Printer2 said:
These are both things that have been on my mind as well. So far, I don't have conclusive answers.
To Printer2's question, we have Shanx personal experience: he can hear a difference between these JFET-based guitar amps, and more recent ones that are built around op-amps.
To Shanx question: I've spent some time tinkering with the compound JFET-BJT stage in LTSpice, and at this point, I don't think I have conclusive results.
At first I thought the BiFET stage was doing exactly what we'd been expecting. When I ran different simulations, I was less sure. It was beginning to look awfully close to "too perfect".
One of the problems I'm facing is how to tell exactly what the amp is doing. The idea of overlaying input and output signals seems simple at first sight, but not so simple once I began to think about it. There are a number of thorny issues to deal with.
For example, if I use the proper values of coupling and bypassing capacitors as used by Traynor engineers, then there will be lots of phase lead at low frequencies through the BiFET stage. This means the output of the BiFET pair will never properly overlay the input signal, even if there's no "squashing"! They won't overlay because the phases won't match.
The logical solution to that seems to be to enormously increase cap values. I did that, but now there's a new problem: the (virtual) circuit takes a long time to settle down to its quiescent operating point. Hit it with the simulated guitar signal too soon, and the signals won't overlay simply because of DC offset.
There's still another problem: we know that good-sounding tube stages do suffer signal-dependent "bias shift". If the BiFET pair is doing this, I'd like to see it, but the methodology of hugely increasing coupling caps will also muck up the time taken to "bias shift". The simulation won't show the right results, in other words.
So how can I tell exactly what nonlinear/imperfect behaviour the BiFET circuit is causing? I still don't have a good answer. (We can see gross changes simply by plotting input and output separately, but I'm looking for small changes, of the order of a few percent.)
Another issue is this: from the time I saw that JFET-BJT direct-coupled pair, I've been wondering just how much open-loop gain it has.
We can make a quick estimate for the open-loop gain of a single common-emitter BJT transistor stage (with a fully bypassed emitter resistor, and no load on the output) using this equation:
Voltage gain, Av ~= 40 x (DC volts across collector resistor).
If we make the first-pass assumption that there will be 15V DC across the collector resistor (centre-biased, +/- 15V supply), we can expect the transistor to have an open-loop voltage gain of around 600 times (55 dB).
I thought the JFET might have some voltage gain as well. It's being operated at very low current (look at the value of R4 and the fact that R4 only has about 0.6 V across it).
So, to a crude first approximation, I thought the compound pair might have an open loop gain in the vicinity of 65 dB.
This is much higher than I'd been hoping for. If open-loop gain is very high, then the stage will start to suffer from the same problems as an op-amp: too-clean when not clipping, and too-harsh when it does clip.
So I went back to the sim. I reduced R1 down to 1 ohm (and massively increased C2) to see what the simulated open-loop gain of the compound stage was. In a nutshell, LTSpice thinks the JFET has a slightly-less-than-unity voltage gain, but the BJT has +50 dB of gain, so the pair has an open-loop gain of 49 dB.
I reset R1 to the Traynor-designed value (470 ohms). The closed-loop voltage gain was now +32 dB.
That means there is (49-32) dB of negative feedback. That's 17 dB!
On the one hand, 17 dB of NFB is nowhere near as bad as an op-amp. An op-amp set to 32 dB closed-loop gain might have 100 dB open-loop gain, so it will have 68 dB of NFB at low frequencies.
On the other hand, 17 dB of NFB is quite a lot. It will do a pretty good job of straightening out any curvature in the JFET characteristic. Maybe too good a job?
I've thought of three different ways to lower the open-loop gain of this BiFET stage, and tried all three in the simulator:
1) Replace the PNP BJT with a P-channel JFET or MOSFET. This dramatically drops open-loop gain. but finding suitable P-channel JFETs in 2022 might be tricky.
2) Insert a string of small-signal silicon diodes between the emitter of the PNP BJT, and the positive rail. This idea is borrowed from KMG's amazing LND150 MOSFET preamp; each diode adds a few ohms of nonlinear resistance, so if you string about four or six of them in series, you lower the voltage gain of the BJT, without linearizing it at the same time. (You have to rework several resistor values to ensure optimal biasing.)
3) Insert an un-bypassed emitter resistor between the +15V rail and the PNP BJT's emitter. This will cause local NFB in the BJT stage, which has the unwanted side effect of making the BJT stage more linear, but also has the desired effect of reducing open-loop gain, so that the input stage JFET's non-linearity isn't wiped out by too much NFB.
