Low noise amplification with mulitple valves

[yestertech]

I'm investigating the idea of a project utilising multiple paralleled input valves for audio. I recall reading a study somewhere about the various dB's of noise improvement available with each doubling which, of course, I can't now find !!
The theory was along the lines that the for the total array of 2,4,8 valves in parallel, the amplified signal is coherent and hence is simply additive , whereas the noise is not, so this is only partly additive. result : better S/N ratio than with one valve alone.
Of course there are the overheads: More HT mA and more Heater A, but with modern PSU arrangements this is not that tricky ( especially as one can series operate the heaters on DC to keep all that noise at bay )
So my question is has anyone played with this, and/or know of anywhere which describes this in more detail, with any practical examples ??

Thanks

Andy

Given ideal circumstances you will get a 3dB reduction in noise per doubling of input devices. That's -3dB in volts i.e. 0.708.

[G8HQP Dave]

You only get the advantage if you don't have to share the signal power between the devices. Valves have such a high input impedance that you can sometimes get away with this. However, the advantage may be illusory as you can get better noise performance just by using higher gain valves, instead of paralleling loads of lower gain ones.

If you do parallel valves, don't forget that the maximum grid leak resistor has to be divided by the number of valves. This may limit what you can do.

[Lucien Nunes]

Is it not therefore true to say that where there is enough input current available to drive many valves in parallel instead of one, the parallel configuration has better SNR mainly on account of having less of an impedance mismatch into the grid circuit than the single valve configuration?

If that mismatch were not there, i.e. if the source were able to drive exactly one grid without excessive loading, then for the same input power the amplitude at the grid would be correspondingly increased and the SNR proportionally lowered for the single valve case.

Lucien

[cmjones01]

Originally Posted by Lucien Nunes

Is it not therefore true to say that where there is enough input current available to drive many valves in parallel instead of one, the parallel configuration has better SNR mainly on account of having less of an impedance mismatch into the grid circuit than the single valve configuration?

...and therefore, wouldn't a transformer do the same job? (though have other imperfections)

[G8HQP Dave]

Better input matching is a way of getting 'free' gain, by not throwing away signal. If everything is matched then there is no noise advantage in paralleling devices - you gain 3dB S/N at the output but you already lost 3dB at the input.

In some cases paralleling is better than using a transformer. It is certainly likely to be cheaper - decent transformers are expensive!

[acorn]

Dave

Interesting disscusion, what do you mean by better input matching are you refering to the input valve or the passive components or both in what way do they play a part in improving the gain, ie matching the input device like a tuner or C.D player to your input valve with the correct impedence.

Acorn

[G8HQP Dave]

Normally when connecting audio devices together there is a serious mismatch. The output impedance of a CD player might be 200ohms, but the following amplifier might have an input impedance of 100K. This is as intended, as we have plenty of signal and the aim is to drive any cable capacitance. We don't have to worry about noise as the signal levels are so high.

Impedance matching only occurs in most audio systems at low level inputs such as magnetic cartridges, microphones, tape heads etc. Even then the match would be considered quite poor by an RF engineer! If we have a moving coil cartridge which needs a few hundred ohms input then we can either use a transformer and get some voltage step-up or have parallel or very high gain valves. Most people find the transformer is best - which is what theory says should be the case. Even then most of the signal power ends up being dissipated in the grid resistor. Unfortunately in audio the requirements of low distortion and low noise often pull in opposite directions.

[Aub]

This site discusses using paralleled triodes for a headphone amp:-

http://www.glass-ware.com/tubecircui...Amplifier.html

Cheers

Aub

[G8HQP Dave]

That is for output, to get more driving ability.

[yestertech]

Thanks folks for all the interesting thoughts. I feel some "winter experiments" coming on !!

Andy

[daviddeakin]

Using parallel valves will reduce the noise introduced by the valves themselves, relative to the signal.
However, in most applications the major source of noise is the source resistance, not the valves, so paralleling devices gives little improvement in the overall SN ratio. (Unless you're dealing with a low-impedance source, like an MM cartridge?)

[yestertech]

Thanks David,

I envisage a MC cartridge here, so Yes ,V low source impedance.
In theory one could just buy a transformer, but ones with good spec. tend to be expensive.
Plus of course they need mumetal shielding.

Andy