Member
Joined 2009
Paid Member
I think this concept (small valve guitar amp into big clean power amp) has a lot of promise. But, at least in my mind, there are still a number of unanswered questions.
For example: Do we really need an output transformer? Do we really need a speaker load on the small guitar amp, or will a dummy load resistor and speaker emulation do? What sort of valve circuit do we need in the little amp to make the big amp really sound like a proper valve amp? Et cetera, et cetera.
And the biggest question of all: How good will our fake butter be, using this approach?
I think it could be a lot of fun (and very useful to many of us) to try and answer some of those questions on the electronics workbench.
It seems VOX is still doing something similar. Their current MV50 series guitar amplifiers have two vacuum triodes in the preamp, followed by a solid-state 50W class-D power section: Review: Vox MV50 Series Amplifiers - Guitar World
The VOX MV50 series are tiny, inexpensive, palm-sized amps. Matrix Amplification currently has a much bigger approach (420 watts of SS power preceded by four triodes, i.e., two 12AX7s): Vintage British 800
Incidentally, VOX in 2002 certainly wasn't the first to try this approach. Around 1965, i.e. about thirty seven years before VOX, "Gar" Gillies of Garnet Amplification developed his Herzog. The Herzog was basically a Champ (12AX7, 6V6, output transformer, etc) designed to be used like a stompbox, feeding a bigger guitar amp.
I don't know if Gillies was the first to try this or not. It's possible someone else tried the same thing even earlier, who knows.
More on the Garnet Herzog here: "Gar" Gillies' Herzog® - All Tube Guitar Effect - garnetamps.com - Home of the Garnet™ Amplifier Company
-Gnobuddy
I finally found the forum where KMG describes his circuit.
Fet version of the JCM800
His method was to build a 2203 preamp with 12AX7, then build one initially with 2SK216, and later, with LND150. He tweaked the LND150 version until the waveforms matched the 12AX7 version.
An excerpt, where he explains some things
I don't know if LND150 have as much Vgs variations as Jfets, my understanding is that mosfets are more consistent in that regard.
So it seems likely KMG's use of individually buffered negative supply for each gain stage has more to do with the grid current simulation rather than device variation. And if you look closely, the 1st and 3rd gain stage use the same buffer circuitry. It is the 2nd gain stage that is different, and only slightly so.
There is a -12v rail and the buffers reduce this.
Accounting for the nominal 0.6v drop in the base-emitter junction, I calculate the 1st and 3rd stage buffers to deliver about -3.0v and the second gain stage about -2.8v So that's not a whole lot of difference.
It's clear KMG was obsessive about matching the 2203's 12AX7 gain stages precisely.
I'm thinking you can just derive a solid -3v rail and do without the buffers. Would simplify things greatly.
The "bottom" source resistors closely match the actual 2203 cathode resistors and the "top" source resistors are 180,270, and 200 ohms.
I'm thinking a 1k5 "bottom" source resistor and a 220 ohms "top" source resistor would be a good starting point, to match a typical Leo Fender gain stage.
Agreed.
The nice thing about an actual transformer is you automagically get some bandwidth limiting under overdrive conditions and some peak limiting and good stuff like that.
Since we are talking about building a preamp with a strictly line level output, I was thinking along these lines:
1. Emulate a Fender Blackface style 3 stage preamp in high voltage, using KMG techniques.
2. Couple this to a low voltage stage, which emulates a valve amp's power amp.
This stage is not intended to drive a speaker, just for additional "tone-shaping".
One possibility here is the BJT differential stage as presented by voltwide.
Perhaps a simple Jfet stage is adequate. Or a Jfet "mu-follower".
Another possibility here would be to use a low cost 600 ohm microphone transformer, driven to a max of maybe 200 milliwatts or so, using Jfets or bjt, whatever.
I'm talking about the cheapo Xicon $2.89 transformers, nothing exotic.
These transformers have a rather dismal bandwidth of 300-3400hz at full rated power of 200 mW, but approaching full audio bandwidth at lower power. Just what the doctor ordered, or so it seems. It might also sound like total rubbish, I don't know.
3. Couple the output of step 2 to a low voltage "de-nastifying" circuit or speaker simulation circuit.
4. low voltage buffer to line out.
It's entirely possible that step 2 is not needed at all.
Here is a little Power Supply Challenge:
This project is going to need roughly +300VDC and -3VDC for the KMG circuitry and at least +9VDC for the low voltage side (preferably at least 12v). The KMG circuitry is only going to draw maybe 10mA or so total. The low voltage side maybe 50ma tops, probably a lot less.