I guess the nutshell version is that I don't think I have a good handle on the Traynor-designed JFET/BJT compound stage yet, in terms of how it behaves with a guitar signal.
Can't you compare the output signal to that of a valve guitar amplifier in simulation? After all, you don't want the output signal to look like the input signal, but like the output signal of a valve guitar amplifier.
By the way, when you make the coupling capacitors really large, like 1 kF or 1 GF or so, there will be practically no DC shift and no need to recover from it during the simulation time. I assume you run the transient analysis from a bias point found in an operating point analysis, like SPICE does by default, not from zero.
By the way, when you make the coupling capacitors really large, like 1 kF or 1 GF or so, there will be practically no DC shift and no need to recover from it during the simulation time. I assume you run the transient analysis from a bias point found in an operating point analysis, like SPICE does by default, not from zero.
Mind you, 1 F is 1 femto in SPICE, not 1 farad.
The late PStar had a default DC leakage time constant of 1 second, so a capacitor of 1 farad had a leakage resistor of 1 ohm. I haven't seen anything like that in any other simulator, but then again, I haven't used all of them.
The late PStar had a default DC leakage time constant of 1 second, so a capacitor of 1 farad had a leakage resistor of 1 ohm. I haven't seen anything like that in any other simulator, but then again, I haven't used all of them.
I don't think it will look like a valve amplifier output, but comparing to an op-amp based guitar amp there are some differences that I think are noticeable. The JFET/BJT pair is one gain element that is interesting, and there are other stages on that amplifier with common source Jfet configuration, which contribute to the overall sound.
I think there is a way to import data from a .wav file into LTSpice and export the result to a .wav file, so if you can make a completely clean and unclipped recording of an electric guitar, you could listen to the effect of virtual guitar amplifiers.
Sorry to have been away so long, guys. Unfortunately I've been dealing with some lingering health issues that may come down to the vague and unsatisfactory umbrella term of "long COVID".
Marcel, I suspected (but didn't know for sure) that SPICE starts with a steady-state operating point before doing a transient analysis. That's good to know. Thank you.
I quite agree that a good valve amplifier is the benchmark, but can't think of any objective ways to assess closeness to that goal.
To my ears, at any rate, a single triode gain stage of the sort used in so many vintage Fender amplifiers doesn't have any audible magic to it. There is some cumulative effect on "clean tone" as the signal goes through the entire chain of tube stages, so that by the time the signal reaches the speaker and turns into sound wave, "tube cleans" sound better than "op-amp cleans".
Even that is subtle enough to be hotly contested, with plenty of people on online music forums claiming that there is no audible difference, that their ten-year-old Digitech RP-something digital multiFX pedal sounds the same as a tube amp. And then there are others going to the other extreme and taking the position that even a single solid-state device in the signal chain somehow poisons the sound and ruins it.
It would be really nice to have some objective (rather than subjective) ways to assess "goodness" of a solid-state guitar preamp or power amp. So far I don't really have a handle on how to do this.
I wonder what the big boys do? When the people at Yamaha, or Atomic Amps, or behind the Flamma FS-06, were developing their amp models, how did they test them? Purely by ear? Or did they have some sort of sophisticated test suite that can tease out valvey goodness from solid-state harshness?
On this thread, I think we've got past the zeroth-order effect, though: we now have pretty good objective evidence that a typical e-guitar plugged into a typical op-amp gain stage will indeed suffer from repeated harsh clipping, for a long enough period at the start of each note to be audible.
The remaining question - how to do away with that harsh clipping, and how to add "tubey" or "valvey" sound - is a much bigger, and much more open, one.
If one must use an op-amp, it may be that something as simple wiring an anti-parallel pair of red LEDs across the op-amp feedback resistor and setting gain conservatively might provide a substantial reduction in harshness. Rather than let the output of a TL072 go to +/- 3V and slam into a brick wall, maybe the somewhat softer (logarithmic) transfer function of the red LEDs kicking in to clamp only the early big signal peaks to maybe +/- 1.8V would be an improvement?
I have some other JFET-related ideas kicking around in my mind, which will have to be tested on an actual breadboard; LTSpice sims of those ideas have failed utterly, but I think the problem is with the way the JFET is modeled, not with the idea itself.
