Spice Analysis of Tube Screamer - Frequency Response

[aziltz]

Sparked by the recent discussions of how the TS Tone Control REALLY works, and all the simplified explanations of it, I decided to model the circuit in 5Spice to try to shed some light on the function of specific components, specifically the RC filters throughout the circuit.

Here's the basic TS circuit, minus input buffers, with a few adjustments to the input output bias resisters so as not to add unwanted filtering.  I used an LT1498 Model because I didn't have a 4558, and for this study we weren't focusing on that.


PART I - Gain Response

First off, you can see 3 test points:
TPv1 - (RED) - Output of Clipping Amp
TPv2 - (GREEN) - Circuit Output (Volume Pot is always at Full)
TPv3 - (BLUE) - Signal after R6/C5 Low-Pass but Before Tone Control Amp.

Lets look at all three Points for Various Gain Settings, using a LINEAR Pot.  Tone Pot is set at Noon, remember this is without Output Attenuation.

First, Minimum Gain.  Full Treble content coming from the Clipping Amp.  Nice Mid-Hump Post R6/C5.  No Effect from Tone Circuit.
I think its most apparent here that the C4/R5 Frequency Cut-Off (720Hz) limits the gain to frequency OVER 720 in the clipping amp.


50% Gain.  Less High Frequency Response from the Clipping Amp.  20dB Louder than Minimum Gain.  Same Mid-Hump Post R6/C5.  No Effect from Tone Circuit


Maximum Gain.  Even Less High Frequency Response from the Clipping Amp.  Additional 5dB louder than 50% Gain.  Same Mid-Hump Post R6/C5.  No Effect from Tone Circuit.  So it looks as if we are seeing the effect of the 51pF Cap across the Gain Control of the Clipping Amp.  I'm not going to make an attempt to explain it entirely, but I believe it is common knowledge that increasing that value of cap to 100pF or 220pF will roll-off even more highs and "soften" the distortion, but the effect is seen most at high gain settings.  Again, I have not gone into details regarding the specific RC Filter/Feedback loop response.


PART II - Tone Control

Next I will fix the Gain setting at 50%, and adjust the Tone Control Pot.  The 4 curves are evenly space settings on the Tone Control Pot, 0, 1/3rd, 2/3rds and Full.  As ycan see the sweep of the Pot does not have a steady affect on the frequency response.  I believe this is why people have used the radical S or W taper Pots as mods instead of Linear Pots.  The effect of the tone control is bunched up at the extremes on a Linear pot, whereas a W Taper allows a more rapid change of resistance near the center, and so the effect seems to have a smoother sweep.  Just a theory, I don't know exactly what the W or S Taper actually does, but I have one in my TS and it works much better!  Its also evident here that the Tone Control shifts the mid hump higher in its upper settings.


PART III - Changing Values in the Tone Section

The Dubious 220 ohm resistor (R9).  In some earlier simulations I did, we discovered that a resister in this configuration (Low Pass Filter) creates a High Shelf in the frequency response.  This shelf levels off the 6dB/Octave created by the original R/C network, at a frequency that appears to be given by 1/(R9*C).  Anyway, in this configuration its a bit more complicated, and I'm not prepared to explain it in words at the moment.  BUT, I simulated the effect of shorting R9 and how it effect the Frequency Response. 

First off, Tone set to 50%.  R9 Does NOTHING!  I believe this is because the Tone control actually has no effect at all in this setting.  If we refer back to PART I, we'll see that there were no significant differences between the signal before and after the Tone Control Amp.  Thus, I did not create a graph for this.  What I did do is simulate the effect of shorting R9 at Minimum and Maximum Tone Settings.

1.  Minimum Tone Setting (Darker Tone from TS, same as lowest curve from PART II)
Here we are seeing a slight DECREASE in Treble Content when R9 is shorted.



B.  Maximum Tone Setting (Brighter, Highest cuver from PART II).
Here the effect is most prominent.  With R9 shorted, we lose a lot of upper harmonic content out of the TS Tone Control.  This is the setting where the RC Filter is applied to the feedback of the Op Amp.  Filtering the feedback would increase the amount of treble content in the output.  Thus, a treble shelf in the feedback filtering would limit the highest treble increase, controlling the highest harmonics (usually considered harsh).  Also notice that the frequency response begins to deviate significantly around 3kHz or so (3.2kHz being the 1/RC value of R9 and C6).


