SOLDANO SLO preamp stage analysis in question

[Killing Floor]

As a side note, I have been drooling over the new Soldano SLO pedal... I gotta have one, LOL... ($229.99 MAP)

Soldano SLO Overdrive Pedal

Get the sound of a legendary amp on your board! The versatile Soldano SLO pedal has a three-band EQ, plus, volume and gain controls, and a Deep switch.

www.zzounds.com

I’ve been thinking of this too. It’s actually Soldano. But for a couple years my favorite “distortion” has been this clone of the GTO preamp section with 2 actual 12AX7 fully utilized. But I’m always down for a new pedal.

 

[GearGeek01]

You can never have too many pedals

 

[ltournell]

@ to all

Hello.

Richard Kuehnel has answered. First he had noted my mistake about the headroom of this particular stage ( the very one that AndrewRneumann had revealed to me in the discussion up there). Richard K considered himself responsible for my mistake , his article being not clear enough ! So he has rewritten it : many thanks to him.

I replied that my main concern was about the "tonal" characteristics of this stage since I did not see why , with so poor a gain ( 1.9 = 5.5 dB) and so poor an output impedance ( 100k // 4Mo) , we could be satisfied just with a high headroom without knowing the real purpose of that stage. He admitted that there was a lot more to say, which he had done in his latest book about the Soldano ! So this article in the ampbooks.com page is some kind of "teaser" to buy his book, I suppose

Now here is how I see that :
Of course in any natural sound (clean sound) you will find a series of harmonics (or multiples of the fundamental frequency) particularly numerous on a rich sound such as the guitar sound. This is not what we talk about . We talk about the ones produced with the overtones generated by overdriving an amp stage. They are of two kinds : ODD and EVEN. Generally it is a mix of both but it is interesting to make the ones or the others prevail. ODD harmonics are the ones early popularized by the 60's fuzz pedals (remember Jagger-Richards' Satisfaction ?)
If you preserve symmetry of the wave when overdriving a stage you will produce ODD harmonics and cancel the EVEN ones :

Here we see a square wave recorded at the output of an EL34 SE amp on a dummy load . To cancel the EVEN harmonics you must respect the symmetry axle shown , which includes preserving a duty cycle of 50%. Of course perfect square waves do not exist in real life and you can see that the EVEN harmonics on the spectrum are only considerably reduced ( - 35 dB when compared to the fundamental)

The maths behind that say that the X(t) signal is equal to :

It is clearer if if take ω = 2 π f
Then we have 4/π ( sin 2 ωt + 1/3 sin3ωt + 1/5 sin5ωt ...)
so the ODD harmonics appear clearly.

In fact to produce symmetrical overdriving you use stages with a middle bias point enabling equal headroom up and down , the same as we use for clean signals ! Here is a famous example :

But some do not like ODD harmonics ! They prefer EVEN harmonics . To be precise , it is almost impossible to cancel ODD harmonics and even to reduce them considerably. All we can do is promoting back the EVEN harmonics in breaking the symmetry of the signal and this is what this SOLDANO stage does. With its cut-off at -4.3 V it beheads the signal at the output and as AndrewRneumann justly remarks we ensure the positive side never clips ! So at the output we get an overdriven signal which is rich in EVEN harmonics. Now what it becomes after is another story ! I think that EVEN overtones have charmed generations of guitarists ( think of the BASSMAN 59 5F6-A for example)

I have made a coarse prototype of a "mixer" following that principle :

On the following video I play with the MIX button to make EVEN harmonics come and go on a 440 Hz overdriven sine :

 

[ltournell]

Here it is just the 50% duty cycle that is broken and the EVEN harmonics are back :

 

[YellowBoots]

@ltournell that’s an interesting analysis. Thanks for writing that up and including images and videos. Always a treat to have something to look at.

My question for you: it seems as if you are working on the principle of superposition in your model. In your model, the even harmonics mix with the odd harmonics by layering on top of one another. But is that a fair way to model two cascading gain stages? I’m not convinced that harmonics are built up that way once clipping is reached.

This is why I asked if you could do a harmonic analysis of the output of stage 4 with stage 3 clean and then with stage 3 clipping. Is the effect the clear and obvious addition of even harmonics?

This is not to cast doubt, but to further expand my understanding.

 

[ltournell]

My question for you: it seems as if you are working on the principle of superposition in your model. In your model, the even harmonics mix with the odd harmonics by layering on top of one another. But is that a fair way to model two cascading gain stages? I’m not convinced that harmonics are built up that way once clipping is reached.

