Tube Tone

[Alex/Tubewonder]

What is it they say about the duck? If it walks like a duck, quacks like a duck and looks like a duck than it is a duck.
Same can be said about tubes but simpler. If it tests and measures all parameters within limits specified by original specs than it is what the specs say it is.
I agree that current production tubes are all over the place spec wise and there have never been so many variants with the same generic designation 12AX7 or ECC83 + new letter combinations slapped on them. Only a few pass the eye of the needle in rigorous test. Some of the failed ones don't even fit within the 12AX7 specs. It's engineering, has to be cut and dry. Sound is too emotionally charged to be exactly specified. Two synesthetics will describe same sound in different ways, one using smells the other colors as descriptors. Translate that to technical specs. EOR :-)

[Enzo]

If it measures within the 12AX7 parameters, then it is a 12AX7. Fair enough. But just how tight are those specs? Shirts come in small, medium, and large, and shirts may fall into the large category, but that leaves a range of size that large encompasses. I think about those 20% resistors Leo used to build his amps with. That 100k resistor could be anywhere from 80k to 120k and be within spec. 12AX7 specs I'd wager include a lot of space, enough to include a lot of variation.

Even 40 years ago, when RCA made great tubes, I bought matched pairs of 7027 for my Ampeg. Said 7027MP on the boxes. They recognized even in those old glory days of tubes that 6L6 or 7027 spec left a lot of room for variance.

Of course that allows us to argue that perhaps the differences between tubes of any one brand could be greater than the differences between two brands. But if for example one brand of those resistors tended to be over 100k, and another brand was consistently lower than 100k, then aside from individual part variations, we'd also have brand tendencies. I can see the possibility that one brand of 12AX7 due to its construction and materials might consistently trend towards one end of the specs while another brand went the other way.

Engineering is amazing, my volt meter reads to 0.3% accuracy or some such. Good specs. Nothing like the specs on those 10-digit Agilent meters with accuracy I can;t even spell. Both are volt meters, neither spec is wrong. Cut and dried has to have a context.

A lot of parts have specs, but the specs do not cover every possible parameter. A tube might spec a gain range, a voltage range, an emission spec for the cathode coating, etc. I don;t know all the specs certainly. Is ambient temperature specified? Inter-element capacitance might be spec'd, but does that include to the heater? How about the hardness of the vacuum? I bet there are a lot of things not specified in the engineering description of that tube.

Then again I could be wrong, this seems to be the day for it.

[olddawg]

Then there's microphonics and resonant frequencies of bottles and such. That in itself can cause a change in tone just like a guitar body. Shape, glass, metal all factors. As are the shape size and materials of the other individual parts in a tube. The nice glassy sound of tubes is one of the things that they can't seem to emulate with an algorithm. And it's basically a construction artifact.

[Steve Conner]

Alex, the precise distortion behaviour of tubes was never specified in the datasheet. A tube could have any distortion spectrum whatsoever and still be a "12AX7". The first few harmonics are probably nailed down by tube physics, but the higher order ones that add "crunch" to distorted tones will depend on the fine details of construction, like the uniformity of grid wire spacing, position of support rods and so on. (IMO. Contains original research, as Wikipedia would say.)

Olddawg, someone on the forum has a signature that says: "It's not microphonics, it's undocumented reverb." I wouldn't be surprised if microphonics were part of tube amp sound. DSP programmers could probably emulate them using a modified reverb algorithm, if they weren't stuck in the mindset of microphonics being "one of the reasons we're not using tubes."

[Steve Conner]

Alex, the precise distortion behaviour of tubes was never specified in the datasheet. A tube could have any distortion spectrum whatsoever and still be a "12AX7". The first few harmonics are probably nailed down by tube physics, but the higher order ones that add "crunch" to distorted tones will depend on the fine details of construction, like the uniformity of grid wire spacing, position of support rods and so on. (IMO. Contains original research, as Wikipedia would say.)

Olddawg, someone on the forum has a signature that says: "It's not microphonics, it's undocumented reverb." I wouldn't be surprised if microphonics were part of tube amp sound. DSP programmers could probably emulate them using a modified reverb algorithm, if they weren't stuck in the mindset of microphonics being "one of the reasons we're not using tubes."