I do have an extra valve amp power transformer I could use.
It's a Hammond 270DX with 550v center tap secondary, a 5V filament supply and a 6.3v filament supply. There is no tap for fixed bias.
I can get the 300VDC with the secondary winding, with the usual techniques.
I can get the -3V from the 5V winding.
But the 6.3v winding is only going to give around 8VDC or so, accounting for diode drops.
Ideally, I'd like to series connect the 5V and 6.3v filament windings, rectify that and get around 15VDC.
But then I'd lose the -3V that the KMG circuit wants.
I have a ton of wall warts that I can scavenge for an extra transformer to generate the -3V, and that may be what I need to do.
Unless someone has some suggestions.
Perhaps take the 6.3v winding and feed into a voltage doubler and some good filtering?
Or is there a way to derive the -3V from the 550vct secondary in a way that doesn't ripple?
(unlike a fixed bias situation in a valve amp, this negative supply will need to supply some actual current.)
I had found and read through that, but you got more out of it than I did. Thank you!
I seem to remember a comment by KMG to the effect that all stages were now using the same -3 V bias. Can't find it now, though.
I too was hoping that a single fixed negative rail would do the trick.
If I understand correctly, it seems KMG uses one source resistor to set the (DC) source current, and a second one (with series cap) to set the AC gain.
If so, it should be possible to scale up both the negative voltage, and the DC source resistors, so as to keep the same DC drain current, but with increased stability against FET parameter variations, temperature swings, et cetera.
Meantime, the AC coupled source resistor will still set the AC gain. (Okay, there will be a slight change because the AC impedance is actually both source resistors in parallel, but that can be allowed for by decreasing the AC coupled resistor a bit when the DC one is increased, if it turns out to be necessary.)
I assume that these, along with the (1/gm) internal dynamic source resistance of the MOSFET, are setting the AC voltage gain of the stage.
I can distinctly hear a reduction in bass when I turn up the output power from my Super Champ XD, which has a pretty tiddly output transformer.
Interestingly, I don't hear the same loss of bass from my Princeton Reverb, which has an almost identical push-pull 6V6 output stage - but a beefier output transformer.
I like both amps. For me the sliding bass filter created by the too-small OT isn't a make-or-break "feature".
But other guitarists may feel differently.FWIW, I have never noticed a corresponding reduction in treble as I drive the SCXD harder and the deeper bass starts to go away.
Yeah, but that's exactly what makes it a fun experiment to try.
I hope you do try it, and tell us how it sounds to you!
This is the route I'd probably take:Here is a little Power Supply Challenge:
This project is going to need roughly +300VDC and -3VDC for the KMG circuitry and at least +9VDC for the low voltage side (preferably at least 12v).
1) 12V, 2A switching power supply:
inShareplus 12V LED Strip Power Supply 2A 24W, Wall Mounted 12V Switching Power Supply, 110V to 12 Power Supply for LED Strip Light with 5.5/2.1 DC Female Barrel Connector to Screw Adapter - - Amazon.com
2) 45V - 390V DC-DC boost converter for the high voltage: Amazon.com: SMAKN(R) High Voltage Boost Converter 8-32V to 45V-390V 110V/220V ZVS Capacitor Charging: Home Audio & Theater
3) 3.3V, 5W switching power supply for the -3.3V rail: https://www.digikey.com/product-det...cs-usa/KTPS05-03315U-VI-P1/62-1234-ND/5820199
Using new-fangled switching power supplies saves quite a bit of weight, time, and cost, compared to the old "big 60 Hz transformer with lots of big caps and diodes" approach.
Having to use two separate power supplies and a DC-DC converter board is a little annoying. But if you mount a single IEC inlet on your project enclosure, mount both switching power supplies inside, and wire them directly to the IEC inlet, all the mess is internally contained. From the outside, you only use one AC cord, and plug it into the one IEC inlet.
However, if you don't like the switching supply route, and go with the 60 Hz big-iron approach instead:
That was going to be my suggestion as I was reading your post. I like voltage doublers, they are a great solution for relatively lightly loaded supply rails. With today's affordable huge-value electrolytic caps, I don't think it will be too hard to filter it well enough for your needs. Maybe add one extra RC filter stage after the basic doubler. Even a series 10 ohm resistor gets you a good bit of filtering (25 dB at 60 Hz, 50 dB at 120 Hz, 75 dB@180 Hz) if you have a 4700 uF filter cap following it!
