Power Amp Sim - A little experiment, needs testing

[lucem]

Hi there,

I'm going to start off my first post into this exquisit forum by sharing something I have drawn up today.
I haven't simulated or verified it as of now, but it should work as intended and so I decided to post this
without running tests first - maybe someone is curious and faster than I am at testing this.

It's a little circuit emulating the interaction between a tube power amp and a speaker load
(so as to complete the JFET preamp craze into something made out of silicon that totally behaves as close to the real
thing as possible with the technology) - including supply voltage sag.

I made a PDF with the schematic and parts list, I haven't done a layout as of yet.
Adjustment description and general design explanation is in the PDF.

http://www.aronnelson.com/DIYFiles/up/lucem-powerampcabsimv1.pdf

Note that I use low powered general purpose BC550C / BC850C npn BJTs here.
They're biased so they're pretty linear and don't influence the sound too much - however,
it should readily be possible to replace them with 2N7000 MOSFETs or at that low current consumption
even J201s with high IDss values (20mA and up), or BF245C with proper biasing.
If someone wants to try that, there's a start - you would however need to operate them in the
pentode region of their transfer curve (which is not that different from a BJT at the set bias).

You should not attempt to bypass the emitter resistor with a capacitor, this will most likely cause nasty distortion.
One thing to try there is to bypass with a lower value resistor + capacitor to provide a little more gain.

In terms of overall gain, the circuit should on average over the frequency range have a gain of 10 (in regard to input level
across the input transformer's primary). You can adjust the trimmer before the input transformer to reduce the input level.

If anyone is faster in actually building this thing than I do, I'd love to hear about the results.
Any suggestions, corrections and possible mods are of course welcome.

Happy tinkering!

[PRR]

Welcome!

Interesting brain-work. Nice drawing.

Little doubt it will pass more signal than smoke.

Some choices I am not understanding.

800mA fuse, but the 100r at R6 means it "can't" ever suck more than 15V/100r= 150mA even if everthing else were dead-short. Without looking further, we would suspect that other stuff is not a fancy short, but takes some voltage. With some hasty guesses, I doubt it sucks over 38mA. As a 1/2W R6 is cheaper than a fuse, a 2W R6 is cheaper than a fuse and clips, I wonder why a fuse.

Loading T1 600r Secondary with 10K RB1 RB2 does not use the available signal power (leaves signal gain on the table). It also affects transistor linearity, but I am not sure how I feel about that.

RB flows at least 0.29mA (to 15mA!) into Q1 Q2 Bases. BC550 can have hFE 110 to 800. So even at full 47K the emitter current tries to be 0.29mA*800= 230mA. This would cause 23V drop from 15V supply in R6 100r, so that can't happen. There is DC NFB happening. RB divided by hFE 110-800 gives an effective device resistance 59r to 427r, working against R6 100r, gives collector voltage of 5.5V to 12V (off 15V supply), *just* from hFE variation. Voltage much lower if RB is trimmed-down. If only the 2*10k are in the RB path, near 1V to 5V.

OH WAIT. I am neglecting RE, and the note for RB. If you want RE at half supply, nail the Bases a bit above half supply. 7.5V in RE 220r plus 1.7V LED is 26mA. 13mA each transistor. This also causes 2.6V drop in R6. Vce is 5V.

Bases will be near 8.1V. Base current will be 13mA/hFE or 16 to 120 uA. This current in the 10K RB1 RB2 causes 0.16V to 1.2V drop. The 1V variation over hFE is "not large" compared to the target "1/2 supply at emitters" (7.5V). I would just nail T1 CT to 8 or 9 Volts and eliminate a trim.

If hFE varies 110 to 800, why pay for 0.1% precision RB1 RB2? Even though you select the transistors "matched", you can't get 0.1% match.

I still have real doubts about these RB12 resistors. At 13mA, the re is about 2 Ohms, so the rb is 220 to 1600 Ohms. 9/10th of the drive voltage is lost.

Note that a Tube amp has no such drive-wasting linearizing grid resistors. (The 1K-5K grid resistor is moot for audio because grid resistance is super high.)

T2 primary is 600r CT. Given 5V Vce, peak signal current is 17mA. The amp runs mostly class A. I do not think it will "sag". In fact it can't, because if one side cuts-off, the 4 Ohm re=re shifts to 2r+220r, and gain vanishes.

Output at T2 10K side appears to be 40V peak, quite large.

This side could drive around 10K, but the 1Meg pot won't drive even a medium-Z load well. It sure will not work well with to 10K-22K Line inputs on modern consoles and DAWs or ADCs, it wants a buffer.

T2 could be 600:600, all secondary loading scaled down, then R5 pot could be 10K, a workable value for going into modern inputs.

Or given that you have R3 10K, make that the output pot. Only at 40V output, it may need a >10:1 divider after it. Even 33K+1K, which gives light loading on the 10K side, and nice low Z on the output.