[Alex/Tubewonder]

Steve, what about changing the slope of dynamic load line to shape distortion characteristics? And not to forget that tube parameters like transconductance are affected by the operating conditions. Check original Tesla (not JJ) datasheets, pages 13-18, these diagrams say a lot of interesting things. http://www.mif.pg.gda.pl/homepages/f...83/e/ECC83.pdf
As for the parameter spread, even though it's not specified in standard datasheets, you can get a pretty good idea by looking at column II in the Philips SQ sheets. http://www.mif.pg.gda.pl/homepages/f...09/e/E83CC.pdf
We have here an illustration of "everything is connected": tube affects the circuit and circuit affects tube parameters. The end result can be a mediocre sounding amp or close to magical. It's like cooking really, you need to know your tools and how ingredients interact with each other. Sometimes the sauce needs just 3 more grains of salt, sometimes you need to start over. OK, I'm watching too many cooking shows....

[exclamationmark]

You'd be surprised at how different the curve traces can be for real life tubes. I've seen some real life curves of 12AX7's, and whilst you could vaguely guess they were 12AX7's by looking at the small signal gain figures, but by directly overlaying it over your usual curves, you would think it was another tube entirely. One of the major differences I've found is that the Vg = 0 line can sometimes be way steeper in some tubes compared to others. Depending on the load line, you can get an extra 10 to 20 volts peak when driving it hard.

I think there is also an underestimation of how complex the clipping behaviour of tubes is - as far as I know there are no models that describe this behaviour (apart from 3D field solvers, which can take the processing power of a desktop PC to a supercomputer to compute). I've worked on some tube amplifier modelling plugins, and while it's trivial to emulate different grid curves, the same cannot be said about clipping. The best we have so far are empirical solutions - no one really knows how to describe the island effect, or the asymmetry found in grid current limiting via physically based models. For example, in real tubes different parts of the cathode and grid wires go into cut-off at different times! Thus even the grid winding type or density, can affect the characteristics of the tube, even if it meets the cathode-grid spacing criteria for that particular tube. Spangenberg is a good read if you really want to get into the details of how tubes work.

IMO the point that you should re-bias each individual tube to the same operating condition is a moot one. I would sort of lump this fact into the reasons why different tubes can have different sonic characteristics in the same circuit. However, I completely agree that the tube and circuit work in tandem to achieve the desired result. Although since the tubes are replaceable, and the circuit usually isn't, sometimes replacing the tubes can prove to be the 'glass slipper' for some circuits. I'm not into tube rolling myself, but I do recognize, like Chuck has said, there is often a difference in tone when tubes are replaced.

[Steve Conner]

I agree, a 3D field solver would be the way to go. Pretty sure Eimac, Svetlana, Varian et al will use one in the design process for their latest RF tubes. They are certainly used to design CRT and X-ray tubes.

Or use an actual tube, measure its curves right into cutoff and saturation with high precision, then put them into your tube plugin as a bigass lookup table with spline interpolation. If I were doing it, that's how I'd do it, except I think Cliff Chase of Fractal Audio has done it already.

To emulate clipping, you have to emulate grid current and the bias shift that comes from it. You end up running more or less a realtime Spice simulation of the amp you're emulating, with the tube model a "piecewise linear" component in the middle of it all. Multirate DSP techniques can help with the workload, because the bias shift is slow compared to the audio signal. Again, I'm pretty sure Cliff Chase has done it already.

[kleuck]

Of course that allows us to argue that perhaps the differences between tubes of any one brand could be greater than the differences between two brands.

Having done a lot of tube roll, and owning a lot of (OLD) tubes from my preferred models (ie TFK smoothplates/EI smoothpklates, Tung-Sol Long plates, Matsushitas) , but from different sources, some NOS, some slightly used and some who died in my amps, i am absolutely sure it's not the case. They consistently sound the same in the same circuit, great in some amps, not so great in others, but the same.

[exclamationmark]

I agree, a 3D field solver would be the way to go. Pretty sure Eimac, Svetlana, Varian et al will use one in the design process for their latest RF tubes. They are certainly used to design CRT and X-ray tubes.

Or use an actual tube, measure its curves right into cutoff and saturation with high precision, then put them into your tube plugin as a bigass lookup table with spline interpolation. If I were doing it, that's how I'd do it, except I think Cliff Chase of Fractal Audio has done it already.

To emulate clipping, you have to emulate grid current and the bias shift that comes from it. You end up running more or less a realtime Spice simulation of the amp you're emulating, with the tube model a "piecewise linear" component in the middle of it all. Multirate DSP techniques can help with the workload, because the bias shift is slow compared to the audio signal. Again, I'm pretty sure Cliff Chase has done it already.