The point is that I forgot to say that the "coarse prototype" I speak of is not a model but the "real thing" : a four triodes (=2 tubes) preamp connected to the loop RETURN of an EL34 SE amp recorded on a 8 ohms dummy load and this is what you see on the video. So I would not speak of superposition by layering harmonics on top of one another but real altering of the sound itself by combining harmonics differently. And , as you point out, my schematic does NOT account for cascading gain stages and it is there for theory only (though I intend to make my next amp with a more sophisticated version of this preamp) .
Of course nothing here is purely binary. When overdriving symmetrically a signal , the ODD harmonics prevail but the EVEN ones are always present but strongly attenuated and when clipping a signal assymetrically you just emphasize EVEN order harmonics but they do not prevail, they are just present with the ODD ones.

This is why I asked if you could do a harmonic analysis of the output of stage 4 with stage 3 clean and then with stage 3 clipping. Is the effect the clear and obvious addition of even harmonics?

Very interesting suggestion ! I am sure that the cascading stages do not result in a mere addition of harmonics and the last stage is always the winning stage ! It will take me some time to make this analysis (within the range of my abilites . I reproduce here for the time being the answer by Richard Kuehnel to my question (and to yours too) that can be the start of such an analysis ( you will see that he invites us to read his book) :

When looking at how the stages work together, it is especially important to consider grid current effects, because this stage typically overdrives the downstream stage, even when its input signal is substantially less than +9.5dBV. You are correct - there is much more to this story. In fact, as I recall, an entire book on this subject was recently published...

This is not to cast doubt, but to further expand my understanding.

Casting doubt is the very engine of fruitful discussion !

 

[2L man]

Now if I bypass the 39k resistor with say a 10 µF capacitor, I will only increase the gain and get this :

View attachment 1096272

This simulation look strange to me when there come so much time difference between positive ~75% and negative ~25% halfs?

Without bypass capacitor simulation was about 50% vs 50%

I understand higher gain what bypass capacitor bring but not uts effect to time span.

 

[ltournell]

This simulation look strange to me when there come so much time difference between positive ~75% and negative ~25% halfs?

Without bypass capacitor simulation was about 50% vs 50%

I understand higher gain what bypass capacitor bring but not uts effect to time span.

Overdriving affects amplitude as well as duty cycle. With an unbypassed cathode resistor the gain is 1.9 (5.5 dB) so the output except the clipping can still look like the input ( clean 20Vpp) . With a bypassed resistor the gain is now 55.1 ( 34.8 dB) so we are no longer overdriving the stage but producing heavy distortion with an output that would reach 1102 Vpp ! The clipping is nevertheless still there, so what we get is a small portion of the signal , hence the far larger duty cycle you see.
Be sure that LTSpice is a very highly reliable simulation tool.

 

[ltournell]

To moderate what I say about the reliability of LTSpice I must say that there is an inconsistency left for me in the evaluation of the low cut-off frequency of a partially bypassed cathode resistor.
I speak about this because the stages of the SOLDANO preamp are all in this case with their 1µF capacitor across the Rk resistor.
LTSpice seems to stick to the good old mistake of Fc = 1/ 2 * π * Rk * Ck as if the pole of the RkCk network was the same as the one of the stage itself ! If we remember the expression of the gain A = µRA / (RA + rak + (µ + 1) Zk ) the low cut-off frequency will be the value for which the denominator will reach 1.414 x its value in the nominal bandwidth. Thus for me :

Doing so the LTSpice simulator is not consistent with the Bible ! : the Radiotron Designer's Handbook Ed.4 . Of course at the time ( the 50's !) the -3dB boundary was not yet in use but is is easy to compute and compare the results which are totally consistent with the formula above.

This is the only objection I have ever found when using this simulator.

I attach a special note on the subject for those interested .. and go on on my side with the analysis of this preamp.

 

[YellowBoots]

To moderate what I say about the reliability of LTSpice I must say that there is an inconsistency left for me in the evaluation of the low cut-off frequency of a partially bypassed cathode resistor.
I speak about this because the stages of the SOLDANO preamp are all in this case with their 1µF capacitor across the Rk resistor.
LTSpice seems to stick to the good old mistake of Fc = 1/ 2 * π * Rk * Ck as if the pole of the RkCk network was the same as the one of the stage itself ! If we remember the expression of the gain A = µRA / (RA + rak + (µ + 1) Zk ) the low cut-off frequency will be the value for which the denominator will reach 1.414 x its value in the nominal bandwidth. Thus for me :

View attachment 1100249
Doing so the LTSpice simulator is not consistent with the Bible ! : the Radiotron Designer's Handbook Ed.4 . Of course at the time ( the 50's !) the -3dB boundary was not yet in use but is is easy to compute and compare the results which are totally consistent with the formula above.

