Welcome to Solid State Guitar Amp Forum | DIY Guitar Amplifiers. Please login or sign up.

March 28, 2024, 05:41:43 AM

Login with username, password and session length

Recent Posts

 

Fet version of the Sharp channel of the Bogner Triple Giant

Started by KMG, February 01, 2011, 04:54:32 AM

Previous topic - Next topic

KMG

Schematics (pdf)

Stage by stage frequency response in comparison with the original tube version


Ready to transfer PCB (pdf)

Upper side assembly (pdf)

Bottom side assembly (pdf)



Sample (gain control only by volume knob on gutar)

Steve Conner

Can you explain the purpose of T1, D1 etc? I'm guessing the diodes are to emulate grid current in a tube, but why the transistor?

KMG


Steve Conner

Very cool! :) I like your other projects too. Thanks for taking the time to make them available in English.

How is the temperature stability of your FET building block?

KMG

QuoteHow is the temperature stability of your FET building block?
Right now I did some tests. I used a soldering iron to heat the flange of the transistor up to 70 ° C. Changing of static drain voltage in range from 20 ° C to 70 ° C was 4-6 V (for different stages). It`s about 3-5% of change.
"Don`t worry, be happy" ;)

J M Fahey

Steve: I guess you refer to T10 to T13. They are for *real* biasing of the fets, instead of the brain-dead ROG method of using trimpots on the *load* side.
The diodes are for grid_to_cathode diode simulation; as Peavey uses in Transtubes.

KMG : congratulations, you've taken the harder path of *really* cloning tubes.
Brilliant results are your well deserved prize.
Sound is *very* good.  :tu:
Thanks for waiting for your friend to play, he did a very good and tasty job.
Using >160V plus accurate , precision biasing is the way to go, not those klunky 9V monstrosities.
Thanks again. :tu:

Steve Conner

#6
JM: No, I figured out what those ones do. I meant the transistors used as emitter followers hanging off the drains.

My best "triode" so far is a cascode of a dual-gate RF MOSFET and a MPSA42, running off 250V. I previously experimented with a JFET/MPSA42 cascode, but it had curves more like a pentode, and needed a negative bias arrangement similar to KMG's. It didn't sound too exciting when overdriven, either.

By dividing down the "plate" voltage and feeding it back to the MOSFET's second gate, the curves are changed into triode-like ones, and it seems to run OK with the source just grounded, no negative bias voltage is needed. The divider ratio sets the "mu" of the tube.

A zener on gate 1 protects it from ESD in the reverse direction (which the RF MOSFET needs badly!) and simulates "grid" current in the forward direction. Peavey seem to have a patent on that.

I haven't tested it for thermal stability yet, though. (From the datasheet, the 2SK216 has a really nice tempco at "tube amp" drain currents, which will be hard to beat.) Nor have I run it alongside a 12AX7 as KMG did, or even listened to it. It's still sitting in a drawer at work, so the above should be treated as just random musings. :)

rowdy_riemer

I wouldn't be too hard on ROG, JM. A big part of their goal is to keep it simple, and I think they do a pretty good job. Obviously, a more complex circuit topology will do a better job of triode emulation than ROG's fetzer valve concept. I do agree, though, that having the trimpot on the load side doesn't make a lot of sense.

Steve Conner

Well, it was seeing the Fetzer Valve that got me started messing around with this stuff.

The pot in the drain kind of makes sense. FETs that bias up hotter will have more gm, so will need a smaller drain load resistor to get the amount of gain the designer wanted.

I just feel that sticking more voltage into the circuit will somehow make it sound more like a tube circuit.

rowdy_riemer

Well, I think the gain-to-headspace ratio is more important than the gain itself. I see a need for designing circuits arround a 9v supply for stomp boxes. But if your supply voltage is that low, your gain needs to be low to have the same gain-to-headspace ratio as the emulated circuit. Frequency response is important, too, and having a lower value drain resistor will reduce the output impedance of the gain stage, which might have a noticeable effect on the frequency response. I would rather have  a trimpot for the source resistor, and, if possible, select fets that bias well for the drain resistor used. I agree that when it makes sense, a higher voltage supply is best.

KMG

Accurate emulation of grid current often means more than a shape of clipping.
Look how tube stage respond on "burst" input signal.



K1 - Anode voltage
K2 - Input voltage
K3 - Grid voltage
K4 - Cathode voltage
Output signal  "floats" in time at a constant input level.
Average current varies almost 2 times.
Amount of changes depends on interstage circuitry (impedance, time constant etc).
That's why guitar players often say that tube sound "live" and the sound of semiconductors "dead" (uniform).
This refers not to the color of harmonics but to reaction on the sound picking.

Steve Conner

#11
Yes, I fully agree. The grid current charges the interstage coupling capacitor, and that changes the duty cycle of the clipped waveform.

It adds interest to the tone by changing the harmonic content, in exactly the same way as the PWM effect on an analog synth.

I experimented with DC servo loops to make JFET stages stable with temperature, but as an unwanted side effect, it adjusted the clipping duty cycle to 50%. That made them sound really dull and "dead" as you say.

J M Fahey

Sorry, didn't want to hurt anybody, all the contrary :(
Beyond that, we are on a technical forum here, and if somebody says 2+2=5, hurt feelings or not, I'll state the result is 4.

