Pentode Noise Tests

[buildafriend]

While your doing it again, I would throw the input transformer, shorted primary, and 150 ohm resistor across the primary. It would be interesting to see how much the transformer adds to the noise, shorted and with the resistor. With the mu metal case and keeping the transformer physically as close to the tube as possible, you shouldn't get too much interference. perhaps some grounded tin foil just around the secondary to the tube grid for extra shielding. Seems if you have everything set up you might as well go the extra mile just to see.

On old tube testers they would some times have two banana plugs that were meant to accept a carbon headphone. These apparently had very high impedance and could be hooked up to the plate of the tube to listen for noise.

Got a model number for one of those?  ;D

 

[bluebird]

Got a model number for one of those?  ;D

Ha! I did see a couple at the surplus store I use to work at.

These have 2K resistance and will work with crystal radios.

 

[bluebird]

Combining both would end up adding an additional uncorrelated noise source that would just make reading the measurements more difficult and teach nothing.

Dang Abby you can be a bit blunt sometimes

 

[DaveP]

OK, I have repeated the measurements as far as possible but I was unable to measure noise with the input shorted to ground because of instability.  As a second best  I soldered in a 20k grid resistor as Merlin suggested and everything calmed down. 

Because the readings were so much lower this time, I had to make up a special screened lead from the output to the DMM to measure the noise.  The noise from the old lead was considerable and I'm surprised I managed to get any results with it, even though I carefully placed it in the same position every time.

I can't post a chart because after dividing by the gain, all the results turned out to be approximately the same at 4.4uV +0.4 -0.4.  There was no trend, just random variation due to the shortcomings of my equipment I expect.

The circuit became unstable using 470k Ra so that was omitted  from the series.

I admit that these figures are quite different from my previous set using a 100k grid resistor, but  they still show that increasing Ra and reducing the gm has no effect on the noise for this EF86.  Later on I will quadratically subtract the effect of the 20k grid resistor to see what noise level it actually was.

@Buildafriend I will post the schematic  later today.

Best
DaveP

 

[DaveP]

While your doing it again, I would throw the input transformer, shorted primary, and 150 ohm resistor across the primary. It would be interesting to see how much the transformer adds to the noise, shorted and with the resistor. With the mu metal case and keeping the transformer physically as close to the tube as possible, you shouldn't get too much interference. perhaps some grounded tin foil just around the secondary to the tube grid for extra shielding. Seems if you have everything set up you might as well go the extra mile just to see.

Jensen are making this IPT to order so I can't include it.  I think they have some very old school people in charge of this company, they have not yet acknowledged my order or even the receipt of my credit card details, not used to this!

DaveP

 

[DaveP]

Buildafriend,

This is the circuit I used, the cathode cap should probably be 100uF
Ck is 25V other caps need to be 400V

You have to solder in the Ra, Rk and Rg2 each time.

Power supply needs to be adjustable to keep at 250V when changing resistors.  I used a supply of various values and added others in parallel, between three 47uF 450V caps to get the voltage right, I was working live with croc clips, I can't advise you to do the same, a resistance box is smarter.

The heater supply was a TX with 12 volt secondary with two SR1503 (solar duty) diodes and centre tap to earth. The output goes to 47,000uF resistor and 47,000uF.  resistor needs to be several watts at around 12 ohms.  this gives 6.3VDC  You may have to adjust the resistor as above to get exactly 6.3V

I put the power supply a yard away on the bench and ran the supplies and earth wire to the tube.  You have to do this at a quiet time for electricity usage to avoid neighbours glitching the mains supply.  Power transformers were both toroidal to avoid noise radiation.

I think that's about all, if you need to know anything else, just ask.

Best
DaveP

 

[DaveP]

Quadratic correction for the 20k grid resistor in the new test.

Average tube noise found 4.4uV

Noise of 20k resistor            1.789uV

4.4 squared = 19.36

1.789 squared =3.2

19.36-3.2 = 16.16

Square root of 16.16 = 4.02uV

I think the next thing to try is to keep Ra constant but vary Rg2 upwards to test the effect on gm, gain and noise.
Rk will probably have to be adjusted to keep g1 to k voltage constant.

