Vox AC30 transfer curve and clipping sim circuit

[Steben]

Based on measurement paper of the AC30 behaviour I broke the response down in 3 parts.
What I notice is a large curved gradual part into saturation before reaching a sloped part, very line like.


The upper part can be simulated with a non inverting soft clipper with high enough treshold. The curve is less easy and needs some ladder style clippers.

[ThermionicScott]

Which paper, out of curiosity?

[Steben]

https://wenku.baidu.com/view/cc6a677a1711cc7931b7167a.html

[iainpunk]

i'd add a hard clipper above the existing highest threshold, like 8 diodes, to make sure the output doesn't get 'too clean' when the input is very high, just in case, but for the rest, that's a very interesting topic.

cheers, Iain

EDIT: image:

[Steben]

i'd add a hard clipper above the existing highest threshold, like 8 diodes, to make sure the output doesn't get 'too clean' when the input is very high, just in case, but for the rest, that's a very interesting topic.

cheers, Iain

EDIT: image:

Amps in fact work the other way around. In front of the power amp other stages get clipped before that but the power stage works with higher signals.
Some scaling between the stages can do the job.
Or like in some overdrives one can put a protection clipper before this stage.

Is that Falstad?
How can you add signal images and transfer functions?

[ThermionicScott]

https://wenku.baidu.com/view/cc6a677a1711cc7931b7167a.html

Ah, interesting.  So the goal was to improve PSpice models by using signal bursts to measure the transfer characteristics of a real one?  Looks like much of the text from page 3 and 4 is missing from the preview.

Since my head is in MOSFET-land these days, the similarity in transfer curves came to mind.  Note how there are 5 "zones" here as well.  Fairly linear in the center and outer zones, curved in the transition areas:


(from https://slideplayer.com/slide/6611213/ )

To be frank, I doubt this is the missing key to perfecting an AC30 emulation.  It seems to confirm much of what you would get by playing around with composite curves of EL84s.

[Steben]

https://wenku.baidu.com/view/cc6a677a1711cc7931b7167a.html

Ah, interesting.  So the goal was to improve PSpice models by using signal bursts to measure the transfer characteristics of a real one?  Looks like much of the text from page 3 and 4 is missing from the preview.

Since my head is in MOSFET-land these days, the similarity in transfer curves came to mind.  Note how there are 5 "zones" here as well.  Fairly linear in the center and outer zones, curved in the transition areas:


(from https://slideplayer.com/slide/6611213/ )

To be frank, I doubt this is the missing key to perfecting an AC30 emulation.  It seems to confirm much of what you get by playing around with composite curves of EL84s.

That is a classic transfer of a CMOS inverter:  It is much closer to feedback style amps where points 3 and 5 are more towards clipping than in fe an AC30. The linear center zone is small in an AC30 and than goes into clipping

[ThermionicScott]

That is a classic transfer of a CMOS inverter:  It is much closer to feedback style amps where points 3 and 5 are more towards clipping than in fe an AC30. The linear center zone is small in an AC30 and than goes into clipping

The broad zones on their chart (colored green on yours) aren't clipping, they're just nonlinear amplification*.  Pentodes don't have to get very close to clipping or cutoff before they become less linear.  It's why Mullard published amplifier circuits with negative feedback while introducing the EL84 and EL34.


(* Unless I'm reading these graphs totally wrong, which is always a possibility.)

[Steben]

That is a classic transfer of a CMOS inverter:  It is much closer to feedback style amps where points 3 and 5 are more towards clipping than in fe an AC30. The linear center zone is small in an AC30 and than goes into clipping

The broad zones on their chart (colored green on yours) aren't clipping, they're just nonlinear amplification*.  Pentodes don't have to get very close to clipping or cutoff before they become less linear.  It's why Mullard published amplifier circuits with negative feedback while introducing the EL84 and EL34.


(* Unless I'm reading these graphs totally wrong, which is always a possibility.)

Yes that is what i ment with closer to clipping. In essence there is almost no clipping only more non linearity. But without feedback the linear part is small.

[iainpunk]

How can you add signal images and transfer functions

i just use a scope, right click on a wire, click 'view in scope' or 'view in undocked scope'
they both have their advantages,

the transfer function is generated with a sawtooth generator and snipping off the straight walls with 'snipping tool'. i used a 2v sawtooth, 50Hz and a 180 degree offset to let the slope cross 0V at a decade line on the time scale, but if you make 2 scopes and combine them, you can click properties and get it to make a XY graph, if one scope is the input and the other one the output, you get a perfect transfer function, but that also shows phase changes, and i am personally not that interested in that, so i use the sawtooth method.

cheers,
Iain

[GFR]

https://wenku.baidu.com/view/cc6a677a1711cc7931b7167a.html

Author Stephen Moeller had a site with schematics for an AC30 simulation (with opamps and diodes to shape the waveform). The parts values were not public. Apparently he later sold the design and deleted the page.