> tiny power amplifier with a reactive load in current mode and thus will behave very much like a tube output stage with an output transformer would

True; but is T2 a Guitar Amplifier transformer? Transformers are available in a VERY wide range of performance. "Line" transformers (typical 10K:600 parts) may be fairly good. "Guitar Output" transformers are generally just-good-enough, or even "poor" performers.

At first I thought the intent was to emulate speaker interaction. That is a BIG part of many older g-amps' sound. Differences between Fender, Marshall, and Gibson or Ampeg. Different with every speaker choice. The load network you have gives an adjustable drop-shelf response above 110Hz. (CL pretends to add a pole at 111KHz, but you will find far more than 100pFd in any practical transformer, so this isn't doing anything worth the solder.)

There are simpler ways to build a hi-Z transformer audio path. GE Transistor Manual 1964 is filthy with them. It thinks they are transistor radio output stages, but one was kidnapped to a rock-n-roll stage and is now infamous as the Deacy(sp?) pedal.

[lucem]

Hi,

thank you for your reply.

> 800mA fuse
Simple - 800mA slow blow will allow to also supply the whole preamp section before this, a Hi-Z buffer after, and is dimensioned high enough so inrush currents do not cause it to blow - this circuit only makes sense as a part of a larger system, and the fuse calculation is adapted to that.

> True; but is T2 a Guitar Amplifier transformer?
No, it's not. However, it doesn't need to be, since the selected part has a pretty similar frequency response curve in the audio band than the whole output transformer-speaker system in a guitar amp (when driven actively; when not, it's dismal). This in combination with the output network (which can be made more complex than this, given) with its frequency dependent loading in theory (that's why it needs verification and possibly value tweaks) produces an overall output response of a guitar speaker, and since it's a current mode amplifier the amplifier is going to react on this load.

> BC550 can have hFE 110 to 800
800? Where would you find that? The datasheet says otherwise, and I've never come across one with that high gain. BC550C (!) is in the range of 100-150 typically, and if you read the schematic I provided with the notes on it, it calls for a matched pair with a hFE of around 100. The same goes for the base resistors, the .1% is an indication for the people who like to spend money that you can get resistors that closely matched, but other than that it reads as "as closely matched as possible". Both components matched on both sides of the long tailed pair mean maximum balance on both branches, and at the operating point far in the linear region with a collector current of 10 mA each.

>  I am neglecting RE, and the note for RB. If you want RE at half supply, nail the Bases a bit above half supply. 7.5V in RE 220r plus 1.7V LED is 26mA.
This goes into the right direction, but still is a bit off. First of all, for 15V input supply, Vcc is going to be around 12V; for 12V supply, Vcc is going to be around 10.4V.
The 100 Ohm resistor and the following whole circuit form a voltage divider of 1:6 ratio, which is loading the supply enough so the power sag effect will actually appear.
So, emitter voltage is actually 5V-6V, the rest of the calculations for static currents are in essence correct.
Btw, the 50R coil resistance on each side of the branch of the output transformer's primary are also non-neglectable in this configuration.

> Loading T1 600r Secondary with 10K RB1 RB2 does not use the available signal power
T1 is a 10K:10K isolation transformer which is just a phase splitter. T2 is 600R:10K.

About the gain calculations; for any meaningful AC signal, the transistor gain is limited by the collector and emitter impedance, which is closely a factor of 2.5; the output transformer will create a voltage gain of about 4 (3.7 to be exact), which in sum means a gain of 10. So a 1V signal across the input transformer will create an output of about 10V at the output transformer. The output wants a buffer, that's true,; however, for anything following the output network not to affect the frequency response, a buffer is necessary anyway (as already stated above).

Does that make sense?

> The amp runs mostly class A.
Nope. It's a Class AB push-pull design. For any signal causing a significant current flow in comparison to the static current consumption, it will load the power supply and sag.

Now, given all that, it might be possible that values need some tweaking, and maybe some minor modifications need to be made, but in general, the circuit should work as intended.

Thanks again for the constructive criticism.

[ggedamed]

> BC550 can have hFE 110 to 800
800? Where would you find that? The datasheet says otherwise, and I've never come across one with that high gain.

Here is one place: http://www.mouser.com/ds/2/149/fairchild%20semiconductor_bc550-320122.pdf. Page 2.

[lucem]

Hm, to conclude: I stand corrected.

In that case, BC550A is sufficient, matching of hFE values is still required.
Or substitue with a different npn model with generally lower hFE values (still matched).

Mistake on my part.

edit:
For clarification, in theory one could increase the base resistors to reflect a higher hFE rating, but this would change the loading characteristics of the secondary of the input transformer and would require great care (right now the input secondary load is overall somewhere around the theoretically ideal 10k at the expected trim value of 12k). So, in essence, touching the base resistors is a bad idea unless one would go about and completely change the operating point of the long tailed pair along with it.

edit2:
I will probably redimension the whole circuit to reflect the deficiencies mentioned. This might take some time though, one has to work inbetween having fun at the drawing board