DaveP

 

[DaveP]

Does anyone have any ideas for a filter network to reproduce these noise weighting curves?

I don't want to re-invent the wheel if someone has already done the work

This one looks like a Spice attempt

Is there any agreement as to which one is relevant today?

DaveP

 

[buildafriend]

Buildafriend,

This is the circuit I used, the cathode cap should probably be 100uF
Ck is 25V other caps need to be 400V

You have to solder in the Ra, Rk and Rg2 each time.

Power supply needs to be adjustable to keep at 250V when changing resistors.  I used a supply of various values and added others in parallel, between three 47uF 450V caps to get the voltage right, I was working live with croc clips, I can't advise you to do the same, a resistance box is smarter.

The heater supply was a TX with 12 volt secondary with two SR1503 (solar duty) diodes and centre tap to earth. The output goes to 47,000uF resistor and 47,000uF.  resistor needs to be several watts at around 12 ohms.  this gives 6.3VDC  You may have to adjust the resistor as above to get exactly 6.3V

I put the power supply a yard away on the bench and ran the supplies and earth wire to the tube.  You have to do this at a quiet time for electricity usage to avoid neighbours glitching the mains supply.  Power transformers were both toroidal to avoid noise radiation.

I think that's about all, if you need to know anything else, just ask.

Best
DaveP

Copy that. I'm book marking this thread for my future pentode experimenting. I might even go as far as making a little PCB of your schematic with a tube socket that's all labelled clearly with while using turrets for installing and removing the experimental component values or for clipping decade boxes to.

I just built a cascode mic amplifier that achieved 0.5% THD+N post it's output transformer when driving a -37dBu signal up to +4dBu using dual triodes. This is all while not even taking my wiring very seriously. My question is how much better can I do than that with other configurations, like perhaps with an EF86 on input into whatever else my research leads me to.  This thread was really right up my alley.

I think I have pretty much everything I need to smack this experiment together in terms of power supplies, I just need a few experimental R's and C's and to make a board, maybe I'll build a high wattage decade box when finances permit.  Even if a couple of thing are effecting my end results, I'm happy having a metric to go by. 

Thanks!

 

[abbey road d enfer]

Does anyone have any ideas for a filter network to reproduce these noise weighting curves?

http://www.beis.de/Elektronik/AudioMeasure/WeightingFilters.html
I think you should also have a 22Hz-22kHz bandpass, because it's probably the most quoted figure for preamps. Also called "Lin".

"A" is quite frequently used also.
For microphones, both "A" and 468 are requested.
Beware that you'll need to precede the filter with a VHZ buffer.

 

[DaveP]

Thanks for that Abbey.

Its a very good article.

The final circuit is incredibly complicated and it must produce noise of it's own, all those special resistor values will cost a bomb as they only seem to come in packets of 50.

I may try to make a simpler passive circuit in a shielded box.

DaveP

 

[PRR]

> must produce noise of it's own

Clean transistor or opamp hiss will be similar to pentode hiss.

The pentode has gain of ~~~100, right? So the filter resistors may be 100^2 or 10,000 times higher than what's on the grid of the pentode. For the special-case 100K, you can go not-quite 100MEG in the filter. For 10K grid, don't approach 10MEG in the filter.

I don't see that any of those filter values are really 1% critical. 10% might be 1dB error, and you can't hope for 1dB error between tube variability and test costs (*proper* shielding to get stray noise >10dB below circuit self-noise). You should have a box of all 5% values. IMHO that's close enough for your work; you can maybe find "more-exact" values in the 5% parts.

I would settle for low-Q 2nd-order filtering at 400Hz and 16KHz. The <400Hz range is 1/f and hum, two special problems to be treated differently from broad-band hiss. I doubt anything odd happens >20KHz (except breadboard instability which makes all readings dubious) but your RP values will give cut-off that vary from >50KHz to <10KHz, and we should at least discount the supersonic hiss.

You should be listening also. Is it broad smooth hiss? Erratic irritating hiss? Hum? Rumble? High reading and no sound?

 

[DaveP]

Thanks PRR, good advice as usual.

