A Tube Screamer experiment

[Mark Hammer]

Here's a question for you kids.

The difference between the TS-9 and the TS-808 lies primarily in the output buffer stage. The TS-808 would appear to provide more loading of the signal, with the consequence being that it "warms it" a bit. But loading is often a function of the cable length that must be overcome. How different do the 9 and 808 sound when using cables of different lengths, and especially very different capacitance? Is it the sort of reputed difference that simply disappears the moment you stick a buffer stage of some sort between the TS and the amp?

I'm curious. If any of you have experience with this sort of thing, please report back.

[R.G.]

I have a short cable, a soldering iron, and burned fingers, but alas no TS-9 or 808 to experiment with. 

[EBK]

Is this a bit like comparing two different energy drinks in the same way that one would compare two different wines? 

I'm also unable to gather the supplies for the experiment. 

[Mark Hammer]

I have a short cable, a soldering iron, and burned fingers, but alas no TS-9 or 808 to experiment with. 

So you're the roadie, and the guys in the van with the gear drove off and forgot you?

And just a word or two of advice, buddy: you're supposed to use solder.  Burned flesh certainly smells impressive, but it makes for lousy connections. 

[R.G.]

Dang! There was a page torn out of the manual, and I bet there were some important items on that page! 

[ElectricDruid]

Is this a bit like comparing two different energy drinks in the same way that one would compare two different wines? 

I'm also unable to gather the supplies for the experiment. 

I've drunk two different energy drinks and two different bottles of wine, and now I'm wired and rdunk and now I don;'t even care if there's any difference between the two pedals. Does that help?

[nickcordle]

I got hung up on this stuff about the TS output buffer and took some measurements a while back.  I never managed to hear / measure anything around the cable length, but the output shunt resistor does something pretty glaring, at least on the breadboard.  The signal has to be HOT to see it.  I was running 1 kHz tones through both 100R/10k and 470R/100k buffers.  Blue is input (AC coupled on the scope), red is output.

470R/100k:



100R/10k:



... yeah.  I'm gonna need to know what that's about ... so at some point I was reading RG's "Technology of ..." and was struck / confused by this bit:

"An emitter follower with a resistor load is often assumed to have a very low output impedance, which is true, as long as the signal is going up. However, when the signal is going down, the transistor can only turn off, and the signal is pulled down by the net loading, DC and AC on the emitter."

Okay so let's assume the falling part of this wave is descending fast / far enough that Vbe drops and the channel somewhat shuts down, is that at least kind of what we're saying?  So if I redraw the circuit a little bit, treating the channel as an open circuit:



Now wait, we still have a circuit, and there's still current flowing at t=0+ after the channel shuts off, because of all the charge stored in the 10u electro, right?  So what's the voltage across the cap at t=0+?  Well, it's the output coupling, so based on the DC points of the buffer, should be roughly (Vcc/2) - Vbe.  Let's call it 3.8V (assuming Vcc=9V)?

Redraw this again:



So this is my idea in a nutshell: on the negative side of a hot signal, we wind up clipped in the neighborhood of -2V because the output ends up on the low side of a 10.1k/10k voltage divider with ground in the middle.  10.1k because 10k shunt + 100R series, and the other 10k is the emitter resistor.

Okay, so now if we switch the resistors back to 470R/100k and draw that divider again:



Now it's the same 3.8V but the divider is ~100k/10k, with the 100k on the lower side of ground, so we shouldn't see clipping until (3.8*(10k/110k)) - 3.8 = -3.45V, roughly.  So if I go back and turn up my signal generator a bit hotter with 470R/100k, what do I see:



Yeah, seems to agree.

One more bit: if what we're seeing here is essentially a divider on the stored charge in the cap after the channel shuts off ... that charge leaving the cap should eventually bring the output signal back up toward ground, right?  It's just very slow because of that huge 10u cap.  But if we replace it with a much smaller value, should we see the clipped part of the wave drift up (faster because of the lower capacitance)?  Tried that too, 100R/10k, with 47n instead of 10u:



Yeah, now it's fast enough that we can see it move within the period of the 1 kHz signal.

So at the moment I want to put the entire difference on the output shunt R.  A little compression / asymmetry / even harmonics where they might not otherwise be ... this is what we typically mean when we say "warm", right?  And it would only happen on pretty hot signals, which lines up with the fact that people say it's a subtle difference, sometimes inaudibly so.

Anyway that's my best idea about TS9/808 differences.  Thoughts?  LMK if I'm barking up the wrong tree, I'm here to learn ...

N

edit: i said "descending side" where i meant to say "negative side"

[Rixen]

Dang! There was a page torn out of the manual, and I bet there were some important items on that page! 

[Rob Strand]

Anyway that's my best idea about TS9/808 differences.  Thoughts?  LMK if I'm barking up the wrong tree, I'm here to learn ...

FWIW,  it makes a lot of sense to me.   It clearly has an effect and the effect can be seen with the normal voltages in the ckt.

