Thank you for posting the schematic. I can see the intent of what they designed. There are some similarities to the JFET CS stage with soft limiting diode that I proposed on the other thread https://www.diyaudio.com/community/threads/building-a-ss-guitar-amp.381758/page-19
Without the soft limiter, you still get some dynamic bias shift due to inherent non-linearity of the JFET curve, but adding this further non-linear feedback seems to audibly enhance the 2nd order harmonic content as well as contributing more to bias shifting.
A lot of the JFET based guitar amps seem to do quite well (at least to my ears) on their clean channel, but we saw a couple of examples where the distortion channels simply use diodes as "hard clippers" (Randall and Traynor Bloc series).
I think you can get JFET circuits in the ballpark of 5 to 10% THD , playing close attention to the incoming signal levels, not trying to drive it too hard and designing the JFET stages for relatively low gain. I have been playing through the 3 FET preamp as the "distortion channel" for my Traynor Bloc 100. The first two stages are JFET and the last stage is Mosfet. That last stage is intended to be the "hard distorter" if pushed with a large enough input signal but I have a pot there to attenuate the incoming signal and it can get to "clean-ish" performance. That pot acts more like an Overdrive control. There is a volume pot between the first and second JFET stages , which I consider as a "Gain" control (and a Boost switch for the second stage source bypass cap). As I am using this into a full guitar amp, which has the TMB tone controls, another thing I would consider adding is a single knob tone control, probably be for high frequency cut. I could add switching such that the third Mosfet stage is bypassed and the overdrive is more subtle. Hope to have some time to do more work on this little project. I plan to post something up on a new separate thread if there is any interest. Cheers!
I had a chance to consider this a bit further. So we have bias excursion (dynamic bias shift) that occurs within the preamp due to large signals and resulting single sided squishing/partial rectification of the preamp stages. The preamp is "decoupled" from the "real" power amp down stream and the bias shifting coming from any power supply sag would be essentially a variable that is normally fed back from the power amp. But it can be configured in a way where the preamp output is monitored and based on its final signal level out. We derive a DC signal which is another control voltage, and it will affect the supply rail voltage of the FET preamp. The threshold levels can be set on the detector such that at low preamp output volume there is no sag introduced, but at higher output levels where significant distortion would be occurring within the preamp/amp, the DC Sag variable will increase. We can scale the "virtual loading" as well as the recovery time constant for V(sag). Maybe a bit too much to incorporate control knobs for this, but perhaps a switch to select a couple of options?
I can only echo this sentiment. That broad transition between clean and crunch needs to happen over at least a 20db range. This is both from personal experience (somewhere there's my attempt at a subtle, nice sounding multistage diode array and is supported by the analysis by others - e.g. Ampbooks on the Soldano)
On +/-15V rails that means non-linearity kicking in at no more than 150mV (peak). If this is all done in a single stage.
I'm developing doubts that that single stage with pre- and post- EQ is going to be sufficient to get quite what I'm looking for. But I'm going to (Real Soon Now (tm) - in my usual lazy way) build a second clipping board (it's two wires and a little hot glue to swap it out) with 20db of non-linearity
Definitely use variable controls in a prototype, to find all the "good" settings.
Use Schottky diodes (germaniums may work as well but are less obtainable) to get more "distortion points" over the +/-15V range, and/or use higher voltage op-amps (5532/5534) for a little bit extra dynamic range, or just use higher voltage discrete transistors. I may be doing this myself. I already have ideas in LTspice.
Oops. What I do know is that Si diode (0.7V) to 4.7 V (ish - can't recall exact value) chain is not enough.
This was the output of the Mark I board at the sweet spot, but it kicked in too late and then hard clipped quickly (at about 5v )
Didn't sound as good as a single-12AX7-into-6L6 amp. That's got a master volume and just the preamp alone sounds better. Back to the drawing board!
(I've just got to find the original sketch to remember the resistor values - or tear it out and check)
I will be using Shottkys for the first step on the next attempt. The voltage rails, op amp, Rout (1k) and Rload (c.80K) are fixed as the clipper board replaces the back-to-back red LEDs in a Vox Pathfinder 15R. It makes audible testing easy as I've added a DPDT centre-off switch to toggle between diodes, nothing and the clipper board. See this thread
Last edited:
Would you consider adding another asymmetric waveshaping stage, in addition to your new diode/resistance board you are planning? Having a look at the schem for the Vox amp, if I am correct there's a R7 on IC1B which is 470 K Ohm, I'd suggest just connecting a single Si diode across that and check the output of stage. With the series resistance of 100K or so in the NFB, there should be a gentle squish on half the wave cycle. It is similar to my JFet with Soft Limiting shunt feedback. It may be very subtle but I hear a difference comparing with/without the diode in the feedback loop. It should only add a low % of H2 harmonic..warming up the sound.. Cheers!