My little 6AK6 push-pull amp gets several different HT rails from one single 48V Hammond transformer: -75V, +150V, +225V, +300V. All through the magic of voltage doublers, triplers, and quadruplers, with a few $1 surplus-store camera flash capacitors (270uF, 330V : PHOTO-FLASH CAPACITOR, 270UF 330WV | All Electronics Corp. )
Heater power comes from a thrift-store Sony 8.1V switching power supply, with a suitable power resistor in series to drop the voltage to 6.3 VDC at the heaters.
The idea to use an inexpensive off-the-shelf contemporary power transformer, and triplers and quadruplers go generate valve-friendly HT rail voltages, came from Roly Roper's writeup ( http://www.ozvalveamps.org/ava100/ava101lamington/ava100-1ps800x566x8.jpg )
Roly was a very smart, very knowledgeable, very helpful man whom I only knew through his prodigious number of helpful posts to the Aussie Guitar Gearheads Forum, and a couple of emails back and forth. If he hadn't been on the opposite side of the earth, I would have loved to have met him and had a long chat over a cup of coffee.
-Gnobuddy
Mebbe use a -5V DC rail from a typical USB switching power supply, and tweak DC bias source resistors to suit?
Any reason why that won't work?
-Gnobuddy
Regarding matching waveforms between tube and SS circuits, I have reservations that using fixed sine wave test signals would be adequate. Some effects such as sag and blocking recovery are dynamic and probably should be tested with more complex waveforms. Maybe two tones gated on and off, or level modulated high and low, to see how dynamic signals evolve over time. Perhaps similar to how one would observe the behavior of a compressor.
Also, maybe just record a clean guitar into an A/D and play it back as a standard test signal to compare waveform behavior.
In addition, maybe circuits like diode wave shapers could be used to some effect. In terms of trying to modulate the time behavior of tube circuits, IIRC, there may be one or two expired Peavey patents on that.
Also, maybe just record a clean guitar into an A/D and play it back as a standard test signal to compare waveform behavior.
In addition, maybe circuits like diode wave shapers could be used to some effect. In terms of trying to modulate the time behavior of tube circuits, IIRC, there may be one or two expired Peavey patents on that.
Last edited:
I agree, and I think those are all valid reservations to have.
I too am very skeptical about all claims to emulate "tube sound" with semiconductors, because we have all heard so many terrible-sounding circuits that claim to sound "tube like". It's "The Boy Who Cried Wolf" effect. We've all been tricked before, and don't want to be tricked again.
However, KMG's guitar clips are what made me sit up and take notice. There are some links early in this thread.
I consider good clean tones to be an acid test for SS guitar amps, one at which they usually immediately fail. Here is a clip of KMG's FET version of the Bogner Ecstacy: http://milas.spb.ru/~kmg/files/proje...mg_xtc_mix.mp3
What do your ears say?
To me, it sounds like a very acceptable guitar tone, probably from a semi-hollow guitar with considerable sweetness of its own, regardless of amp. Still, the subtle "valveyness" at the start of the clip, and the little bit of crunchy overdrive later, sound pretty good to me.
Until I hear it for myself, from a guitar in my hands and a speaker in the same room, I won't know for sure how it sounds & feels. But I am sufficiently impressed by this (and other) clip of KMGs work not to dismiss it out of hand.
-Gnobuddy
Last edited:
It sounds pretty close, but not exactly like a good tube amp at least not quite like the one I have here. But not bad at all for that piece of music. I would like to hear it without so much reverb and with some additional playing techniques.
Also, I downloaded the mp3 and tag data in the file says, "Denn + Ibanez RG3120TW;" That would suggest its a solid body guitar with humbuckers. RG3120 | Ibanez Wiki | FANDOM powered by Wikia
To me, the onset of overdrive sounds a little more abrupt, and the actual overdrive tone sounds a bit "drier" and "crisper" or "crunchier" than the valve amps I'm familiar with. BUT - I have heard newer higher-gain valve guitar amps that have a similarly dry / crisp sounding mild overdrive.
I think the main thing is that I would not immediately be utterly disappointed if I plugged into a guitar amp and it sounded like that clip. That is already a huge improvement over 99% of the SS guitar amps out there. Heck, 99.9%.