The thing that makes it more obscure is you would have to have the level control advanced before you saw the effect.   Is that when people see a difference?  I don't know.    The clipping diodes are going to prevent the level getting too high so it also needs a hot guitar signal before the clipper output can get to a high level, or, perhaps the tone control also needs to be boosting a bit.   So there's lots of conditions which can affect the result.

The TS808 buffer will also behave slightly different when a capacitive load is present.  However, I don't think that's the issue.

What further complicates things is do we need the input tube of the following tube amp to clip ?   That could forward bias the grid diode and charge up the cap.   This case is complicated further by the fact the BE junction could be forward biased and then the 100nF cap on the base could start charging up.

If you repeat your tests/sim driving into a 1N4148 diode you can see the waveform is somewhat different between the two cases.

What does come-up is the clipper mod with two diodes in one direction and one diode in the other would sound different depending which way the diodes were pointing.

Edit:
- If you plot Vbe of the transistor you can see the Vbe swing is much higher on the TS808 case.

- If you bias the base of the TS9 to a lower voltage, say 2.5V, you can get *similar* waveforms.

[nickcordle]

The thing that makes it more obscure is you would have to have the level control advanced before you saw the effect.   Is that when people see a difference?

Yeah, it's a fairly specific scenario, sure, but maybe a pretty common one?

I've always used a TS-style boost in front of an amp with drive ~9:00, tone 12-2ish depending, and level all the way up, or very close.  This into a Marshall or 5150 or whatever.  I only know my own world (metal) but it is extremely common to see it used that way in that music.  It's the slight boost and the mid hump in that scenario, the clipping isn't really the main thing.

Also yeah the NFB diodes earlier will keep things within a diode drop of the input ... but that input signal can be a lot bigger than I used to imagine ... I can pretty easily generate 8V PtP swings on pick transients with thick strings, tuning down a couple steps, digging in on palm mutes ... don't really have to do anything out of the ordinary to see that.  That's on something like Duncan PATB or JB pickups.  Not even considered hot pickups these days.  (Oh and the 7-8 strings now ... eeeeek)

- If you plot Vbe of the transistor you can see the Vbe swing is much higher on the TS808 case.

I want to say I tried to measure this aspect of it at one point, don't remember how far I got.  I do recall the issue that the instant you stick a meter or scope probe directly on the base, you've now screwed up the base bias by giving it a new DC path to ground, so everything starts to tell lies.  Maybe with coffee in the morning I'll remember more ...

[Ben N]

I've drunk two different energy drinks and two different bottles of wine, and now I'm wired and rdunk and now I don;'t even care if there's any difference between the two pedals.

You certainly are rdunk.

[Rob Strand]

Yeah, it's a fairly specific scenario, sure, but maybe a pretty common one?

I've always used a TS-style boost in front of an amp with drive ~9:00, tone 12-2ish depending, and level all the way up, or very close.  This into a Marshall or 5150 or whatever.  I only know my own world (metal) but it is extremely common to see it used that way in that music.  It's the slight boost and the mid hump in that scenario, the clipping isn't really the main thing.

I see a lot of people setting the Tone at 12:00 and Level at the high-ish settings (often maxed out).
It's definitely a scenario where the effect you outlined could come into play.    I'm pretty sure not everyone can hear the difference and a lot of dudes are happy with just a TS-9.

I'd stick with your explanation until someone can can come-up with something better.

I want to say I tried to measure this aspect of it at one point, don't remember how far I got.  I do recall the issue that the instant you stick a meter or scope probe directly on the base, you've now screwed up the base bias by giving it a new DC path to ground, so everything starts to tell lies. 

Unfortunately that's life.  You can use a differential probe or two probes (set to x10) then subtract the result.

[Mark Hammer]

I will reiterate my original question, which is whether any audible differences between the two different output-buffer arrangements might arise, or perhaps disappear, with different cables, and especially cable lengths, and the cumulative capacitance that comes from that.

I appreciate Rob's measurements at the output, but that still leaves the question of whether all of that is still in effect at the end of 25ft of 49pf/ft cable.

[Rob Strand]

I appreciate Rob's measurements at the output, but that still leaves the question of whether all of that is still in effect at the end of 25ft of 49pf/ft cable.

Actually Nick did the measurements.

For a 6m (20ft) cable I've got doubts it would make a difference but without doing the experiment it's only speculation.

The real question is perhaps what changes when a buffer is added between the TS9 and the amp.  The capacitive loading is only one thing that gets decoupled from the output of the TS-9 when a buffer is present.  When you overdrive a tube stage the grid diode can conduct and that forces a strong interaction between the TS9 and the amp.    I'm suspecting that mechanism explains the loading effect.

When a buffer is added the interaction between the amp and the TS9 is lost.  However it's still not that simple.
One detail is what type of buffer?  Opamp or transistor?  Opamps are likely to truely buffer the signal and produce a different interaction with the amp.  But what if two TS9 buffers were cascaded would that lose something or is the interaction still present and just transferred to the second buffer? 