CONCLUSIONS

Its been said that to make a good distortion box, you need to control the EQ/Harmonics between and after gain stages.  Here's a good example of what that means in a classic box.  Personally I had no idea the Mid Hump was so apparent, but remember, these simulations are done with a flat frequency source, whereas an actually guitar has a lot more amplitude in the mids and lows.

From this data I believe we can conclude that the Mid Hump is induced as a combination of the C4/R5 (720Hz) cut-off in the Clipping Amp Gain, and the R6/C5 Low Pass Filter immediately following the clipping stage (cut-off also ~720Hz).

Honestly, this was just me looking for an excuse to try out a new tool.  I hope you find this info useful as I did.

[aron]

Very cool! Thanks for the post!

Aron

[aziltz]

One thing I'd like to add.  If the output seems abnormally large in scale (dB), its because I'm using a 1 Volt input signal.  Using 100mV does not change the curves, only the overall level.  For the purpose of this frequency study, the scale should be considered relative.

[grolschie]

Does this info yield possible new mods to reduce mid hump? 

[MohiZ]

Good work, thanks!

[Projectile]

Does this info yield possible new mods to reduce mid hump? 

Jumper the 220 ohm resistor to ground and the mid hump is gone when the tone control is maxed. You still lose some bass in the gain stage, but the whole tone control section goes almost perfectly FLAT. Try it!  Unfortunately a tubescreamer without the mid-hump sounds pretty harsh in the highs. I still think it's a useful mod though, so I put it on a switch in my build.

That discovery is what led me to start this thread where I calimed that the GeoFX article on the Tubescreamer must be wrong. The arguments on that thread were what led to Aziltz conduct his spice analysis which cleared everything up, and I thank him for that. Apparently nobody was following my original post, which is okay, because most of it is just a lot of pointless arguing and mis-communications. I was wrong about some of my initial assumptions anyway, which were just my noob explanations for what I was seeing on frequency plots that directly contradicted what I was reading. If anyone is more curious about the details, I would suggest just skipping to the 3rd page of that thread.

Anyway, carry on...



 

[Projectile]

Since I never got around to posting these in the other thread, I figure I'll put them up here...




Here is a triangle wave swept over the widest frequency spectrum possible within the parameters of the synthesizer I was using as a tone generator. It represents a flat frequency response for reference:









Here is the isolated tone section of the tube screamer circuit with the tone knob at max:










Now, here is the isolated TS tone stack at max again. The only difference this time is that the ***220ohm resistor is jumpered***:



Notice that the entire tone section goes completely flat! In listening tests, it sounds very similar to a bypass of the tone section entirely. The 1k/.22uf RC filter combination in the feedback loop acts as a boost that just mirrors the cut made by the first 1K/.22uf RC filter, effectively leveling the signal. What the 220ohm resistor does is put a shelf in this boost above 3.2kHz, causing frequencies above 3.2 kHz not to be brought back up to unity from the earlier cut.  So, in the end the 220 ohm resistor acts as a high frequency cut to the overall signal above 3.2kHz. This is opposite of what is claimed in the GeoFX "Technology of the Tube Screamer" article.








Just for the heck of it, here is the frequency response of the isolated gain section with the gain knob at it's lowest setting:









And here is frequency response of the gain section at full combined with the tone section at max:





This is basically the same frequency response you should get with a stock pedal at those settings, since the rest of the circuit does little to effect the frequency response.

---



These plots are from real world analysis done with the circuit isolated between two buffers on a breadboard. A software synthesizer was used as a tone generator.  I had plans to do more, but ran out of time as other priorities in my life came to my attention. I hope these are of use to someone.

[grolschie]

Does this info yield possible new mods to reduce mid hump? 

Jumper the 220 ohm resistor to ground and the mid hump is gone when the tone control is maxed. You still lose some bass in the gain stage, but the whole tone control section goes almost perfectly FLAT. Try it!  Unfortunately a tubescreamer without the mid-hump sounds pretty harsh in the highs. I still think it's a useful mod though, so I put it on a switch in my build.

That discovery is what led me to start this thread where I calimed that the GeoFX article on the Tubescreamer must be wrong. The arguments on that thread were what led to Aziltz conduct his spice analysis which cleared everything up, and I thank him for that. Apparently nobody was following my original post, which is okay, because most of it is just a lot of pointless arguing and mis-communications. I was wrong about some of my initial assumptions anyway, which were just my noob explanations for what I was seeing on frequency plots that directly contradicted what I was reading. If anyone is more curious about the details, I would suggest just skipping to the 3rd page of that thread.

