Pentode Noise Tests

[DaveP]

has link to DIY filter and amp

https://tavishdesign.com/pages/audio-tube-noise-measurments

Well found CJ.
That is an amazing body of work, a great deal of time has been spent on that survey.

Couldn't you be a bit more generous Abbey?

DaveP

 

[moamps]

No schemo for the 100x amp.

It's published in the downloads section.
https://tavishdesign.com/pages/downloads

 

[ruffrecords]

Well found CJ.
That is an amazing body of work, a great deal of time has been spent on that survey.

DaveP

I am pretty sure there was a long discussion about this at the time he did the work on Gearslutz IIRC. I was never quite sure why he use a C weighting network. I think it was also suggested he extend the work to pentodes but I don't think this ever happened.

For me, the interesting result is that most of the triode  NF values are clustered around the 8 to 10dB region. This implies that to reduce the overall NF to a small number of dB you need to precede the triode with a low NF gain stage of about 20dB in order for the triode noise not to be significant. Hence the common 1:10 step up transformer in mic pres.

Cheers

Ian

 

[CJ]

what about an instrument amp into a power amp which feeds a speaker system? that way you could listen to the pentode noise which occurs in the hearing range?

 

[ruffrecords]

what about an instrument amp into a power amp which feeds a speaker system? that way you could listen to the pentode noise which occurs in the hearing range?

Listening to tube noise is very instructive. My Lindos test set has a little built in speaker so you can do just this. When you turn a tube on, after the heaters start to warm up you get this sudden rush of hiss and then, as the space charge develops, it fades to a lower level. When you turn off, as the space charge disappears the hiss initially rises then fades away as the heaters cool.

But, the best bit is listening to the tinkling of the internal tube metalwork as it heats up and cools down. Some older tubes have 'saggy' grids. You can hear them boing when you tap the tube.

Cheers

Ian

 

[DaveP]

So we need a good space charge then for low noise.

DaveP

 

[ruffrecords]

So we need a good space charge then for low noise.

DaveP

Very definitely yes. The reduction in noise is significant. I have been meaning for some time to stick a mic in front of the little speaker and record it - if I get a chance I will.

Cheers

Ian

 

[abbey road d enfer]

Couldn't you be a bit more generous Abbey?

What do you mean? You seem to be constantly irritated at my posts. However, you finally act like you admit that my comments were justified (using 22k grid resistor, considering a weighting filter,...). Was it ungenerous of me to warn you of the many issues in measuring microvolt noise levels at high impedance, and suggesting solutions?

I've found the 100x amp details. Note its impedance is 220k as is, which is not adequate for measuring the output of a pentode with an anode resistor of 100k+.  Since the input Z is defined by a single resistor, one just needs to replace it with a higher value (4.7Meg would yield only 1dB error with a 470k anode resistor), as long as the opamp is a FET type (bipolar would not work correctly with 1Meg+ in this postion, due to input bias current). OPA2134 is a good substitute to LF353, for its input noise voltage is about 10dB less.
This is not a criticism, just a warning.

 

[DaveP]

What do you mean? You seem to be constantly irritated.

Never mind, it just seemed a bit dismissive that's all.  It was a lot of work done, freely given.

No point in going backwards, we have to move on.

DaveP

 

[DaveP]

From Wiki
Space charge tends to reduce shot noise.[6] Shot noise results from the random arrivals of discrete charge; the statistical variation in the arrivals produces shot noise.[7] A space charge develops a potential that slows the carriers down. For example, an electron approaching a cloud of other electrons will slow down due to the repulsive force. The slowing carriers also increases the space charge density and resulting potential. In addition, the potential developed by the space charge can reduce the number of carriers emitted.[8] When the space charge limits the current, the random arrivals of the carriers are smoothed out; the reduced variation results in less shot noise.[7]

6. Terman, Frederick Emmons (1943), Radio Engineers' Handbook (first ed.), New York: McGraw-Hill, pp. 286–294
7.  Terman 1943, pp. 292–293
8. Terman 1943, pp. 286–287

So I wonder if  very low currents in a pentode allow the space charge to strengthen and so reduce the noise.

DaveP

 

[PRR]

The no-space-charge current of most small tubes is 10X-100X more than the currents you are working at.

 

[DaveP]

I have a copy of Terman, I shall study it tonight, I may need some help!

He has a table of tubes with calculated noise against noise found, but none of them are operating at the currents that commercial designers used.  This is the difference between academia and what has to perform and to sell.  Commercial secrets are not always published.  I also think that while commercial amps with pentodes were being used at non published currents, electronic research had switched to transistors so no-one was interested in pentodes any more and why we are still puzzling over them.

DaveP

 

[ruffrecords]

I have a copy of the general noise formula for triodes on my desktop (attached). The two major components are shot noise and 1/f noise. Shot noise is pretty much totally characterised because all the factors are known or measurable. The second term is 1/f noise which depends critically on the constant K which is empirically derived but IIRC is or the order of 10 to the minus 10 or 12.

Most recent publications concentrate solely on shot noise which explains the audiofile fraternity's penchant for high gm tube run at high currents but 1/f noise is a very important factor in the audio band. The causes are not well understood which is why the ancients resorted to a careful selection process for tubes used in the first stage.

Cheers

Ian

 

[DaveP]

I have attached a pdf which Google translated very well from the original German.

This guy has tested many tubes for audio, but what interested me, was his test of an EF86 at various currents, at low current the noise was similar to that of triodes.

His test circuit is also included

DaveP

 

[DaveP]

I have been looking at Terman in the light of what PRR said about voltages.  Understanding something often depends on how you look at something.  I have attached part of my table of results including voltages.

In particular I have been trying to imagine what is happening to the space charge on either side of g2.

As g2 has a gm of its own, I think that at very low currents it starts to become ineffective.  After the point of maximum gain, comes the point of lowest noise and then the collapse of both.

I'm guessing that low voltages and low currents reduce the effect of faster electrons missing their targets and causing noise.

DaveP

 

[CJ]

what about partition noise? 

pentode noise talked about here>

http://www.diyaudio.com/forums/tubes-valves/160773-lowering-noise-signal-pentodes.html

need a graph for pentode noise vs heater voltage vs trans-con,  ;D

 

[DaveP]

Would someone please check my calculation for the noise level of a potential mic-pre on the basis of my measurements.

Noise found at the grid of my EF86 = ~4uV.

0.775/.000,004 = 193,750

20xLOG 193,750=105.75

1:10 Jensen = 20dB

Therefore potential minimum noise is -125.75dBu

I realise that the transformer will produce noise of its own and the layout and stray fields will take their toll.

I have not allowed for NFB on the EF86, so maybe this will help a little.

Thanks
DaveP

 

[ruffrecords]

That calculation looks about right except the EIN is in dBu not dB but I assume that is just a typo. You should also state the bandwidth of the measurement and the type of detector used e.g true rms, average calibrated for rms sine, peak or quasi peak) but you already know that.

Cheers

Ian

 

[DaveP]

Thanks Ian 

DaveP

 

[ruffrecords]

Thanks Ian 

DaveP

My pleasure. Bear in mind that -125dBu unweighted in a 20KHz bandwidth is a a very respectable EIN figure for a tube mic pre. In practice anything better than -120dBu would be considered good. The transformer itself will have a noise figure of 1 or 2dB mainly due to the dc resistance of the windings so with a 150 ohm source, a  noisless tube pre will achieve at best around -129dBu.

Cheers

Ian