Peavey Rage 158: calculating TDA2040 power amp gain

[aion]

I'm working on adapting the preamp of the Rage 158 and had a question about the TDA2040 power amp stage. I'm trying to work out what the total gain is between C14/R4 and the preamp output jack, with the idea of substituting a gain recovery stage that would serve the same function without the power amp.

Here's the relevant excerpt:


(Full schematic here)

The TDA is not dissimilar from the LM386, a power amp on a chip with some limited control over the parameters using external components. The datasheet doesn't say much about how to control it other than some basic circuits, all with fixed gain of around 30. This write-up has a lot more info about the particulars of adjusting the gain, with R24 and R29 setting the gain ratio, more or less (numbers referencing the Peavey schematic).

Where I'm getting hung up is how R28 and R26 affect the gain calculation. The preamp out jack takes the signal from the TDA2040 and cuts it down to 9% using R20 and R21... but how do I calculate the gain level of the TDA2040 before this cut? All the example circuits show the speaker grounded, perhaps with a Zobel network close by, but in this case the negative side of the speaker is going back to the power amp (or taking the output from another part of the power amp - I'm not sure which direction I should be thinking of the signal flow here).

The 68k gain resistor (R24) is also much higher than the 22k from most of the example circuits, and the 330R resistor to ground (R29) is much lower, so it seems like the gain would be significantly higher than the chip is capable of, which based on the page above is around 33dB.

[PRR]

The hidden detail is: this is not normal voltage-gain feedback. It is mixed feedback. It ALSO watches the load _current_. The result is that the output impedance is not "zero", as it typically is for general and HiFi speaker amplifiers.

Lessee.

NO-load: We have 68k/78k at the input, times 68,330/330 as the no-load voltage gain.

Shorted load (may not be safe): 2730/330r.

But with a 3.2r load it works out somewhere inbetween.

EDIT!! fergot image.

[Rob Strand]

There's a bit more to it than just gain, especially if you are trying to mimic what is happening around the power amp.

A tube amp has a high output impedance whereas a solid state amp has a low output impedance.  When you connect a 12" guitar speaker to a tube amp and a solid state amp you get different frequency response.   The signal through the speaker has more treble and bass on the tube amp because the output is more like constant current where as the solid state amp has a constant voltage.   Basically the amp output impedance interacts with the speaker impedance,

https://patents.google.com/patent/US5467400A/en
https://www.aikenamps.com/index.php/designing-a-reactive-speaker-load-emulator
https://patentimages.storage.googleapis.com/ec/fe/f3/59fd3505ea7003/US5197102.pdf

The whole point of R26 and R28 is to make the solid-state amplifier output impedance finite so it acts like a tube amp and reproduces these effects.

As for the TDA2040, typically the gain is around 30dB but (normally) there's nothing stopping you using more gain.   If you try to set less gain you can get problems where the amplifier oscillates.

The addition of R26 and R28 affects the gain.  You can calculate the gain assuming a resistive load and you will get a gain value (amp approx 41dB, and 21dB to pre out) but a flat response from the amp.  You need to model the gain of the amp with a real speaker impedance to see the effect of what is really happening.   The response will be like the plot in the Marshall patent.

[PRR]

Mixed-mode:

We have 207 as voltage mode, 3.3/0.1r times 2730/330r or 273 as current mode. If we pretend that 207=273 then gain would be 237 and output impedance would be 3.2 Ohms. Since the current feedback dominates, a little, I'd pencil the Zout as 4 Ohms.

Note that the voltage gain will depend on the load impedance. Speaker is very variable impedance so you get a mild copy of the speaker resonance and impedance plot. (And potentially a very small reflection of speaker minor resonances and nonlinearity.)

I don't recall if I have seen a speaker-application analysis. But it is an extension of Howland Current source, compromised for non-infinite impedance.

If you have a SPICE with any opamp you can make it solve for gain and impedance. Cheating but it is too muggy today for full analysis.

And what Rob said.

These speaker chips have MUCH more gain than we typically use them at. Say 30db closed-loop and up to 60dB open-loop. They figure like op-amps, assume "infinity" (until you get silly).

[PRR]

.... seems like the gain would be significantly higher than the chip is capable of, which based on the page above is around 33dB.

Page 6, gain vs freq:


If you had to, in the guitar band, you could get 50dB-60dB of gain. OTOH, for HiFi application, you want to leave 20dB of gain in hand to reduce errors, giving 30-ish dB as you say.

BTW: you know you can't buy a 2040 today. The copy-cats may work at low voltage but typically explode long before full rating.

[Rob Strand]

FWIW, very crude simulation with and without speaker,



Doesn't add much that isn't already in the Marshall Patent.

(Something else, the nominal impedance of the speaker is probably 4 ohm not 3.2 ohm.
3.2 ohm looks like the DC resistance of the coil.  I could be wrong.   Doesn't affect the main points though.)

FWIW, here's one way to calculate the observed gain (with a 4 ohm resistive load):



The Av I have calculated is for the speaker terminals.  To get what the preamp out sees, which is ground referenced, I believe it is more correct to use 4.1 ohms not 4.0 ohm for the final output divider calculation.

[Axldeziak]

On my '92 Peavey 158, the 8" Blue Marvel speaker meters as 3.8 ohms, and is labeled 4 ohms.

[aion]

FWIW, very crude simulation with and without speaker,



Doesn't add much that isn't already in the Marshall Patent.

(Something else, the nominal impedance of the speaker is probably 4 ohm not 3.2 ohm.
3.2 ohm looks like the DC resistance of the coil.  I could be wrong.   Doesn't affect the main points though.)

FWIW, here's one way to calculate the observed gain (with a 4 ohm resistive load):



The Av I have calculated is for the speaker terminals.  To get what the preamp out sees, which is ground referenced, I believe it is more correct to use 4.1 ohms not 4.0 ohm for the final output divider calculation.

Thank you, that all makes sense! I'm unclear based on the schematic whether the speaker is disconnected when the preamp out is used - if the jack is wired exactly as shown then the speaker would still be operational, though it's fairly conventional on most amps that the preamp out would disable the speaker. But the manual does say "This patch does not affect the operation of the amplifier" - so based on that, and using it to interpret the schematic, I would lean toward the speaker still being active.

BTW: you know you can't buy a 2040 today. The copy-cats may work at low voltage but typically explode long before full rating.

Yep, although I think the TI LM1875 is a drop-in replacement for most applications and seems to still be in production.

[Rob Strand]

Thank you, that all makes sense! I'm unclear based on the schematic whether the speaker is disconnected when the preamp out is used - if the jack is wired exactly as shown then the speaker would still be operational, though it's fairly conventional on most amps that the preamp out would disable the speaker. But the manual does say "This patch does not affect the operation of the amplifier" - so based on that, and using it to interpret the schematic, I would lean toward the speaker still being active.

My experience with Peavey amps is the Pre-out doesn't cut the speaker (or cut the power amp input - in the cases where the pre-out is taken from the actual pre-out).   That way you can add slave amps.   So I expect it is as per drawn schem/manual text.

[Axldeziak]

On my 158, the pre-amp output does not cut the speaker, and daisy-chains quite nicely with other amps in line.