugly. Putting a tube in a circuit like this will not fix these problems. If a little THD or soft clipping is all that is desired - they could do it with a FET or transistor, but that poses a problem for the marketing and advertising execs. See - it has to be a tube because tubes are warm. A ten year old kid would know you're joking if you told him " It sounds warm because these glowing glass things are warm". Aren't you glad the buzz word isn't "soft". They could be pitching "made with goose feathers" next.
We might hint that there can be an optimum balance between the ideal of minimal THD and acceptable or cool clipping character and transient accuracy. How? Start off with the absolute cleanest circuit using no wrap around feedback, then if needed, use as little feedback as necessary to achieve lower distortion, higher damping and greater repeatability. Years ago, there was a good analogy that the better audio designers lived by. "Negative Feedback is like paint. You can't fix a bad structure with paint but you might improve the appearance of a good structure with just the right amount - Not too much, not too little." Manley Labs either uses minimal negative feedback (allowed by clean designs) or provides a rotary switch that allows the user to choose the amount depending on their taste. Because the feedback has many effects, it is easy to hear, but difficult to describe. We can also hint that if "warm" distortion is what you think you need, then look ( & listen) to good transformers.
The best sounding recording systems today are a finely chosen blend of tube, discrete solid state, op-amps and digital. The key seems to be avoidance of the cheesy gear and only using the best stuff and having a good balance of technologies. Each has strengths and weaknesses. Too much of any kind results in a predominance of that technology's color. A weak link can break the chain.
Manley has a different definition of warm. There is something warm and friendly about a natural life-like acoustic instrument before it gets recorded. If it doesn't get messed up by the electronics then guess what - you got warm and perhaps you have proven you know how to record. The best way is with a great player and instrument, then use the best Mics, the best position, and a minimalist mic pre straight to tape. If the arrangement works it has to sound great. If you are looking for warmth there is a very good chance you skipped at least one of the above items.
This one seems obvious but a lot of people get blinded by this one. Synths never have been very warm. They are wonderful and amazing but usually cold. Always have been electronic sounding. We like synths and project studios and all the new affordable ways to record and distribute music but this topic of warmth was never an issue until lately. Samplers and looping are too cool - in more ways than one. The good news is that more and more people are using them as a creative writing platform and replacing the parts with real natural instruments and leaving some of the nuances and idiosyncrasies in there. Some heat here. Another very unusual statement from a gear manufacturer, " Maybe more gear isn't the answer, it's the problem". How about better equipment, not more toys? Perhaps better tracks & arrangements rather than more tracks in some cases. If the "old classic sound" is the goal, remember, they didn't have tons of gear and tracks. They used acoustics, mic technique, basic simple gear and played & sang. Not as easy as it sounds - many limitations and good pro gear was expensive.
Manley has been building highly reviewed audiophile equipment, professional recording gear and mastering gear for many years. We use tubes, transistors, FETs, MOSFETS, op-amps, monolithic converters and we wind our own transformers. Each technology has its advantages and disadvantages. To us - its all audio gear - all built to be clean, flat, smooth, transparent and musically fulfilling. The engineers call it warm, the audiophiles call it natural, and the mastering folks say its magnificent. How? No big secret -- we minimize the number of parts the signal goes through (particularly active ones), maximize the quality and performance (almost ignoring the cost), and listen. "Less is more" when it comes to audio. This is why we use tubes - they are cleaner, less fix-up is needed, less electronic tricks. "Make it as simple as possible and no simpler".
Engineers tend to get lost in theory and numbers. Nelson Pass points out "that no audio product has succeeded in the marketplace that has taken some spec to an extreme ". The obvious reason is that all the other concerns tend to be trivialized. Do we hear the difference between .01% distortion and .001%? Are there advantages to damping factors above 20? How about below 20, especially for driving tweeters. Should we be impressed with 24 bit converters if the friggin' interface is incapable of 24 bits? Here is what we aim for - lowest basic THD with minimal wrap around feedback. For us, this ensures the best transient accuracy, best clipping characteristics including fast recovery and minimized upper harmonics. With headroom, we seem to be swimming against the current pro audio trends. We aim for what seems to be excessive headroom and get +30dBv out of the majority of our products. We would not use hybrid circuits. We have all-tube, low output impedance, high current & voltage, stable, Class A line driver circuits proven in years of use. Why would we use an op-amp with half or 1/4 the juice, audible crossover distortion and much, much more internal complexity. There are better ways.
With the transformers, we achieve frequency responses usually from 7 Hz to 40Khz or higher. We accept some saturation but only at extreme levels. Of course, we aim for lowest noise, hum and RF. It is not as low as some solid state but it is low enough to be as quiet as tubes can get (compared to other gear). Mostly this is grounding, circuit lay-out, and application of magnetic shielding. Wherever practical, we use "Class A" circuits so there is no crossover distortion. We have found these techniques equally apply to tubes and solid state. It requires optimized gain or power for a real life applications. In power amps, the optimum power relates to speaker sensitivity (efficiency), the desired listening levels, the room, and those important power hungry low frequencies. While there is some good to be said for the concept of low power single ended amps, for the bulk of audiophiles, it is just not worth the trade-offs. It is possible to have all the good stuff, without it costing as much as a Ferrari (especially one that maxes out at 20 mph). Similar is true in converters. We built our first A to D nearly 8 years ago and it still wins shoot-outs with 24 bit/96K units. Why? Probably because it is simpler & straight through. With the Reference DAC, we were one of the first to tackle the problems of jitter and it remains among the lowest jitter specs to date. The analog stage is all-tube, all class A, simple but effective. It seems to be most important to address the weakest links first and foremost, rather than pushing the limit of the best aspects into the realm of pure vanity.
This is not new - all of it can be found in books and in well known articles in electronic magazines from the 40s, 50s and 60s. Names you might look for are "Williamson", "Sowter", "Crowhurst" , "Partridge" & "Blumlein". These old timers were brilliant. Not to give you the idea that brilliance is hard to find today. There is Nelson Pass, David Blackmer, Michael Gerzon and David Manley. Our work is built on solid foundations.
We hope you enjoyed reading this, and perhaps learned a bit more. We absolutely encourage you to try out our products, check out how these ideas translate into real gear and sounds. In many of them, you can bypass the transformers and adjust the feedback to hear for yourself what differences there might be. It is your opinions that count because its your music. Have you ever noticed there has never been a Manley ad with an endorsement or list of users? It would be an impressive list but would you rather be impressed by your results with the gear or by us flagrantly name-dropping? On the other hand, for those engineers, producers and musicians and reviewers who have been saying good things about us in print or otherwise - we sincerely thank you. You know who you are. Ya' makes us feel warm all over...
Written in April 1998