View attachment 1100251
This is the only objection I have ever found when using this simulator.

I attach a special note on the subject for those interested .. and go on on my side with the analysis of this preamp.

I’m surprised that your models seem to fail here. I have found LTSpice to be reliable in every way, excepting the routine user error. I haven’t gone too far into the tube modeling though. It seems to me that frequency analysis of a simple R || C circuit would be no trouble for it. Is there possibly a shortcoming in the tube model?

 

[ltournell]

I would like to begin the "study" with an analysis of the first stage, comparing the "color pencils + abacus" method with the result of simulation. It would be the start of the discussion.

First let us start with the data entries, that is the reference schematics we have !

Now the simulation equivalent, retaining the true supply voltages for each stage and stopping the simu to the FX SEND output, the recovery stage and CF seemingly not interfering with the shape of the output.

I choose to input a 50 mVp signal (-29 dBV as Richard Kuehnel would put it) : it can represent about the lowest attack level of a neck single coil on the first string of the guitar knowing that this level next decreases and raises then the problem of sustain !

On the load line here we see :
the headroom = 1.5 Vp (close to 0.5 dBV) - no PU generating such a level this input is used as a clean input as well.
gm = 1.2 mA/V
hence we deduce rak ( Ri in the datasheets) = 83300 ohms
the real ! low-cut-off frequency is 66 Hz ( see formula and discussion above) NO ! 82Hz (Thank you Andrew and Pete !)
the gain in the bandwidth on an infinite load is 72.5 ( 37.2 dB)
the output impedance ( 220k// 83300) = 62150 ohms

Here to simplify I disconnect the 2 nF treble peaker across the 470k resistor since we use a 1 kHz signal for easy spectrum analysis and this capacitor makes computations too difficult for the abacus !


we can compute the gain on the load ( 470k + 500k )

Now if we compare to the result of the simulation we see :

For an input of 50mVp that is 100mVpp the "abacus" computation above gave 68.1 x 100mVpp = 6.81 Vpp on the load and the simu gives 6.87 Vpp. So this is a good start.

If we connect again the treble peaker across the 470k we get 6.54 Vpp , which is normal considering that 1 kHz is already a treble sound.

 

[ltournell]

I’m surprised that your models seem to fail here. I have found LTSpice to be reliable in every way, excepting the routine user error. I haven’t gone too far into the tube modeling though. It seems to me that frequency analysis of a simple R || C circuit would be no trouble for it. Is there possibly a shortcoming in the tube model?

Yes I agree. It is probably more a problem of modelling than an "internal" fault of LTSpice algorithms. I know that this question of low cut-off frequency on partially bypassed cathode is relatively ignored ( when Randall Aiken from Aikenamps ) wrote his white papers in the early 2000s he himself was not aware of this.

EDIT : no , it was just a problem of me mistaking the plate resistance with the plate total load !

 

[Pete Farrington]

This cathode bypass calculator uses the RDH4 procedure http://bmamps.com/CapCal.html

 

[ltournell]

This cathode bypass calculator uses the RDH4 procedure http://bmamps.com/CapCal.html

Thank you Pete ! This helped me remember that in my formula here :

RA should have been taken as the total load that is RA // RL . Now everything becomes consistent : with the data you give in your link
the -3 dB found there is at 46.4 Hz , with my formula above it is at 45.8 Hz and LTSpice gives 43.44 Hz. Otherwise said all three the same value. The so-called mistake with LTSpice is not the tragedy I fancied , just me forgetting what the true load was. I am reassured

The cut-off frequency of the SOLDANO I give here above is therefore not 66 Hz but 82 Hz that is the sixth string of a guitar in standard tuning.

 

[ltournell]

The problem when you write many things in the same post is that a few mistakes can be made and this forum only lets you 24 hrs to correct them. !
For instance here above I wrote for the first stage : the output impedance ( 220k// 83300) = 62150 ohms
Of course I should have written : the output impedance at 1 kHz (= module of 220k// ( 83300 + (µ+1) x Zk) ) = 62150 ohms with Zk = Rk // ZCk

In fact the conclusion here is that this peculiar 3rd stage acts just the reverse of what I expected . In itself it produces EVEN harmonics as we haven seen but integrated in the chain it helps producing almost perfect symmetrical square waves thus making he ODD harmonics absolutely prevailing ! Simulations can be made to see how alternate schematics for this stage could have achieved the same thing and it would be interesting to simulate the DUMBLE ODS overdrive channel to compare.

I think I'll simplify all this with a synthesis table for the whole preamp in a new thread.