Agree about the large amount of reverb and delay - they are very good at covering up some solid-state harshness in an amp. When I'm too tired to haul my real (valve) guitar amp to a jam, and instead play through my small solid-state P.A. system / acoustic guitar amp, guess what I do? I take my semi-hollow guitar (an Agile AS-820), and slather on lots of reverb and delay. I also use a graphic EQ pedal for some tonal fine-tuning and "de-nastifying".
It can sound pretty good, even with no valves involved. But not quite as good as through a good valve amp...
Fake butter, but it doesn't taste too bad, right?
I heard the humbuckers (I thought it was an ES-335 or something similar). But I have never heard a solid body guitar sound that sweet / woody before...not even Mark Knopfler in his early Dire Straits period managed that with his 'Strat.
Now I'm really curious if the MP3 tag is correct (and my ears are wrong), or the other way around!
-Gnobuddy
For what it is worth I tried the signal transformer thing once. I did not find much widening of frequency response at lower levels. Sounded like crap. What I do also find of interest is the phase inverter. The imbalance some have. Also the difference in bias, you get 1.2k (Vox) down to 470R (Fender).
Thanks for the info, never tried a signal transformer myself, so now I can scratch that off the list.
Another interesting thing about the long tail inverter is how the global negative feedback feeds into the tail.
One would think the second input of the LTP would be the place to connect the NFB, similar to solid state amps with a differential input stage.
Was this a conscious decision, or did someone at Fender make a wiring mistake and Leo liked the sound, and it became standard?
Cathodyne imbalance is easily simulated in a MOSFET "SourceODyne", just make the drain resistor slightly different than the source resistor.
Since cathodynes have so much negative feedback, they also produce very little distortion, so the MOSFET is pretty much an identical-sounding replacement.
But a long-tailed pair is an entirely different story. They are nonlinear, also with a vaguely S-shaped transfer function, and produce audible (mostly odd order) distortion.
I don't know if there is a way to re-create that same signature distortion with FETs. It would be an R&D project, with no guarantee of success. Might be fun trying, though.
-Gnobuddy
One would, indeed!
Merlin Blencowe, in his valve preamp design book, says the bizarre Fender feedback arrangement let Fender make minimal modifications to their pre-existing eyelet board pattern, and was probably picked for that reason. (That's from memory, hope I didn't get any major details wrong.)
I've never built a guitar amp with an LTP phase splitter, or one with any negative feedback. If I ever do both those things in the same amp, the feedback will go where it should go, to the control grid of the other triode in the LTP. No need to endlessly replicate Leonidas' long-ago kludge!
-Gnobuddy
Connected to the cathodes, the NFB connection is out of phase with the 2nd LTP grid. Before LTPs were in use, IRRC, cathode feedback already existed so it wasn't a new idea. One can kind of think of it as providing feedback into the 1st LTP rather than the 2nd. That kind of makes sense to me because the 2nd LTP grid is often connected by capacitor to where or close to where the NFB returns. In that case the NFB drives the 2nd LTP grid and cathode up and down together, so most of the effect should be on the 1st LTP who's grid is not connected the the NFB path. That's my quick take on it anyway. Feel free to correct me if its wrong.
I believe you are exactly right, and supposedly it was exactly because Fender already had a layout for negative feedback to the cathode of one triode in a 12AX7 that they picked the weird bastardized NFB connection they did. It avoided having to change their board, and retrain their workers.
If you like to look at it this way, you can think of the second device in the LTP as a cathode follower; NFB is fed to its grid, and its cathode then drives the cathode of the first device.
In other words, exactly the same as cathode feedback, except you have a buffer (cathode follower) to help apply the feedback!
In the end, what matters is now how we make sense of the story inside our own heads, but rather the precise mathematical equations that tell you the relationships between the various voltages. Our intuition isn't always reliable, but Kirchoff's equations are.
LTSpice or similar software is an easy way to see what the math works out to be, without having to sit down and write out loop equations and solve them.
As far as I can see, there is no good engineering reason for Leo's weird negative feedback topology. It was just a kludge.
That doesn't make it bad, it just makes it weird. Like manufacturing a car that has the rear windshield installed upside down and inside out. It looks weird, but it still keeps the rain out!
-Gnobuddy
Honestly, I'm rather skeptical, firstly because Fender was already using unequal anode resistors to improve PI balance, and secondly, because the best-balanced long tailed pairs (which are in op-amps) don't muck about with feedback to the long tail.