One idea is to subtract the amplified and the raw signals (properly normalized first of course) and only look at the difference.
Another idea is to look at the frequency spectrum in addition to the amplitude (fourier transform). Comparing raw signal, SS amped signal and tube-amped signal in the frequency domain might be informative. This can be done on a computer if somebody digitizes the signals with a reasonable sampling, no need for special equipment.
Apologies - I am not setup with any measurement equipment right this minute, just found this forum but I probably will be.
P.S. I am new here, glad to see such an illuminating discussion. I have some background with circuits and measurement in general but I am new to e-guitars. Learning a lot from this forum already.
P.P.S. I have not read the whole thread yet (standing now at 13 pages) so I am reacting to the beginning of it before I forget.
OK, I'm going to have to try that. Fun bit is I've already done the "gain mod" so I need to think about interactions that arise.
(for those at home, that's the pot in this schematic, which is not there in the stock amp.)
Yes that extra gain pot may affect the response as it's in parallel with the NFB when Q1 is turned off, thus lowering the overall gain. When Q1 is on you probably won't see much, if any 1/2 cycle soft limiting as R8 is much lower resistance than the R6+R7 leg. On my JFET shunt limiter I made the resistance across the diode variable, which would be on this circuit equivalent to the R7. Varying that R affects both the overall stage gain and amount of 1/2 cycle squish. I am trying to figure out how to impart some dynamic bias shift on this particular op amp stage, it's a bit difficult because you'd have to do some changes on the input configuration. I have some ideas but haven't built up and tested the circuit to prove it out yet.
Yes but does it have to be that way? Can all signal processing be handled by ss, so what comes out is ready for hi fi ideally transparent amp?
This is my ideal. But I know next to nothing, and whenever I mention it near guitar people, they just roll their eyes. Could somebody explain why?
Is this outright impossible? Is it possible but very impractical? Does it just go against tradition?
A large paper cone has dozens or hundreds of resonances. Which vary in F with drive level.
If we actually understood that, a supercomputer might simulate it. But it is still early days in level-varying response theory.
A 1930s-style paper cone is still cheaper and smaller than the emulation.
There's no tradition there are countless tastes, styles and technique to play an electric guitar, and modern amplifiers have at least three general modes well known as Clean, Crunch and Drive, and each of them has many variations in ties with a cabinet or another with the type of voicing done through a typical 5 band eq or through a cabinet , preffered with miking or no miking ....There are thousands of guitar players with different styles and gear...Now there are thousands of effects, harmonizers,noisegates, both hardware and software, SVT 's are now replacing most of the old style hardware in a recording studio.. Quilter labs signed a contract with a SVT manufacturer just a month ago to be sure he gets in the guotar bussiness with the right foot...
Nothing's clear and simple.You like it or not...
The RunOffGroove Thunderbird has an interesting circuit that does a bias shift (between U1b and U2a). I had a little problem with too fast bias shift with my asymmetric clipper so ended up shorting out C11 to keep the asymmetry (too lazy to work out a suitable value given the variable impedance down stream of it)
Actually there are many guitar amp simulators that have been sold over the decades. I recall the Line 6 Pod from the 1990s, often called the kidney because of its shape. I just googled guitar amp modeler and Google shows me ads for current ones (all basically based on DSP "models" or whatever of real tube guitar amps), priced from $200 to $1800 or so. Yet actual tube guitar amps are still manufactured and sold.
Yes, the thing with the Vox second stage is I would be looking at the resistor and cap in series to the inverting input. It would be better if they swapped positions, so you could add a diode that shunts a bit of current away from the cap but only at a certain threshold..then the DC offset at the cap would discharge through the resistor to virtual ground at the inverting once the signal settles below the threshold.
Looking at the Vox circuit again, and I think with the single Si diode across R7 doing soft limiting on 1/2 cycle, it may end up having some bias shifting on C11 at that stage output. That would be an interesting test point to check out. I think a shift in the order of 10 to 100 mV is enough to give a nice effect without becoming too clippy or farty, but you need to look at the charge/discharge time constants through the LED1 and LED 2 (plus series resistance).