As you can see from the schematic I posted, the load is a 1M resistor.  I have a mind to sub that for a passive network along the lines you suggest, do you think that is the right place for it?

DaveP

 

[abbey road d enfer]

The pentode has gain of ~~~100, right? So the filter resistors may be 100^2 or 10,000 times higher than what's on the grid of the pentode. For the special-case 100K, you can go not-quite 100MEG in the filter. For 10K grid, don't approach 10MEG in the filter. 

The very high output Z of the pentode advocates the use of a VHZ gain stage before the filter. A simple gain of 100 OPA2134 would add only 1.5uV to the pentode's output noise, and be capable of driving the weighting filter whilst minimizing the influence of its noise.

 

[DaveP]

I have done some tests following the Mazda Belvu suggestion of raising the value of g2 resistance up to 10 times the value of Ra.
It may be possible to do this with Ra =100k, but with Ra=220k I could only get up to 6.8 times due to instability.

The test was with the same tube and set-up as before including the 20k grid resistor.  HT/B+ was held at exactly 250V, the cathode resistance was adjusted to keep the g1 to K voltage as near as possible to 1.12V.  The anode/plate resistor stayed the same at 220k.  The values of Rg2 used were 1M, 1.2M, 1.5M, 1.8M and 2.2M.

Here are the voltages on anode and g2 as Rg2 is increased.

Interesting how they crossover.

Here are the currents on anode and g2

This is how the gain changes

The peak gain was with 1.5 M

This is how the gm changes

It is almost a mirror image of the gain.

This is how the noise decreases

I could not include data for the 2.2M resistor because it became unstable with that value

Finally, the ratio of Ia/Ig2.

The use of a 1.5M resistor with a 220k looks like the optimum value, I wonder if this ratio is the best one for other values of Ra?
I did not expect the gm to climb up again, but that may be because the value of RL is not constant the ra may climb much higher than the value of 2.5M that I used.  Even so the 1M load ensures that it cannot change too much.

DaveP

 

[ruffrecords]

Interesting how the maximum gain is at minimum gm. Not what I would have expected.

You have several times mentioned instability problems. What kind of instability are you experiencing?

Cheers

Ian

 

[DaveP]

You have several times mentioned instability problems. What kind of instability are you experiencing?

Normally the noise measurement is reasonably stable on my DMM, it may flit about from say 1.0 to 0.8 but spend most of the time on 0.9 for example.  But when I get to the extreme values it can go from 2.4 to 0.4 and its impossible to decide what it is so I reject that reading altogether and call that unstable.

DaveP

 

[ruffrecords]

Normally the noise measurement is reasonably stable on my DMM, it may flit about from say 1.0 to 0.8 but spend most of the time on 0.9 for example.  But when I get to the extreme values it can go from 2.4 to 0.4 and its impossible to decide what it is so I reject that reading altogether and call that unstable.

DaveP

Got it. I thought at first you were talking about the circuit oscillating.

This might well be 1/f noise. Interesting you got instability when you shorted the grid. I thought that was a bit odd but you will then see only the tube noise - perhaps the 100K resistor noise was masking the 1/f noise.

Cheers

Ian

 

[PRR]

> a passive network

Do the math. At 1Meg impedance a 20KHz filter starts with 8pFd. Stray lug and wiring capacitance is often assumed 30pFd. While radio layout may get you down near 8pFd, when you turn your head it will be 7 or 9 pFd. Also 0.0001pFd of coupling to power wires will overwhelm hiss with hum.

I know you already but too much in this. But the next step really is a TL072 buffer. Or the proposed gain of 100 OPA2134. The pentode's self-hiss is like 800uV, so the ~~<2uV of the opamp is insignificant. You may as well let it bring up that sub-mV signal to maybe tenth-Volt (Gv=100 for easy math). *Now* put a filter in. And your meter can see a low-Z source; when meters have freq specs they are mostly low-Z source, not a large fraction of a Meg. (We already saw that the typical 30pFd input on lab gear will be a heavy hit on the top of the audio band with near-Meg sources.)

 

[DaveP]

Or I could follow it with a cathode follower to get the high input Z and low output Z?

DaveP