Also let's not forget Leo was using extremely crude +/- 20% resistors from the E6 series, meaning his LTP balance would vary wildly from one amp to the next, up to a worst-case 40% mismatch. This even if he had the most magical feedback scheme in the world. There is simply no way to ensure LTP balance when the triodes and resistors Leo used had these sorts of huge statistical parameter variations.
There is also the fact that a perfectly balanced phase splitter would still inevitably be followed by imperfectly matched output valves, and an imperfectly symmetrical output transformer. The output of the amp would still contain lots of imbalance.
There are also Internet tales about improving Fender cathodyne balance, including claims that the 1.5k cathode bias resistor upsets the balance of Leo's cathodyne, because it adds to the cathode resistor, but not the anode resistor. This is nonsense if you look closer: 1.5k is about 3% of the value of the typical Fender cathodyne cathode / anode resistors. Even if you used modern 5% resistors, the cathodyne is already out of balance by up to 10% because of resistor tolerance, and the extra 1.5k is swamped by the statistical uncertainty in the values of the cathode and anode resistors. Back in Leo's day, with +/- 20% resistors, this was even more true.
There is a great deal of Fender lore that doesn't stand up to closer scrutiny. I mean no offense to anyone who's believed some of it, but if we are to make any progress in guitar amp design, we have to sort out the myths from the cold, hard, often unromantic, engineering reality.
Personally, I prefer the sound of a push-pull amp with at least 1 dB (approx 12%) imbalance, either in the phase splitter or output stage. The slight imbalance prevents complete cancellation of even harmonics, giving a sound signature a bit closer to a single-ended amp.
-Gnobuddy
If we follow Fender's phase splitter evolution in the 1950's we first see the paraphase inverter, for example the Super 5C4, no global NFB.
Next up, the floating paraphase, Super 5D4, global NFB to grid of 6L6.
Next, the cathodyne, Super 5E4, global NFB to cathode of stage preceding the cathodyne. This was the standard technique used in hi-fi amps of the day.
Next is the LTP. The most famous example is the Bassman 5F6A. Here we see the global NFB to the tail of the LTP, and that arrangement became the industry standard.
But take a look at the 5F6 (not 5F6A). Here we see the global NFB to the bottom of the tone stack!
The layout of 5F6 is substantially different from the previous 5E6. The filter caps got moved to the "doghouse", for one thing.
However, the eyelet board of 5F6 and 5F6A is identical.
The tail resistor was a convenient place to attach the NFB, having rejected the idea of going to the bottom of the tone stack.
It is a very minor change to the layout, almost unnoticeable unless you're really looking for it.
So that would support Merlin Blencowe's theory.
But now, one wonders what was wrong with the 5F6?
Was it oscillating? did it sound bad?
The other changes from 5F6 to 5F6A:
5F6: type 83 rectifier, 5F6A: GZ34
5F6: 100R screen resistors, 5F6A: 470R
5F6: 1k5 grid stoppers, 5F6A: no grid stoppers
Fender probably ditched the 83 rectifier for economic reasons, either it was unreliable or too expensive.
The change in screen resistors makes sense.
Removing the grid stoppers probably done to save a few cents.
But why change the NFB to a configuration that will now have the "scratchy presence control" problem? This was later changed to the 4k7 resistor/25k pot configuration, which doesn't work as well as a presence control IMO, but eliminates the scratchiness.
The only thing I can think of is the 5F6 NFB configuration would have additional phase shift problems due to the coupling cap between the tone stack and the LTP. Also possible weird interactions based on tone control settings.
This may have caused oscillations.
Tying NFB to the tail would eliminate this, without having to change the eyelet board.
Since this might have fixed the oscillation problem, Leo may have said "yank out those grid stoppers so we can save some dough."
While the 5E6 cathodyne amp has the same phase shift problem, perhaps Fender was trying for more NFB in the 5F6 than the 5E6.
Many hi-fi amps at that time were quite similar to the 5E6 topology and used a lot of NFB, but also used higher quality output transformers so they could get away with it.
I must admit I have a heck of a time trying to calculate global NFB mathematically in valve amps because you have to account for the output transformer impedance and such. So I just play with the feedback resistor until I get about 10db NFB, and call it a day.
For what it's worth, I have done just that and it works fine, as it should. I can't say it improves the tone or anything like that. There is no savings in components either way.
I will say that when you wire it "correctly" and stare at your schematic, it's a lot easier for your brain to figure out what is going on.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.