As of ROG projects, seeing a typical one gives me headache, the same kind I get when I hear somebody singing absolutely out of tune or playing very fast, missing 80% of the notes.
Let me explain that:
The Fetzer itself is the only serious part, and I was not referring to it.
Its designers know the triode curves they wish to approach, know what they actually have (pentode-like Fets), know all the theory, how things work, do whatever's needed to bend one into the other, brilliant !! :tu:

*But* cloning a tube design , and finding there a typical 1/2 12AX7 with 1K5 on cathode, 100K on plate, 260V B+ , 160V on plate, and substituting it with a Fet, using 1K5 on source ..... (why 1k5 ?).

There's 3 possible answers I can hear:

1) "because I decided I want that FET to pass 1mA and, look at the curves, this particular one passes 1 mA with any drain voltage between 3 and 30V and 1.5 Vgs , so [1.5V/0.001A]=1500 ohms".
*That* is the "serious" answer and the one I expect  :tu:

2) "Duh?? :o"
Even this one is somewhat acceptable, sort of.  ???
Nobody was born knowing, but studying or experimenting anybody can learn and advance.With some care it may end up good, in a near future

3) "Because Fender uses it"  :duh
C'mon, you are kidding, right? This is magical thinking of the worst kind.
Why the worst? Because to the uninitiated it "sounds" logical, even powerful (Mojo) , pure pseudoscience.
Add this to the power of internet, and another impossible to erradicate myth is born and spread. :duh
I repeat, I find *good* design work the Fetzer approach and the work of all others who took it seriously, such as KMG, Lab Series, Mensur, Trace Elliott, Yamaha, Roland, Univox (Westbury) and the granddaddy of them all: Randall.
1K5? 9V? Forget it !! (unless you buy 100 fets to farm the 5 to 10 amongst them which might be happy with that value)
*None* of the above mentioned established manufacturers uses anything even approaching that Fet-unfriendly combination.

Dear Steve: please reread this statement:
QuoteFETs that bias up hotter will have more Gm, so will need a smaller drain load resistor to get the amount of gain the designer wanted.
I think you are mixing two different things here:
1) In FETs, transconductance (which is part of the gain equation) varies with current, with higher current yielding higher Gm, so I agree that on this case, biasing a given FET with higher current (> Gm) will need a smaller drain resistor (<R) (given a fixed +B) Since both terms of the product vary in opposite ways, they tend to compensate, so the gain variation is not as large as might be imagined.
In this case what you said applies fully. , same fet, different biasing.
2) Fet manufacturing is inconsistent, yielding products with important parameter variation, specially Idss, Gm and Vp, all three being related.
Fets with higher Idss and Vp usually have lower Gm.
So in the case of different, unselected FETs, the ones with higher Vp and Idss will pass more current through the famous 1k5 resistor (which can't be modified for Religious motives), so will need a lower Rload (or R drain) setting and will have much less gain, because we are finding lower Gm and Rload at the same time.
I'm afraid what you said applies fully to the other case (always the same FET) and not to the ROG case, random unselected FETs.
In these designs, Rload (the preset) is adjusted simply to have 4.5V on Drain (another mistake), and this increases the gain spread instead of somewhat compensating as in your earlier suggestion.
No wonder no manufacturer uses this approach.
What do you think?
Of course, the usual disclaimer, I might have overlooked something very important, in which case I'll hear whatever can be said about that.
Good luck and happy FETting. :)

PS: a last thought: Leo Fender didn't even "design" those values, and there's nothing "magical" about them, he simply took them from manufacturer's datasheets (excellent !), which show suggested plate values for 12AX7 triodes between 47K and 220K, for voltages between 100 and 300V, and the corresponding cathode values needed to optimally bias them.
If you insert , say, a 12AU7 there, same circuit and +B bias values will be already wrong, even being also a triode; why those values would work for a FET, a very different device, and even worse, using +9V is beyond me. Ugh !!

Once again, congratulations KMG on your brilliant work, doing original research and taking care of every detail for a very accurate emulation.
I guess that that Vd sampling you do , reinjected at the input, bends the unforgiving "pentode" curves into softer "triode" ones.
I also congratulate Steve on his brilliant idea of using a dual gate Fet; using one for amplification duties and the other for curve corrections.  :tu:

Steve Conner

#13
Hi JM,

You could be right about the spread between units. Somehow I thought the pot in the drain was the right thing to fix that, but I might have it backwards.

I think I figured out the "scalable diode" thing. The thread on it was in Russian, but the schematics and scope shots were pretty clear. If the amplifying device were a BJT, the scalable diode would be a Baker clamp: its purpose is to make sure that "grid" current begins before the amplifying transistor saturates. This hides the fact that the transistor's saturation behaviour is different to a tube's.

I think my curve correction approach does something similar. It gives the same rounded bottoms to the clipped wave, because it's a physical model of a triode. It has the same saturation behaviour, so there's no need to hide it.

I was inspired by the "Trioderizer" concept that feeds the curve correction signal into the gate of a regular FET along with the signal to be amplified. The Trioderizer has great characteristic curves, but because the curve correction signal is a kind of shunt feedback, its input impedance is nothing like a real tube's.

I messed around with arrangements of multiple FETs, but the dual gate MOSFET made the most sense. They are still widely available, sadly in tiny SMT packages.


KMG

QuoteI messed around with arrangements of multiple FETs, but the dual gate MOSFET made the most sense. They are still widely available, sadly in tiny SMT packages.
These dual-gate MOSFETs is mainly targeted at the AGC control in the RF-frontend and work with a very low scale input voltage.
Now I`m looking at this high voltage N-channel depletion mode (normally-on)  lateral MOSFET.
http://www.supertex.com/pdf/datasheets/LND150.pdf