Green Citrus AIAB LTSPICE Analysis

[Vivek]

About 1.5 years ago, when I was a raw noob around here, I asked the question "Where do I find data on inter-stage frequency response and gain for multi-stage pedals and amplifiers ?"

I thank the many regulars here who suggested some excellent books and web sites. One advice stuck out though, it was "Simulate, Simulate, Simulate"

Thus began the Coffee and Covid fueled obsession of entering pedals and AIAB into LTSPICE.

I recently came across some thread here about multistage FET based pedals / AIAB, hence I entered the GREEN CITRUS schematic into LTSPICE.

This design was first proposed by Branimir in 2008. It appears that was the high point of "Stack some FET based gain stages, add some tone controls and voila !!"

https://www.diystompboxes.com/smfforum/index.php?topic=71597.0





Finding suitable FET models:
LTSPICE did not have any built in model for the 2n5457 FET, so step one was to google out the models and install into LTSPICE. I found 3 models.

Next step was to bias the FET to the recommended 4.75V. This proved to be tricky. I first chose the model proposed by Purdue University. The FET are very sensitive to trimmer resistance. It's obvious that a multi-turn trimmer will be best, if one wants to bias exactly as per recommendations.

The clipping response was fairly dependent on the bias, producing different curves for slight differences in bias point.

Then I tried to use one of the other FET models. They bias totally differently !!. They also produce different clipping characteristics !!

Here is a graph that shows some different transfer curves based on the setting of the Trim pot. You can see that there is a big variation of responses of the stage based on FET characteristics and trim pot settings.



This was a reminder of three aspects :

A.) SPICE models for same reference number can be very different
B.) FETs themselves have too wide a range of specifications. Maybe each model is modelling a physical FET that is within spec but far away from the other FET with the same part number
C.) The interconnection of trim pot settings with clipping characteristics and frequency response make this an inconsistent system where each physical build will be fairly different than the other. Don't expect consistency with this kind of schematic !!

Teemu K had offered sage advice on multistage FET pedals earlier :

A word of warning: You will not be able to build this circuit with consistent results. Inherent and fairly large variation of FET characteristics ensures that you "trim" every drain load to match the particular FET characteristics and preferred bias point. Effectively you will end up with different drain load for each FET stage.

The generic "drain trimmer" bias of FETs usually suggest to bias to a drain voltage that is approximately half of the supply voltage. Biased this way, the FET stage will introduce very little overall asymmetry and it is evident that it doesn't really perform like tube gain stages that are, for preferred effect, very often biased for deliberate asymmetry.

Not only that, the drain trimmer will alter stage gain (as noted) and its output impedance, which in turn will alter corner frequencies of subsequent RC filters. Source biasing with trimmer is not technically any more consistent. Yes, you can minimize gain variation of AC signals by capacitive bypassing but the trimmer resistance will affect ALL DC bias conditions of the circuit (e.g. idle drain voltage) and therefore clipping characteristics.

You perhaps start to understand now why the most recent "Runoffgroove" project abandoned using FETs and began to exploit much more predictable operational amplifiers and clipping diodes with predictable thresholds...

I tried not to be discouraged by the futility of studying an inconsistent system where the analysis is quite meaningless by definition, but decided to freeze one set of parameters and try and see what else could I find. I decided to use the Purdue FET model and bias at 4.75V as suggested in the FUZZDOG kit build docs.

First Stage:


I placed a label TP11 at the output of the first stage. I used an ideal voltage source for signal. It has no internal resistances, capacitors or inductors.

The input impedance of this circuit is a bit below 220K ohms. It would have been better to design for at least 500K Ohms, if not even higher. There is a possibility of slightly reduced signal at input and attenuated treble due to this lower impedance, but some circuits use this to their advantage / factor it in during the design.

The frequency response of the first stage is extremely flat. Maybe it would have been better to have a high pass filter to cut away sub audible bass, and a low pass filter to remove unwanted high frequency content and RF. Unless the designer is counting on the loading effect of the low impedance of the input to cut the high frequencies.



Now I use that totally blasphemous, completely deceptive and wrong tool which I stupidly call "AC Transfer function". Basically I run input signals between -2Vp and + 2Vpeak at 1Khz, and plot the Vtrough of the output between -2Vp and 0Vp input, and the Vpeak of the output between 0 and +2Vp input. Previously I would manually run about 20 simulations and enter data into Excel to plot. Now I found a way to automate the process and can get Spice to collect and plot 200 to 1000 data points per curve while I sip my coffee.



This graph tells us:

1. The stage exhibits asymmetric soft clipping, somewhat akin to the transfer function of a tube stage. The FETZER stage has similar response too.
2. The gain for very small signals is about 10.6x
3. We can roughly identify two points on the curve at 330mvP and 600mvp inputs, where the output curve is at "edge of breakup". This means that the first stage will add some crunch at start of a note, but cleans up when the note dies.

Here is Stage 2:
I added a test point labelled TP22


Now we do a cumulative Frequency analysis and AC transfer function considering everything from input till TP22.