A different experiment would be to place a long cable between TS9 and the buffer.  That would keep the capacitive loading but still have the buffer between the TS9 and the amp.  No change in behaviour would eliminate the capacitive loading theory.  Again what type of buffer?

You can burn-up a lot of hours trying to answer those questions by experiment.  If you wanted to do TS9 vs TS808 that's blows things out further and you would need a set-up where the difference between TS9 and TS808 is obvious.

[Mark Hammer]

That's what I'm wondering: what sort of setup does make their putative differences obvious?  That accompanies the inverse: what sort of setup makes any potential differences simply disappear?

I'm not dismissing any impact of the differences in output buffer between the two out of hand.  But like a great many legends that survive on the internet, things get "tested" under a single very specific set of conditions, and it turns into a considerably more generalized assertion than might be warranted.  We see this all the time with different issues of a given pedal.  Somebody takes ONE pedal of a particular issue and compares it against ONE pedal of a different issue, completely neglects any component tolerances, and in a flash there is an assumption that every single pedal of issue X and Y will display those identical characteristics.  And nobody ever takes the time to obtain multiple pedals of a particular issue, identify their common tone, and do the same for the other issue. 

In other words, there are legends in abundance on the internet that come out of very weak tests.  So my curiosity was aroused by the TS-9/TS-808/Nutube comparison that Brian Wampler posted on Youtube earlier this week.  The Nutube has a somewhat different design by virtue of the different sort of semiconductor used.  But technically, the 9 and 808, apart from component tolerances, are the same design, but with a different output buffer.   It got me wondering whether the received wisdom about the merits of the 808 were generally true, or merely true for a particular set of conditions.

I have no skin in the game, and am perfectly content with whatever outcome can be agreed on.

[Rob Strand]

That's what I'm wondering: what sort of setup does make their putative differences obvious?  That accompanies the inverse: what sort of setup makes any potential differences simply disappear?

I like to idea of proposing a theory then confirming it or dispelling it.   Based on Nick's results the main factor is the level coming out of the unit.   So tests should be done at high and low level settings.  It could even depend on the amp's first stage being overdriven or not.

I'm not dismissing any impact of the differences in output buffer between the two out of hand.  But like a great many legends that survive on the internet, things get "tested" under a single very specific set of conditions, and it turns into a considerably more generalized assertion than might be warranted.  We see this all the time with different issues of a given pedal.  Somebody takes ONE pedal of a particular issue and compares it against ONE pedal of a different issue, completely neglects any component tolerances, and in a flash there is an assumption that every single pedal of issue X and Y will display those identical characteristics.  And nobody ever takes the time to obtain multiple pedals of a particular issue, identify their common tone, and do the same for the other issue.

Agree 100%.  Audio testing is a tricky business.   It is extremely difficult to compare things equally.   Then on top of that is all the generalizations you mentioned.   If you take two different pedals how can you know one has a control set at 2 O'Clock  like it could be anywhere from 1:30 to 2:30.  On top of that add the tolerance of the pots (especially tapered pots) and you already have a good opportunity to make things different from the start.   Like if I build a TS9 clone with a linear Tone pot and set it to 2:00 then compare that against a real unit with the G-taper pot at 2:00 there's a built-in difference (the same problem can occur when comparing boutique clones).    A nasty test bias is to increase the level of one unit by 0.5dB, something around the point were you can't quite tell it is louder.   Even with AB testing you will get a significant positive result for the louder unit.  People cannot adjust pots to that accuracy!   When I did lot of AB testing I had a switch which switched between two configurations keeping as much common as possible.   For better/worse tests a good method is to record samples then present them to the listener randomly [edit: AB might be better for detecting small differences].  The statistics will show what is better but you need a large number of samples.

In other words, there are legends in abundance on the internet that come out of very weak tests.

Yes, from a scientific perspective it isn't great.   If you had several thousand tests and took the average you might be able to see the true result from the noisy data.

I have no skin in the game, and am perfectly content with whatever outcome can be agreed on.

Same here.  It's nice to know and nice to know why but at the end of the day I've paid enough dues doing audio testing.

On the funny side I saw a few videos comparing tube amps and Peavey transtube amps.  Interesting how it's not so easy to tell.  As far as generalizing goes, the tests were limited to type of stuff they played.

[nickcordle]

Maybe I should qualify my statements above: here's one difference between TS9/808 output buffers, which arises in a particular scenario, which AFAIK should be mostly insensitive to tolerances ... that circuit fragment / behavior shouldn't move much with anything mid-high hFE (Ib & Vref will change but not by a ton) ...  ?

Didn't mean to be overly conclusive about it

But from there we could ask a more specific question, though: what set of transmission line characteristics / terminating impedances might exaggerate the difference, vs. which might tend to tamp it down, yeah?  But someone else is gonna have to get at that analytically because I barely ever wrapped my head around TLines and I don't know tube stages well enough to guess.

N