Anyway, carry on...



 

So one could jumper the 200ohm resistor, max the tone knob, and then tack on a rat-type filter just before the output? Or maybe a runoffgroove type series of filters to roll off the harsh highs? Or does this get us back to a mid hump?

[slacker]

That's pretty cool. The only thing I'd say about jumpering the 220R resistor is that in the simulator doing that creates a nasty looking resonant peak at about 100KHz, I don't know what if anything this will do in the real world. If you replace the 220R with 50R then this disappears and you still get a flat response up to about 10KHz or so, making it smaller gets you a flat response up to above audio, but you probably don't need more than 10KHz. I can't hear much higher than that to start with and by the time it's gone through a guitar amp and speakers there won't be much left above that.

[Projectile]

That's pretty cool. The only thing I'd say about jumpering the 220R resistor is that in the simulator doing that creates a nasty looking resonant peak at about 100KHz, I don't know what if anything this will do in the real world.

That's strange, I haven't noticed anything like that in actual practice. I can't account for what would be causing that peak in the simulation, but it doesn't appear to happen in the real world.

[slacker]

I might not happen in the real world with real opamps, but generally peaks like that are bad because they can cause oscillation.

[MohiZ]

So one could jumper the 200ohm resistor, max the tone knob, and then tack on a rat-type filter just before the output? Or maybe a runoffgroove type series of filters to roll off the harsh highs? Or does this get us back to a mid hump?

If you're going to go as far as to add a whole filter, it would be simpler to just take out the whole TS tone stack and replace it with a rat-filter, for instance.. jumpering the 220 ohm resistor could be considered as a quick and dirty mod.

[Evad Nomenclature]

That's pretty awesome to see the actual frequency response all done out like that.

I've always been one of the guys that has never liked the TS at all... I like some pedals built off of the circuit, but the sound has always been to muddy for my playing.
Now I see the big hump I can definitely see why.  For some reason that frequency range while mixing bugs the crap out of me too =)  (like 300-450hz)

nice work 

dave

[aziltz]

I'm not sure the 220 (R9) on its own is the best solution for getting rid of the mid-hump.   Yes, I agree that shorting it causes a "Flat" response, even in the simulations, but only above 720Hz, and only at the brightest Tone setting.  Shorting 220 actually makes Treble DECREASE in the Darker Tone Setting.  I think its real purpose is to shape the highest harmonics, but remember I did not show the graph of the effect of 220 with Tone at 50% because it had no effect at all, mid hump still present.  Although I think its a great place to tweak to get different curves out of the tone control, we can definitely do more.

i think another way to work on the mid-hump would be to play with the R5/C4 and possibly R6/5 Values.  I think by increasing C4, the lower cut off on the gain of the clipping amp would be lowered and the effect would be to broaden/flatten the mid focus.  I believe this is where many tweakers put their mid/full switches (i've seen it on BYOC as well).  I don't think i've seen many switchable options on C4 though, which might be an interesting place to start.  By lowering C4, the cut-off of that RC Low Pass would move higher.

It seems to me that both of those points being centered around 720Hz causes the mid-focus there.
I went ahead to simulate this.  Now that I've got the circuit built it takes 2s to explore ideas like this.

Here's what happens when you increase C4.  Stock, 2x and 3x.  (Same values as the BYOC Mods).  It seems to pull the mid hump down a bit, into the 400Hz range.


Changing C5 (Decreasing)  I chose Stock (.22uF), .16uF and .1uF.  It looks like the mid hump is pulled up, over 1kHz.


Here's both at the same time, raising C4 and lowering C5 TOGETHER.  This looks a lot more broad, while still centered in the 6-700Hz "Mids", with a lot more output. Keeping things relative, a quick comparison would be to look at what point the amplitude drops 5dBs from the peak.  In the stock setting, that happens a lot sooner than in the "Maximum" test values setting.


I don't think we're gonna get anything near a FLAT response out of this puppy, and I suppose that is ok because the guitar is a mids-instrument.  I would like to suggest that the idea of mid-hump is the relatively sharp hump in the stock settings, and simply broadening that will reduce the effect on our ears.

Does anyone have any suggestions for how to create a signal source that models the frequency content of a guitar?  That might give us a better real world picture out of something like this.