We see that the frequency response is substantially flat, with no reduction of sub bass, no removal of higher harmonics, no "mid hump". This shows different design choices than most Amps, Tube screamer type pedals.

The blue line shows the cumulative AC Transfer function at TP22, with the gain pot at 10% of its resistance value.


and here is the graph of the AC transfer function at TP22 in blue, for gain pot at max gain position.


The green curve at TP11 does not change since TP11 is before the gain pot.

We see that at 10% gain pot, the cumulative gain till TP22 is 12.5x
But with gain at max, the cumulative gain till TP22 is 129x ie the second stage added 12.17x gain to the 10.6x gain of the first stage

Now we look at stage test point TP33





The green curve is the frequency response at the gate of Q3. I tried to make the tone controls have a flat response.

The red curve is the cumulative frequency response from input till TP33.

That is totally different than the "typical" design that has low pass filters to remove "fizz" created by earlier stages.

However, when tone controls are flat, we seem to have inherited a Bass boost of about 6db at about 20 Hz, and a treble cut that makes 10Khz go down about 3.3 dB versus the flat region around 300Hz. These can be altered by the tone controls.

Here are the results at TP33 with tone controls at max and min positions.


We see that it is possible to have a shallow mid hump after all, when tone controls are at min position.

Looks like we will need some filter to further remove high frequency signals, and indeed there is a low pass filter at the end !


Red curves are frequency response at min and max tone controls at TP33, before the final low pass filter.

Green curves are frequency response at min and max tone controls at output, after the final low pass filter. We can see the effect of the low pass filter in reducing the high end and trying to contain it to the range that is reproduceable by a Guitar cabinet. Yet, we can see totally different design choices than other pedals, which will lead this pedal to having its own unique tone.

Here is the AC transfer function with Gain pot at 10% resistance


We can see that there are some interesting "break points" very roughly at input values of

100mVp
150mvP
280mvp
290mvp
330mvp
600mvp



I have begin to feel that the harmonic movement of the output signal of multi-stage AIAB depends upon judicious selection of these break points, and meaningfully distributing them across the normally expected guitar signal range + Slope of curve after these break points. Horizontal AC transfer curve = hard clipping = lower dynamics and harmonic movements.

and here is the AC transfer function with gain pot at max, showing cumulative response at TP11, TP22 and TP33:


The cumulative final AC transfer function at TP33 (Red curve with horizontal stretches) shows hard clipping where all signals above a particular clipping point are strongly clipped to the same level. This will increase sustain but limit dynamics and harmonic movement.

We can see that the maximum gain of this pedal is ~660x, and that input signals over about 6.5mvP will be distorted.

Over to the Transient analysis, first with gain pot at 10% resistance and input signals of 5mv 10mv, 50mv, 100mv, 250mv, 500mv, 1000mv, and 2000mv at 1Khz.


As expected from the AC Transfer function graph with gain pot = 10%, lower level signals are not distorted. For larger signals, there is some slight movement in wave shape and size, leading to harmonic movement. However the differences in output for higher signals is not much, indicating hard clipping that limits the rise of signal above the clipping point. The AC Transfer function had warned us about that, since the red curve is almost flat after the clipping begins.

We can see that the clipping is asymmetrical and the level of asymmetry depends upon the amplitude of input signal.

Now a transient analysis with the gain pot at 100% resistance:


As expected from the AC transfer function with gain pot = 100%, only very small signals below about 10mvp get by unscathed. All others are chopped, diced and shaped into almost the same final shape and amplitude. We gave gained sustain at the cost of dynamics.

If we call these shark fin shaped output waveforms as "Essentially square waves with their leading edge smoothened out", it means the square wave generated at clipper passed through a low pass filter that limited the high frequencies needed to truthfully follow the leading edge of a square wave. And indeed we saw a passive low pass filter after the last active stage.

FFT analysis : Pending, but we can expect from the wave shape, it would be somewhat similar to a square wave but lower levels of the higher harmonics as compared to a pure square wave.

Possible mods :
A ) We could have made input impedance higher, maybe around 500K ohms.

B ) We could add interstage LPF and HPF filters to generate different sonic signatures

C ) We could have added voicing switches that alter final output by providing different gains and filters at internal stages.

D ) We could attempt to increase the harmonic movement at output by avoiding severe hard clipping and instead allowing larger signals to grow slightly.

E ) We could try to incorporate some kind of "Sag" such that amplitude and harmonic content of the output partly depends upon integral of the input signal envelope in the earlier 1-2 seconds.

F ) We could try different knee frequencies on the tone control, to see what suits guitar better. Here is the link to the James tone control at TSC IN THE WEB, with the component values of this design already loaded in : https://www.guitarscience.net/tsc/james.htm#RIN=25k&R1=100k&RB=1M&R3=22k&R4=100k&RL=10M&RT=1M&C1=2n2&C2=22n&C3=330p&C4=2n2&RB_pot=LogB&RT_pot=LogB


Thanks going out to all the good folks here who guided me on my guitar electronics journey so far. Please help me to learn more.

i) Please inform me if there are any errors or shortcomings of this analysis.

ii) What is essential for AIAB analysis but was not covered ?

