Tube power amp simulation

[stm]

I've always thought that depending how hard you drive it, the power amp section added a little bit of "warm distortion", less raw and more musical than the preamp one.
But I can't understand what is doing this.
...

After several years of asking the same to myself (about the little bit of "warm distortion" in the power amp section) I'm pretty sure I have sorted it out.  I'm talking about the "near-clean" warm sound which appears when the amp volume is set at certain level where the warmth appears, and if you take back the volume a bit then this "magic" dissappears.  At this volume it is not about phase inverter saturation, not about power supply sag, not about output transformer saturation, not about speaker breakup (however some speakers may show breakup effects at low volumes), and certainly not about speaker thermal compression.

It's the power amp nonlinear transfer function which adds mostly third order harmonics.  The "ideal" transfer function would be a straight line. The practical transfer function in a push-pull valve amp looks rather like an "S" shape to each side of the origin of the graph.  Negative feedback sort of straightens this wobbly transfer function, but not entirely.  Amps with no feedback at all (like the Vox AC30) have very warm "clean" sounds because of this.  Search around for testimonials of people that have removed negative feedback from their amps, giving their amps a more "vintage" and warm tone instantly.

You may see a practical implementation of this in a future ROG project.

[Sheldon]

I think I have issues with my op amps
none of them works as they supposed to work!

[sajy_ho]

After several years of asking the same to myself (about the little bit of "warm distortion" in the power amp section) I'm pretty sure I have sorted it out.  I'm talking about the "near-clean" warm sound which appears when the amp volume is set at certain level where the warmth appears, and if you take back the volume a bit then this "magic" dissappears.  At this volume it is not about phase inverter saturation, not about power supply sag, not about output transformer saturation, not about speaker breakup (however some speakers may show breakup effects at low volumes), and certainly not about speaker thermal compression.

It's the power amp nonlinear transfer function which adds mostly third order harmonics.  The "ideal" transfer function would be a straight line. The practical transfer function in a push-pull valve amp looks rather like an "S" shape to each side of the origin of the graph.  Negative feedback sort of straightens this wobbly transfer function, but not entirely.  Amps with no feedback at all (like the Vox AC30) have very warm "clean" sounds because of this.  Search around for testimonials of people that have removed negative feedback from their amps, giving their amps a more "vintage" and warm tone instantly.

You may see a practical implementation of this in a future ROG project.

Exactly! I'm talking about the same "Warm Sound". A sound that is not very distorted (from aspect of clipping distortion), but contains a lot of nice sounding harmonics.

[Sheldon]

For now I'm trying to simulate Seljer schematics in Johan's thread :
http://www.diystompboxes.com/smfforum/index.php?topic=107868.msg981504#msg981504
It's basically the same idea with JFET, so I don't have to rely on lt spice tube simulation.
For now I'm not able to have the same scope signal as him (mine are less assymetrical, more clipped on the top), but it's quite interesting as it brings even and odd harmonics, with less even harmonics (I assume it's the effect of the phase splitter) so I think it's like a push pull power amp.
I'd like to know how is it possible with the mosfet to adjust the quantity of harmonics. I assume that the ratio odd harmonics/even harmonics is due to the signal assymetricality? So I have to play with the jfet bias voltage?
And how to adjust the global quantity of harmonics? For instance, how to attenuate harmonics past the 5th order?

[tca]

For now I'm trying to simulate ...

Simulating a circuit is fine but you should never, I repeat, never, avoid testing a circuit on the breadboard based on the sims alone.

Try this on the breadboard:



and feed it with a hot clean signal... You'll have a surprise (http://www.diystompboxes.com/smfforum/index.php?topic=93930.0)!

You should also try Tim Escobedo's bootstrap:



and also Arsenio's Novo - Tube Sound Overdrive



Have fun!

P.S.
Arsenio's - Tube Sound Overdrive is my favorite.

P.P.S.
Have a look at Teemu's book: http://www.thatraymond.com/downloads/solidstate_guitar_amplifiers_teemu_kyttala_v1.0.pdf

Specially from pg. 18 to 26.

[sajy_ho]

And how to adjust the global quantity of harmonics? For instance, how to attenuate harmonics past the 5th order?

Maybe by adding several low pass filters at the end of chain, like this one: http://www.runoffgroove.com/thunderchief.html

[tca]

And how to adjust the global quantity of harmonics? For instance, how to attenuate harmonics past the 5th order?

You don't. At least with an analog circuit, but you could do it digitally.

I think the question is not how to remove attenuate  it, but how do *all* harmonics scale to one another. What are the ratios of the 2nd, 3rd, etc...to the fundamental?

The 5th isn't all that bad it all depends on the amplitude ratio to the other harmonics.

[Sheldon]

I have a question
If you have a transistor phase inverter feeding 2 jfet for soft assymetrical clipping, what is the method to bias the transistor and the jfet?