I was recently listening to an Anderton's Music video showcasing some expensive Universal Audio hardware and software for recording. "Danish Pete" (Peter Honoré) played some licks on a Stratocaster that caught my ear - he was picking some single notes quite forcefully, with the software amp model set for clean tone, and you could actually hear a brief little "chirp" of distortion at the start of each note. It was a musically acceptable sound, though, rather like the airy transient "chuff" at the start of every note blown on a flute. Not harsh like a clipping op-amp.
-Gnobuddy
Marcel, I suspected (but didn't know for sure) that SPICE starts with a steady-state operating point before doing a transient analysis. That's good to know. Thank you.
I quite agree that a good valve amplifier is the benchmark, but can't think of any objective ways to assess closeness to that goal.
To my ears, at any rate, a single triode gain stage of the sort used in so many vintage Fender amplifiers doesn't have any audible magic to it. There is some cumulative effect on "clean tone" as the signal goes through the entire chain of tube stages, so that by the time the signal reaches the speaker and turns into sound wave, "tube cleans" sound better than "op-amp cleans".
Even that is subtle enough to be hotly contested, with plenty of people on online music forums claiming that there is no audible difference, that their ten-year-old Digitech RP-something digital multiFX pedal sounds the same as a tube amp. And then there are others going to the other extreme and taking the position that even a single solid-state device in the signal chain somehow poisons the sound and ruins it.
It would be really nice to have some objective (rather than subjective) ways to assess "goodness" of a solid-state guitar preamp or power amp. So far I don't really have a handle on how to do this.
I wonder what the big boys do? When the people at Yamaha, or Atomic Amps, or behind the Flamma FS-06, were developing their amp models, how did they test them? Purely by ear? Or did they have some sort of sophisticated test suite that can tease out valvey goodness from solid-state harshness?
On this thread, I think we've got past the zeroth-order effect, though: we now have pretty good objective evidence that a typical e-guitar plugged into a typical op-amp gain stage will indeed suffer from repeated harsh clipping, for a long enough period at the start of each note to be audible.
The remaining question - how to do away with that harsh clipping, and how to add "tubey" or "valvey" sound - is a much bigger, and much more open, one.
If one must use an op-amp, it may be that something as simple wiring an anti-parallel pair of red LEDs across the op-amp feedback resistor and setting gain conservatively might provide a substantial reduction in harshness. Rather than let the output of a TL072 go to +/- 3V and slam into a brick wall, maybe the somewhat softer (logarithmic) transfer function of the red LEDs kicking in to clamp only the early big signal peaks to maybe +/- 1.8V would be an improvement?
I have some other JFET-related ideas kicking around in my mind, which will have to be tested on an actual breadboard; LTSpice sims of those ideas have failed utterly, but I think the problem is with the way the JFET is modeled, not with the idea itself.
I was recently listening to an Anderton's Music video showcasing some expensive Universal Audio hardware and software for recording. "Danish Pete" (Peter Honoré) played some licks on a Stratocaster that caught my ear - he was picking some single notes quite forcefully, with the software amp model set for clean tone, and you could actually hear a brief little "chirp" of distortion at the start of each note. It was a musically acceptable sound, though, rather like the airy transient "chuff" at the start of every note blown on a flute. Not harsh like a clipping op-amp.
-Gnobuddy
I think you're right, on both counts.
Some years ago I heard an audible reduction in harshness simply by adding a JFET common source stage at the input of a guitar preamp I was developing. The JFET stage had an unbypassed source resistor to control voltage gain. Predicted THD was too low to be audible!
So what was I hearing? I guessed that maybe I was hearing something the JFET was doing to tame the big +20 - +30 dB initial transient signal.
That was one of my early clues that we couldn't draw correct conclusions about guitar preamps by using steady-state sine wave testing.
-Gnobuddy
Sterkte, as we say over here.
It does unless you specifically ask it not to.
It's difficult because you don't want the output signal to be an accurate scaled copy of the input signal. None of the traditional quality metrics for HiFi amplifiers are applicable. With controlled listening tests, you could at most assess if two guitar amplifiers sound the same, not if they sound good, because there is no clear definition of good. I think it automatically becomes a matter of taste.
No idea, but @JMFahey wrote about long tweaking and listening sessions with musicians in some other thread - unfortunately I don't remember which thread. The musician plays the guitar, expresses in non-technical terms how he/she/it feels about it, JMFahey tries to translate that into technical terms and into a required circuit change, changes the circuit accordingly, the musician plays the guitar again and so on until the process converges and everybody is happy.
Thanks.
yes, that´s about it, patience and interaction.