I've actually tried doing most of this - I think I ended up with like a 50+ MB look up table - I've even thought of stretching it another dimension and getting the grid current, too. One of my minor problems with it, is that it assumes the tube doesn't care about it's past behaviour, which means that it doesn't replicate the asymmetrical grid current clipping thing - I'm not sure if it's just the internal capacitances causing it, or if it's a function of something else. It probably doesn't affect the sound TOO much, but it annoys me anyway. The giant tube LUT is then plugged into my own spice derivative (or should I say, abomination - I'm no Cliff Chase, that's for sure!), which handles all the other crap. The other rather frustrating thing, is that the waveforms still appear as if they've been drawn by someone practising convolution or step functions, and they don't really resemble a real scoped signal all that much. For example, when you have a small (1nF or so) coupling capacitor, and you clip the bejesus out of the stage with a lowish (100Hz ish) signal, you expect to see distortion. Viewed on a scope, it's really quite ugly. But view it on any spice program, and it almost looks too clean. Then there's other issues that come along with signal processing and limited sample rates (as well as optimizing the damn thing to run in real-time!). At this point the project has been abandoned for at least a year, but it gives some insight into how complex modelling real life circuits is, at least for musical purposes.

[Steve Conner]

You're right, the tube doesn't care a bit about its past behaviour. It's the coupling capacitors, charged by grid current, that form the "memory".

[Chuck H]

To get back to the Q in the OP... I'll conceed that an amplifier that meets spec, and therefor agrees with it's designation, SHOULD sound the same as any other. But a whole bunch of issues have been raised to fail this ideal. If, for example, you plug a 12ax7 into a mono block hi fi unit, that is using local and global NFB to linearize the performance and operaes the tube well within it's amplification capabilities, then you probably won't hear much difference between tubes. Now let's consider guitar amps. Where there is far less linearization from the circuit and the tube may be asked to jump in and out of it's ideal operating parameters. Hell, let's even overdrive the piss out it with a cold bias and no NFB whatever! Now the small performance differences betwen tubes starts to stack up. And that's enough about all things being created equal for operating parameters. The spirit of the original question, to me, is: "If you just swap tubes in a guitar amp, will the tone change?" The answer is ABSOLUTELY. Whether because different tubes DO sound different or because they operate different is pedantic. Different tubes, simply swapped in, do sound different. And, IMHE, this is especially true in guitar amps.

[Alex/Tubewonder]

What happened to paper, pencil, a ruler and designing amp stages graphically, even cascading 2 or more stages? You can even do distortion analysis "good enough for R&R" with these graphic methods. I do it most of the time and even run Spice. But when Spice results look too funky, I go back to graphics on paper. It works fine for me most of the time...
Anyway, I see a stepper motor driven nutcracker controlled remotely via web interface under your Christmas tree this year. Just close the FB loop so it won't run amok and do a job on your toes.
PS. It's OK and fun to investigate a bunch of tubes to your hearts content when you have bought an amp, or built one for your own enjoyment. It's a different story when you run a business manufacturing amps to sell. I need my amps to sound just like the master unit - that has been tweaked during R&D and prototyping - without further tweaking. Unless it's a custom job involving individual options, but that's another story.
P+L+U.

[Enzo]

Or design the amp so that it is not hyper sensitive to which tube is stuffed.


I would wonder among all the 12AX7 specs, I don;t recall seeing specs for overdriven abused tubes. RCA certainly never planned for such use. Once we get outside the designed parameters, ther are no guarantees how a tube will behave.

[Chuck H]

What happened to paper, pencil, a ruler and designing amp stages graphically, even cascading 2 or more stages? You can even do distortion analysis "good enough for R&R" with these graphic methods. I do it most of the time and even run Spice. But when Spice results look too funky, I go back to graphics on paper. It works fine for me most of the time...

Agree. I don't do it for three reasons. The first is that I don't usually want the stage to be entirely linear. And I like a little chaos too. Dynamics that can do more than just have a loud peak attack, etc. The secnd reason is that it's not hard to run tubes within "safe" parameters, ignoring ideals. If every stage is center biased with every care taken to reduce any odd distortions your guitar amps distortion character can sound blatty and flat. The third reason is that I can't seem to get my head around the way tube charts are drawn. And IMHO all the on line tutorials suck. So when I do need a linear stage, like an effects return, mixer, reverb recovery, etc. I just use my scope to determine center bias. I'm never over voltage spec anyway.

It's a different story when you run a business manufacturing amps to sell. I need my amps to sound just like the master unit - that has been tweaked during R&D and prototyping - without further tweaking.

Unless you use the same make tubes for every position in every unit that's just not possible WRT guitar amps. But you didn't say that you make guitar amps. But like I said before, I've never adjusted operating conditions for different brand 12ax7's to be certain.

My band (from long ago) use to cover Elvis Costello's version of What's so funny bout'... That was one of my personal favorites to play.

P+L+U X2