[samhay]

Something I've been cooking up lately.....

Thanks - this gave me a push to dust off some old ideas about envelope-controlled clipping and actually get them working.
Have started a new thread:  http://www.diystompboxes.com/smfforum/index.php?topic=110346.0

[mac]

P.S.
Arsenio's - Tube Sound Overdrive is my favorite.

Arsenio Novo =  Bazz Fuss 

http://www.diystompboxes.com/smfforum/index.php?topic=98141.msg857451#msg857451

mac

[J0K3RX]

I have to ask, what are you going to do with just a tube power amplifier simulation circuit?

[Sheldon]

For rehearsals or small gigs, instead of carrying an amp, I'd like to use a pedalboard with tube preamp ->  power amp sim -> cab sim
Like a preamp into a torpedo cab that have a power amp sim section that brings beef up and warm up a bit the preamp distortion.

For the moment I have to admit I'm a bit discouraged.
Teemuk schematics seems nice but too complex for me.
With lt spice I finally undertood how bjp-op amp phase inverters work, how to have assymetrical clipping with op-amps or jfet (even if I have trouble with biasing), how to sum the two push-pull signals. But at the end of the days, it gives nothing more that some "cold" harmonics, nothing like the Torpedo Cab power amp simulation.
I think I should try to implement a feedback inverter with presence and depth filters, but for the moment, I don't know how!

[Quackzed]

this might be a job for a diode ladder type soft clipper... i messed with a few different variations and you can get a pretty smooth 's' shape or 'saturation' type shape with it... it's not a common trick, other than your typical ' diodes with a resistor in series' type softness control, or warp control. this approach takes the transfer curve a step further and allows you to stack several thresholds, each with additional soft clipping added. so the signal hits the first lowest threshold and bends a little, then hits the next and bends a little more, then the next and bends a little more etc... like an s curve... there are a few ways to do it, and it offers some stepwise control of the curve, so you can adjust it to bend/clip the signal 10% say, then at the next threshold an additional 33% then at the next 75 % etc... you need a decent voltage swing to use the diode ladder type setup, as you need to go through 3 or 4 diode thresholds without hitting the rails, but theres another way to do it where you set up the soft clippers one at a time and just voltage divide the signal down after each stage to make it the right size for each separate soft clipper...

to simplify   

first stage:  b2b leds(1.2 threshold-ish) with a small value resistor in series to ground... clips just the peaks alot say 85% but not quite  flat.
second stage ; voltage divide to where the already clipped off part is juuust above this stage's clipping threshold : si with a medium value resistor in series... clips the signal say 50% (anything above the si diodes is half clipped)
third stage ; voltage divide to where the already clipped off part is juuust above this stage's clipping threshold : ge' with a large value resistor in series... clips the signal only 20% say -bending the transfer juuust a little bit...

so a signal big enough to get through all the thresholds will...
1st. hit the ge's and gently bend a bit... a little compression, and a little low order hamonics added ( 20%)
2nd. hit the si's and bend a bit more ... more compression and more  harmonics added (50)%
3rd. hit the leds and bend alot ... compressing peaks and adding higher order harmonics (85% clipped)

a big signal will get progressively 'rounded' a small signal will just get a bit bent and a medium signal will get softly squashed but still have dynamics... never hit a hard ceiling... the smoothness of the curve keeps the added harmonic content from getting buzzy, by having less high order harmonics added till you slam it, but you dont have to slam it....
i should have just drawn it up... 
i do have a falstads simulation of the idea i could post if you're interested. lemme know. that way you could play with the values etc... to get  the curve you want...

[Sheldon]

of course I'm interested!
I'm so lost that I will be glad to test and learn any idea!

[Sheldon]

is it possible to add a negative feedback simply by adding an op amp before the phase inverter fed with input and a signal taken from the output?
I tried it with lt spice with different resistors values but it seems I've done something wrong!

[Quackzed]

heres a falstad sim of the idea...
go here http://www.falstad.com/circuit/ go to file/import from the app window and copy/paste the code into the import window.