It´s hard for Tech or Engineer to resist the urge to tell the musician:"no!!! that´s WRONG!!!" n 🙄
Two examples from "the First World"
* in a Rolling Stones Movie,they are recording at some fancy studio,and remember, all were "BBC Engineers", weaned on recording Classical Music, or at worst , Jazz or Big Bands
Since the camera rolled continuously, tons of "between the scenes" footage.
At a given moment, you see the Studio, a large one, where different amps and of course the drummer, were "all in the same room" but separated from each other by chest height acoustic panels, to somewhat improve mic separation.
One Engineer points at a large amp, head and 4 x 12" cabinet, probably a VOX, and says "use that one, it´s the best" and Keith Richards, somewhat sheepishly answers, pointing at a little knee high combo (an AC15/AC30?) :"ummmm .... if you don´t mind, I´d prefer to use THAT one.
Expressing some disbelief, the Engineer shrugs and answers: "ok, IF you insist", not too convinced about it.
* just anecdote, but might have some substance, apparently some early The Who tapes were returned from mastering or Record Cutting to original recording Studio because "the guitars were distorted". 😉
yes, that´s about it, patience and interaction.
It´s hard for Tech or Engineer to resist the urge to tell the musician:"no!!! that´s WRONG!!!" n 🙄
Two examples from "the First World"
* in a Rolling Stones Movie,they are recording at some fancy studio,and remember, all were "BBC Engineers", weaned on recording Classical Music, or at worst , Jazz or Big Bands
Since the camera rolled continuously, tons of "between the scenes" footage.
At a given moment, you see the Studio, a large one, where different amps and of course the drummer, were "all in the same room" but separated from each other by chest height acoustic panels, to somewhat improve mic separation.
One Engineer points at a large amp, head and 4 x 12" cabinet, probably a VOX, and says "use that one, it´s the best" and Keith Richards, somewhat sheepishly answers, pointing at a little knee high combo (an AC15/AC30?) :"ummmm .... if you don´t mind, I´d prefer to use THAT one.
Expressing some disbelief, the Engineer shrugs and answers: "ok, IF you insist", not too convinced about it.
* just anecdote, but might have some substance, apparently some early The Who tapes were returned from mastering or Record Cutting to original recording Studio because "the guitars were distorted". 😉
They say the Star Trek episode with the alien woman, painted green,
came back from developing "corrected" to flesh-tone.
---this was apparently an ongoing problem....
https://www.startrek.com/article/trouble-keeping-her-green
I'm reminded of a shipping box with the phrase "Rattle OK."
"Clipping OK, if it's the GOOD kind."
"Clipping OK, if it's the GOOD kind."
Easy to solve: subject him to one too many dubious quality burritos and 5 too many drinks and next morning he WILL look green.
At least I do 😉
Red eyes and bloated eyelids add to the Alien effect 🙂
Jokes aside, yes, in the roll film era machines were usually calibrated to render "flesh tone" as "healthy" European hue, which sometimes made the rest of the scene look weird.
ONE of the reasons colour Magazines and Pro Photographers preferred transparencies, say Kodachrome (best), Ektachrome (average) or other brand equivalents, NO "Lab Technician" or auto machine "opinion" influencing results.
That and the much higher optical Dynamic range.
Even with the early DSP modeling, I think engineers had to start somewhere, and there was a certain raspy-ness to ''digital overdrive''. They could get a decent clean sound (albeit with some latency) but I remember a reviewer saying the overdriven amp effect had sonic artifacts to it that would be difficult to EQ out. It would seem that in the industry eventually it was realized.. to get beyond the presumption of steady sine wave testing being the reference point, it`s ''okay'' for basic testing. A better indicator of the circuit response would be using a spectrum analyzer and feeding it a range of sine waves at various frequency and amplitudes, looking at the impulse responses. Then feeding it .wav file of actual guitar signals. It does ultimately come down to whether the player themselves like the sound and the feel of the preamp response. I play guitar, and I do some design in audio circuits, but when focusing in on the tech side I usually recruit the ears of other players and let them try things out for comments and feedback. Sometimes they can come up with things that I had not realized, or came up with some useful suggestions or feature. It`s not design for and by the masses, but it helps to get a different perspective on what I am trying to achieve.
Shanx said:
So the biggest and most important ingredient in the recipe is this: access to a pool of good guitarists with excellent ears.
Unfortunately, that's an ingredient I don't have! 😀
-Gnobuddy
I remember all the ones I tried in the 1990s and first decade of the 2000s sounding harsh and kazoo-like when overdriven, and "too clean" when set for clean tone.