[earthtonesaudio]

Just my $.02, I put a trimmer in parallel with the 220Ω resistor in my TS-7, drilled a hole to make it externally adjustable, and found that shorting the 220Ω did "practically nothing" except in the very brightest setting of the tone control.  In that setting, it makes a very harsh treble boost.  I'd agree that something like 50Ω is probably a good value if you want a bit more brightness at max treble without things getting too crazy.

[Cliff Schecht]

Nobody here is interested in phase response?

[aziltz]

Nobody here is interested in phase response?

I wouldn't say nobody! but phase can take a while to explain if you haven't had a basic AC electronics intro... whereas I think even the newest of newbs can understand the amplitude no problem.  But check one box when you run these simulations and you can get the phase response as well, as I'm sure someone like you knows.  Something i'll do when I make this into a article/website.

[Cliff Schecht]

People don't see the immediate purpose of phase plots and tend to ignore them completely. At least from what I've seen, a lot of people really don't know how to interpret phase - especially when it comes to guitar type effects that are looking for distortion and not specific phasing sounds. Filters aren't hard to understand in terms of amplitude vs frequency response - that I completely agree with - but they can get a bit more fishy when you start looking as phase plots, group delay (negative derivative of phase), linear and nonlinear filters, etc.. The tonestack of a tube screamer is a good example of a nonlinear filter, essentially meaning that the filter sections can't be rearranged in any way without altering the response (it doesn't obey the principle of superposition).

It's always easier to apply this stuff to synth style circuits, because the waveforms typically used are so much less harmonically complex (e.g. square, triangle, ramp, sine) and the effects of phasing can be easily seen. The best example of phasing is running a square wave through a filter with a bad phase response (Butterworth, Chebyshev, Elliptical). You can watch the harmonics of the square wave literally shifting in time (phasing can be equated to a time delay at a specific frequency). You slowly start to see the third harmonic, fifth harmonic, etc start to appear as you raise the cutoff of the filter - this is what causes ringing. It not only looks cool on a scope, but it sounds fantastic!

[MohiZ]

I'm one of those who tend to ignore the phase plots. But seriously, what does it matter? I thought the human ear can't hear a difference in phase anyway. I get it that phase differences can be used to advantage in phasers, etc. but in a distortion circuit such as this one, is there a difference anywhere else than the screen of the scope?

[Projectile]

Just my $.02, I put a trimmer in parallel with the 220Ω resistor in my TS-7, drilled a hole to make it externally adjustable, and found that shorting the 220Ω did "practically nothing" except in the very brightest setting of the tone control.  In that setting, it makes a very harsh treble boost.  I'd agree that something like 50Ω is probably a good value if you want a bit more brightness at max treble without things getting too crazy.

Yup, like I said "The tube screamer without the mid-hump sounds pretty harsh in the highs". It's not for everyone, but it IS technically flat above 720Hz when the knob is maxed, and even lower if you combine it with the fat bass mod. Once you start rolling the tone knob back the highs start getting cut again and you bring the mid-hump back. If you want a "scooped" sound when the knob is near max, then you could lower the value of the cap on the wiper. This would shift the frequency response of the feedback network up in frequency, so it doesn't start compensating for the 720Hz roll off until higher in the spectrum. I haven't actually tried it, but this should produce a dip in the mid range between 720Hz and 3.2KHz. If you are going to try this, I would also suggest doing the bass mod, or else it is just going to sound mostly like a treble boost, since most of the bass is already lost in the clipping stage. You could also play with different values for the 220 ohm resistor to determine at what frequency the high rolloff starts.

Another interesting idea would be to replace both the 220 ohm resistor and the 1K resistor after the clipping stage with trim pots, 500ohm and 5K respectively. Then you would have control over both the rolloff points of the initial 720hz rolloff, as well as the 3.2kHz high sculpting. Unfortunately you cannot also change the frequency where the feedback network boost starts without swapping out the cap, because if you change the value of the 1K resistor in the feedback network, you change the gain of that stage significantly.   

Overall, I don't think the TS tone stack is a very good design. I don't know anybody who actually uses the tone knob below the first 1/4 or maybe 1/3 turn from the top of it's rotation. It's fun to discuss what happens when you turn the knob down that far, but in reality the whole bottom two thirds of the knob's range is wasted, which seems pretty pointless.  The problem is that once you start messing with it, you'll find that, for that particular style of tone stack, the values that were originally chosen actually sound pretty good. If you want to have a more useful tone stack it really requires a pretty drastic re-design, which is what is discussed in the AMZ article.