$ 1 5.0E-6 2.275989509352673 7 5.0 43
a 496 432 608 432 0 4.0 -4.0 1000000.0
c 400 416 352 416 0 1.0E-7 -0.039382139813982045
r 352 416 352 496 0 100000.0
g 352 496 352 512 0
r 496 416 448 416 0 22000.0
w 496 416 496 368 0
r 496 368 608 368 0 270000.0
w 608 368 608 432 0
g 496 448 496 464 0
R 352 416 288 416 0 1 500.0 0.3 0.0 0.0 0.5
c 608 432 656 432 0 1.0E-6 -0.03528020967951877
O 992 432 1040 432 0
r 656 432 736 432 0 22000.0
r 736 432 816 432 0 10000.0
r 816 432 880 432 0 10000.0
r 736 432 736 496 0 22000.0
r 816 432 816 496 0 10000.0
w 736 432 720 448 0
c 992 432 960 432 0 2.2E-7 0.01345104037952944
r 992 432 992 544 0 47000.0
g 992 544 992 560 0
r 720 448 720 496 0 47000.0
w 816 432 800 448 0
r 800 448 800 496 0 47000.0
w 880 432 864 448 0
r 864 448 864 496 0 47000.0
d 736 496 736 544 1 0.6
d 752 544 752 496 1 0.6
w 752 544 736 544 0
g 736 544 736 560 0
w 880 496 896 496 0
d 896 496 896 544 1 0.6
d 880 544 880 496 1 0.6
w 880 544 896 544 0
g 880 544 880 560 0
d 816 496 816 544 1 0.6
d 832 544 832 496 1 0.6
w 816 544 832 544 0
g 816 544 816 560 0
r 880 496 880 432 0 4700.0
w 864 496 864 544 0
w 864 544 880 544 0
w 720 496 720 544 0
w 720 544 736 544 0
w 800 496 800 544 0
w 800 544 816 544 0
w 832 496 816 496 0
w 752 496 736 496 0
w 608 368 608 336 0
w 496 368 496 336 0
c 496 336 608 336 0 4.699999999999999E-11 -0.044992037532930275
r 400 416 448 416 0 2000.0
w 960 432 896 432 0
w 896 432 880 432 0
o 11 2 0 34 0.4703509377162163 0.0 0 -1

[Sheldon]

thank you

[DDD]

Hi guys,
There's "pure BJT" simulation of tube push-pull amp in the link below.
You can find stompbox version as well as 1,5 Watts RMS output stage there (also on the picture below plus output waveform)

http://forum.gtlab.net/cgi-bin/yabb2/YaBB.pl?num=1427446870/20

[PRR]

> As far as I can understand it, a push pull power tube section is attenuating most of the even harmonics

No. It does not *add* even harmonics. Any evens which come in are passed like any other tone.

A push-pull stage will always ultimately add odd harmonics when it overloads.

> but I thought that it was those even harmonics that give the warm and musical tone?

Very substantial amounts of even harmonic are scarcely audible. While even I can tell the difference between a Fender input triode and another input with so much NFB that it is "clean", the difference is very subtle.

ODD harmonic distortion is the sound of rock-n-roll, the Power Chord.

Single-ended stages may add 2nd harmonic to medium-strong signals, but *always* add 3rd harmonic as they get into overload.

Come home drunk in a loose-steering car and park in the garage. If the steering slop is one-sided you will tear-up one side of the car. If both-ways you tear up both sides. Even if one-sided slop, if you are drunk enough you will tear-up both sides of the car as you bang from wall to wall. So don't drink and park.

High-order distortion is typically weak and masked by low-order until you get to Gross Overdrive. Yeah, we do that. Up to the 9th(?) harmonic is on the musical scale, the 11th and up is nasty; also for guitar fundamentals a spurt of partials up around 3KHz is ear-piercing and some may say "unmusical". Bandwidth limitations (guitar speaker top-cut) and rounder clipping reduces this, but also reduces the "good" 2nd and 3rd harmonic.
_______________________

> Transistors work in another way, they do not add much harmonics to the signal at clean level

Naked BJT transistors distort *worse* than tubes or FETs. Just below hard-clipping they will do 26% of 2nd, versus 5% for tubes. However they have SO much nonlinearity and gain that we almost "never" run them without significant negative feedback. When we do, it is because we *want* hard clipping as soon as possible, and never work much lower.
_______________________

The waveforms you see in steady-state 'scope tests are not what you will get with guitar. Usually (not always) there are caps, and an overdriven stage will charge-up these caps as it overloads. For tube interstage coupling caps, a typical response is a burst of distortion on the initial transient and then the stage un-biases itself to a different, sometimes lesser, level of distortion, and lower still when/if the note decays. In the extreme a tube grid with a HUGE signal can de-bias itself so far it quits amplifying, "blocking". Dancing on the edge of grid-block has been a key tone for some classic guitarists. In any multi-stage amplifier chain you can get a variety of tones this way, but you can also "stun" the amp so it is tricky business.
_______________________

All "simple" distortion mechanisms are VERY level-critical. 3rd harmonic typically rises as the *square* of the signal level. The idea of a distortion effect which will take "any" signal of any level and deliver a constant racket is probably folly, or at needs a very complicated mechanism.

[DDD]

"... Dancing on the edge of grid-block has been a key tone for some classic guitarists. In any multi-stage amplifier chain you can get a variety of tones this way, but you can also "stun" the amp so it is tricky business. ... " - that's why the famous Tube Screamer is so popular.