[Axldeziak]

https://web.archive.org/web/20030803202853/http://sound.westhost.com:80/project47.htm

[Vivek]

Is that the overall composite transfer function considering all stages ?

I suppose it is hard to approximate

EQ clip EQ clip EQ clip EQ

With

EQ clip EQ

[Eb7+9]

it is hard to approximate

EQ clip EQ clip EQ clip EQ

With

EQ clip EQ

the problem here is that your premise lies in a totally different universe than the real (3-D) amplifier ... case in point, one cannot replicate a non-linear dynamic system - let alone a highly non-linear one - by a composition of LTI systems smothered in diode strings ... only to try to reduce it further again, what an op-amp distorter does ... anybody who's ever tried recording direct with a tube screamer knows this

here's the original paper:

http://dafx.de/papers/DAFX02_Moeller_Gromowski_Zoelzer_measurement_nonlinear.pdf

compare figure 4, the result of a full NL simulation - which clearly shows bias shifting between successive cycles (typical of NL circuit transient response) - to the averaged equivalents shown later in Figures 7 (and similarly Figures 8, 9,and 10) ... the later is what an op-amp and diode combo will 'hardware' simulate ... even with mismatched diode strings there's no bias shifting of an op-amp // same math reduction, same lowered order of dimensionality ...

in this sense one can only attain an equivalence at the spectral level (not the dynamic) using a cascade of LTI systems artificially clipped, and attaining pretty much the same level of complexity with only one LTI system and diodes ... this is also why/how a DSP emulation of an amplifier sounds/feels at best 2-D since the DSP transform itself is based on LTI equivalency ... the last part of my book explains this part in more detail ... (pp 235-239)

https://viva-analog.com/product/ifmta-book-pdf/

one did considerably better by going the isomorphic jFET simulation route ...
as for the output stage, well ... you know

although not spectrally equivalent, NL dynamics are exactly what the Mullard/Deacy and PigNose one watters bring to the table ... , just saying

https://soundcloud.com/eb7-9/kat-deacy-rangemistress-lowgain

[Vivek]

I heard that the QUAD CORTEX is using modified Impulse responses to be able to capture dynamic responses.

I read a paper somewhere on extended Impulse response for non LTI systems

[Vivek]

To be frank, I doubt this is the missing key to perfecting an AC30 emulation.

My friend Dr. Buffa made an online Amp Simulator where you can choose amongst around 20 Transfer functions.

And have 2 or 3 or x number of tube stages in the emulation

with EQ in between all stages


and plug your guitar to the PC and hear the effect online

or choose one of 15 Wav already loaded


https://mainline.i3s.unice.fr/AmpSim5/index.html

It is enlightening !!!

[Vivek]

there's no bias shifting of an op-amp // same math reduction, same lowered order of dimensionality ...

There are simple ways to get Bias shifting effects in Opamp and FET circuits.

Most commercial pedals do not use these techniques.

[teemuk]

I must add that even very early (mid 1990's) DSP modeling schemes did feature techniques to estimate the dynamic DC shifts of gain stages. Such are, for example, discussed in Yamaha and Line 6 patents. If they weren't ignored then they surely aren't ignored now. 

[Steben]

How can you add signal images and transfer functions

i just use a scope, right click on a wire, click 'view in scope' or 'view in undocked scope'
they both have their advantages,

the transfer function is generated with a sawtooth generator and snipping off the straight walls with 'snipping tool'. i used a 2v sawtooth, 50Hz and a 180 degree offset to let the slope cross 0V at a decade line on the time scale, but if you make 2 scopes and combine them, you can click properties and get it to make a XY graph, if one scope is the input and the other one the output, you get a perfect transfer function, but that also shows phase changes, and i am personally not that interested in that, so i use the sawtooth method.

cheers,
Iain

et voila! With diode clipper in front to cut too large signals

[iainpunk]

et voila! With diode clipper in front to cut too large signals

looks nice!

@Eb7+9
a lot of amp in a box pedals use mosfet or jfet stages that most definitely have a lot of dynamic DC shift going on, since all of the sages clip asymmetrically. remember that jfets without source capacitors behave just like triodes. and stacking them in a cascode consisting of a jfet in the bottom and a mosfet at the top is really really close to how pentodes react.

someone should make a cascode ''poweramp saturation drive'' with actual output tranny and realistic power sag scheme etc.

cheers, Iain