IMO, Line 6 products of that era made the worst sounds on the market. On top of the harsh/too clean problems, they also chose to add an incredibly unpleasant nasal honk to every product they sold.
I can only guess that this came from trying to emulate the frequency response of some guitar speaker that had a huge midrange peak. But something went wrong, and somehow, the process was badly botched.
However the Line 6 "honk" originated, it condemned a decades worth of their amp emulator products to sound absolutely horrible.
Line 6 ad copy promised the sound of dozens of wonderful-sounding vintage tube amplifiers in the box. IMO, what you actually got was the sound of dozens of kazoos in one box.
Those horrid-sounding Line 6 amps sold very well, though, so I guessed the manufacturer laughed all the way to the bank.
Evidently I was a slow learner, as I allowed myself to be convinced by peer pressure to buy three different Line 6 products over the years. All three sounded horrible. Each time, I thought the fault must have been with my guitar or my playing, so I spent weeks trying to get a less-horrible sound out of the infernal things. Each time, it sounded just as bad on the last attempt as it had on the first.
To my ears, the first affordable DSP amp modeller I heard that had good clean tone was the Boss Katana.Shanx said:
IMO, when set correctly, clean tone from the Katana can sound very "tubey". Even more tubey than my rather sterile-sounding Fender '65 Princeton Reverb reissue, which is an actual tube amp!
The second affordable DSP amp modeller I've heard good "tubey" clean tones from is the Flamma FS06 Preamp. IMO this is a shockingly good little device, at a startlingly low price point. I can't say enough good things about it.
I've recently tried running the Flamma into the Katana (with a DIY signal attenuator in between). I set the Katana set to its "Acoustic" channel, which is supposed to produce a fairly flat frequency response. I find the results surprisingly good - the Katana actually makes a very good powered speaker for use with the Flamma.
To my ears, the clean tone from solid-state Fender modelling amps is nowhere near as good as that from the Katana or some of the Flamma models.
This is rather surprising, considering that the Fender brand achieved its fame based entirely on clean guitar tones.
But there it is. The Fender modelling products I've heard have about the same sterile too-clean sound you get from an op-amp.
Case in point, both the Fender Mustang Micro and the Flamma include models of a clean Fender amplifier. The Flamma version sounds more tubey and more like a real Fender tube amp to me...and the little Fender Mustang sells for 50% more!
We can't forget the absolutely staggering improvements in microcontroller performance per dollar since the early Line 6 products in the late 1990s and early 2000s. In that time, we've gone from feeble and expensive 8-bit microcontrollers to cheap and powerful 32 bit microcontrollers.
Back then, I'm sure tube models had to be stripped down to the minimum to run on the feeble hardware.
Today it's possible to run much more sophisticated amp models on powerful (but affordable) hardware. But you still have to have a good software model to run. I don't think Yamaha or Fender software models are there yet.
The Flamma is a mystery to me. I've never heard of the brand until quite recently. Out of the blue, they suddenly release this remarkably good-sounding, remarkably affordable DSP modelling pedal.
Where did those remarkably good amp models come from? Were they licensed from somewhere else?
-Gnobuddy
Before everything shut down during Covid, I used to go open mic jam nights fairly regularly. I met quite a few players, and I sometimes would bring along some of my electronic creations for them to try out. Some of these players were semi pro or pro, had good ears and would be happy to try out some new gear..it was a good type of venue to do this, as it's more for fun and relaxed than an actual paid gig. People like to be asked their opinion, and as a bonus I sort of got known as a local tech and people did bring me repair business, and some bought some of my own designed gear. Leo Fender could barely play a few notes on guitar, but he relied on musicians feedback to create and innovate new products. He used to go see Dick Dale play (famous Surf music guitarist). Dale played clean with reverb but needed a more powerful amp with lots of clean headroom..Leo wondered why Dale would need 80 + watts until he went to a show filled with noisy screaming teenagers... Hence the 100 W Dual Showman (with 15 inch speakers) was conceived.
I took some guitar lessons at a Los Angeles area studio many years ago, and the teachers there all had superb ears.
But since then, I haven't encountered other instrumentalists with the same sort of good ears. I must be going to lower-quality jams. 🙂
Imagine if he'd attended some of the Beatles concerts!
The tragedy is that almost every single guitarist from the next several decades ended up with serious hearing loss, tinnitis, or both.
-Gnobuddy
Last edited: