Valve preamp.

[Mark Hennessy]

Thanks Mark!

Going through your views:

Bearing in mind that this amplifier design was a very big learning curve for me as the last, and only, push pull amplifier I have ever made was a more simple example from a kit quite a few years ago.

Being honest, it's a pretty accomplished design for a first effort. Definitely something to be proud of. My feedback was intended to help guide you in your learning "journey", so I hope you found it to be constructive. The joy of audio is that there's always more to learn, but even quite basic circuits can give excellent results

I would agree that the input impedance of the amp is low. Probably 5 or 10k would be better. When I determine what pre-amp I will make that decision. My original thoughs were to use the valve with emitter follower design that I tend to favour. This has a fairly low output impedance (if I remember right between 600 and 800 Ohms but is below 1k) so would work reasonably but there would be some loss. Your idea of an op-amp circuit is good as my main issue would be with tone control which may be easier with op-amp networks. A valve amp is another option. I would likely instigate the option for a plug in pre-amp. Here the low impedance may benefit if longish lines (being aware of possible issues driving by an emitter or cathode follower). I do have the odd small transistor matching transformer but I suspect their characteristics are not ideal.

Emitter followers work well. In simple terms, the output impedance from an emitter follower depends on what is feeding it. If you take it from a valve circuit - say, a triode with a 100k anode load - then the output impedance will be around 1k. More (but not completely) precisely*, it will be 100k divided by the current gain (hfe) of the transistor. And that's in parallel with the resistor you put on the emitter.

*: As you probably know, hfe varies wildly from sample to sample, and with operating conditions, so "rules of thumb" are more than adequate for estimating this sort of stuff

Of course, there's nothing stopping you using a Darlington pair or a CPF to further reduce the output impedance.

You're right to consider the capacitance of long cables, but in practice it's usually not a big problem. Just to put some numbers on it, a good rule of thumb might be 100pF per metre. From a moderate source impedance like 1k, that gives a -3dB point of 160kHz for a 10 metre cable. Ideally, it's best to aim for something in the 50 to 100 ohm region, but that again is just a rule of thumb. Clearly, a 10k output impedance would be bad, as that's more like 16kHz (not that there's much up there in practice).

What might be more significant is loading effect.

With an output impedance of 1k feeding into an input impedance of 1k, you have a potential divider that is halving the voltage. Whether this matters depends on the situation, but obviously it's worth keeping in mind.

With regard to the Vbe multiplier, I am glad you have mentioned this. I got very confused when I was designing things as one of the sites I spent some time on was Elliot Sound Products, which has helpful information. https://sound-au.com/amp_design.htm With regard to the circuit though their information, although now I have investigated and tested and know what they are getting at, is incorrect with a slider going OC. It originally threw me as I knew they were wrong. Your suggestion of the resistor going OC is much better and is probably what they really meant. Last night I put together an LTSpice simulation with the constant current drive and determined that, if the slider went OC, the voltage across the transistor would be just below 4V. Today I did some tests on one of the amps (carefully) and could momentarily turn the pot full up. This gave 4.4V across the transistor with about 3.3A supply current at 22.5V battery voltage so compares well with the simulation (the output transistors are rated at 10A but the heatsinks cannot handle the power for long and heated rapidly). Obviously an issue. For another option which I did notice on the website, I am thinking of protection diodes across the transistor, which would also help if the transistor failed too. I did vary the pot and take a few readings for reference. I will also think about the change of transistor as I do have some BC556's.

If you've ever had a transistor amp blow up, or go into thermal runaway, then you will have been forced to think a lot about these innocent-looking Vbe multiplier

Putting the pre-set between base and emitter is done because we know that pre-sets can get noisy with age, and the worst-case scenario is crossover distortion (as opposed to fireworks when it's between the collector and base). Plenty of designers make that mistake though. If using a good quality pre-set, you'll probably be fine in practice. Rod's suggestion of putting a diode string across the Vbe multiplier is not something you see all that often - it could be considered cheap insurance, but get the Vbe multiplier right in the first place.

I will also think more about the transistor on the heatsink as it is not at present, but the Elliot site also suggested that it should be if a Darlington output but not with a compound pair. All very confusing but doing some testing for limits is really helping. There is nothing like 'hands on'.

Ah yes, as your output stage is CPF, the Vbe multiplier transistor needs to be in thermal contact with the drivers rather than the outputs.