[PRR]

> when tone controls are flat, we seem to have inherited a Bass boost of about 6db at about 20 Hz



Where did C9 come from? It does not seem to be in the original plan? It affects DC bias (and perhaps for the better?) but will give a significant Bass boost.

Also C13 is not the suggested value.

[Vivek]

Dear Paul, thank you for reading through that rambling !

I entered the schematic from http://pedalparts.co.uk/docs/GreenCitrus.pdf




I will now try the values/ connections from the original schematic as well.

PS : Is there any specific name for the configuration of the last 2 FET ?

[antonis]

PS : Is there any specific name for the configuration of the last 2 FET ?

You can name them as you like..
(e.g. PMAPS - phase mixer of already phase splitted..)

[ElectricDruid]

That's a very interesting read, Vivek. Thanks.

I particularly like an analysis with graphs and pictures! So much better than just descriptions!

[Vivek]

Thanks Tom, for your encouraging words.

[anotherjim]

PS : Is there any specific name for the configuration of the last 2 FET ?

You can name them as you like..
(e.g. PMAPS - phase mixer of already phase splitted..)

Signals & paths are not matched. I suspect it's the feedback path from an imaginary  "output transformer" but nulled out when C13 is fully in circuit.

[teemuk]

Yes, those two stages are runoffgroove attempt to "emulate" the power amp section. They could have used more imagination to build a more realistic simulation (e.g. differentual circuit) but I guess the know how was limited to common source circuits with 100k drain trimmers. So, they merely emulate a negative feedback circuitry, probably at wrong magnitude and yes, the resonance circuit has morphed to something that actually controls NFB across the effective bandwidth.

The circuit in general - like name hints - is an attempt to emulate vintage Orange amps. (Take the tube circuit, replace tubes with fets and plate resistors with trimmers, and erroneusly expect that result is similar to original  tube circuit). And yes, these amps typically did not have much hi-passing prior to distortion, which you can also hear in their extremely flubby and fuzzy tone of overdrive. They have very different sound than that of Marshall-derived high gain amps.

[POTL]

1) there is no FAC control in the circuit, a very important part of the sound and flexibility of old oranges and matamps.
2) Over time, I began to realize that all these preamp circuits sound boring, without imitation of amplifier.

[TheMythicalNarwhal]

I was lurking around reading random threads this morning, trying to absorb some second-hand knowledge, and came across this thread. I just built two of these Green Citrus PCBs. I had to register an account (what took me so long?) to say thank you for this great write up! Such great info at the perfect time! Your analysis confirmed what I was already hearing- there is truckloads of bass in this design. Playing in a lowered tuning, it's Flub City, and quickly worsened if you boost it with a full spectrum drive pedal without good EQ. (It also helps if I would have put the right value capacitors in the tone stack, for starters!  ) Like POTL said, they left the VAC control out of the circuit, and opted for the highest cap value/most bass as the default. Matamp from what I have read used a 4-way called Depth, and placed it after the first gain stage, opposed to the Orange FAC 6 way after the tone makeup gain stage. I'm going to try to add a version of the Matamp depth control on a switch which should tighten up bass considerably.

I will have to give your post a thorough re-read, and spend some time dialing the bias on the JFETs, they seemed particularly sensitive while biasing, but maybe I should consider upgrading to multi-turn trims.

FWIW I can't compare a vintage Orange or Matamp/Green to the pedal, but I do love it for what it does, and it seems to capture that spirit! Even in it's work-in-progress state I am using it as my primary distortion component. It has a unique, strong character with a bit of feeling of transparency, and takes fuzz and modulation before it really well.

Thanks again!

[Steben]

Great work Vivek.
Don't get us wrong. Most of us really adore your efforts.

But ... they prove many things that have been said. Again, tests have value whether the result is affirmative or not.
Multiple stages of eventual clipping "against the rail" - or any bias point for what its worth - will lead to hard clipping curves in high gain mode.
And how much I used to be in love with jFETS in earlier days, those curves can easily be mimicked with opamps. With the addition of variable current draw (because of AB mode of (most) integrated amp circuits).
You see, any hard clipping amp with power sag has this benefit of having more dynamic playfulness between clean and saturated in a picking style. A feature which still comes in my mind when thinking of why vintage amps sound or play so good. Any correct EQ and a power sag will get you closer to classic "marshallesque" amp feel than the roundness of the knee.
There are many amps, but I keep coming back to "marshallesque" and "voxy" (where Fenders can be placed in either of the two with some EQ mods). Or in other words: feedback fixed bias and non feedback cathode bias.

[Vivek]

Great work Vivek.
Don't get us wrong. Most of us really adore your efforts.

Thanks Bro,
But what makes you feel I am not part of the "us" ?

[Steben]

Great work Vivek.
Don't get us wrong. Most of us really adore your efforts.

Thanks Bro,
But what makes you feel I am not part of the "us" ?

"us" = the ones just looking at your work  Can't be you