Pentode Noise Tests

[DaveP]

I was afraid you would take it badly, but you can't blame me for pointing out procedural errors.

You're right, I can't blame you for pointing out procedural errors.  However, I do blame you for spending so much time on them and so little time  over the conclusions.  In fact those acknowledged procedural errors made no difference whatsoever to the conclusions that could be drawn because the results were always relative rather than absolute.

You might like to ask yourself whether your approach is likely to encourage me (or anyone else) to share their work on this forum.
I am more surprised than anything else, but maybe you have had a bad week.

Yes I have already seen the  French Mazda EF86 data hints, thank you.

DaveP

 

[bluebird]

I will be doing some tests myself later with some mumetal foil and copper foil.

Are you talking about shielding the tube or power transformer? Sounds like your thinking Jensen input so that is already in a mu metal can.
I found copper to be pretty worthless for any real shielding. Making my own input transformers I found even nesting steel copper steel couldn't come close to the performance of one layer of .5mm thick mu metal. Problem is its expensive. $130 for a 12"X30" sheet.
I would consider shielding your power transformer or output transformer with mu metal if you wanted to do a compact cassette design like the Tele V pres. But if your going to have a nice big chassis you may not need any shielding.
If your talking about shielding the tube.  For some reason I feel like its more important to just have the chassis ground extended around the tube and it doesn't really need heavier metal than aluminum.  I've never seen a mu metal shield for a tube.

P.S. why don't you just try the tests with a 20K grid resistor for shits and giggles

 

[DaveP]

The Germans used to put their transformers and chokes in mumetal boxes, I can't do that for reasons of cost.

From experiments that I've done, I have found that magnetic flux does not like changing from one medium to another.  So I plan to line a steel box with foil layers making four in all, iron, aluminium, copper and mumetal.  My sheets were $55.

I have ordered the Jensen IPT with the double mumetal case JT 115k e60.  Their ordering process ignores the digital age, no online credit card or Paypal, you have to download a PDF form, fill it in, scan it and send it back, preferably on horseback.

My customer prefers a 2u rack mount, so as you say I have much more room to play with, nevertheless it will be a challenge to equal  the spec of a design  that's around 60 years old.

DaveP

 

[abbey road d enfer]

I do blame you for spending so much time on them and so little time  over the conclusions.  In fact those acknowledged procedural errors made no difference whatsoever to the conclusions that could be drawn because the results were always relative rather than absolute.

As I mentioned, your conclusions may be utterly wrong. Let me give you an example that is pertinent to capacitor mic design:

It is  known that noise in a parallel RC circuit obeys the formula Vn=sqrt(k.T/C). In order to achieve adequate LF response, the input impedance must be high enough vs. the capacitor's (the capsule) reactance. One could conclude that , as long as this criterium is verified, there is no further optimization, in particular regarding noise, because the RMS value of noise doesn't change vs the resistor value. However, the effect of the resistor value is to shift the noise spectrum, so a very high value will shift the noise spectrum towards infrasonics. A wideband meter will not measure any different, but one with a psophometric filter (or simply a 22-22k bandpass), and your ears will tell the difference. This is very counter-intuitive, since it is common knowledge that increasing resistors increases noise.

This pertains to your experiments, because you don't know what part of the noise spectrum you are actually measuring; are you improving flicker noise below 20Hz, or Johnson noise at 100kHz?

If you had formal training in metrology, as I had, you would know that controlling the measurement BW and the rectifier type are essential prerequisite in noise measurement, even for comparative tests.

You might like to ask yourself whether your approach is likely to encourage me (or anyone else) to share their work on this forum.

Believe me, I have asked myself, that's why I took the time to try to explain what you should change in your procedure, but I see that the constructive comments from Ian, Merlin and me have been generally met with defiance. Indeed, I have tried (and failed) to encourage you, but I don't want to encourage you to persist in error.

I am more surprised than anything else, but maybe you have had a bad week.

On the contrary, I'm in a very good mood (new convertible and two good gigs). I just think you take badly the fact that someone points out your mistakes. I know it's frustrating, working hard on a project and being met with criticism, but I think you should consider the advices given by Ian and Merlin and go back to the bench.

 

[DaveP]

This pertains to your experiments, because you don't know what part of the noise spectrum you are actually measuring; are you improving flicker noise below 20Hz, or Johnson noise at 100kHz?

At last you have a point that makes sense.  So in effect you are saying that that no-one can take noise measurements without a psophometer and the appropriate filter, because the noise may simply move to another part of the spectrum,  I give you that point.  However, You spent some time telling us that a my DMM did not have sufficient bandwidth to take proper measurements, now you are saying that the bandwidth must be restricted with a filter, I don't think you can have it both ways.

You can't include Ian and Merlin in this conversation, Ian has made a more positive response. Merlin a one liner.

I included a grid resistor in order to apply a signal  to measure the gain and subsequently calculate the gm.  This kind of work is very sensitive as to positioning of leads and components and even where I stand to take the measurements (Like Heisenberg's Uncertainty Principle).  A series of noise measurements with different components soldered in and out must not take too long or baseline drift will set in.  In these circumstances I decided it was better to leave the 100k grid resistor in place rather than keep disturbing the components with soldering.  Simply shorting out the resistor with a croc clip does not help because the loop of wire acts like an antenna.

I am not reacting to your criticism out of defiance, I have spent most of my 68 years in solid state chemistry observing similar disciplines to that which you uphold.  After so many years on the bench one develops a feel for experimentation, what corners can be cut and which can't.  I stand by my work even though there are "procedural errors", I suggest someone with better equipment than I possess repeats the experiment, this is after all how most scientific work is tested, by peer review.  I am confident they will observe similar results.  But it won't happen because the common belief is that pentodes are passe', Frank Blöhbaum’s work for Linear Audio was a rare exception, generally, only triodes are deemed worthy of investigation nowadays.

Best
DaveP

 

[abbey road d enfer]

However, You spent some time telling us that a my DMM did not have sufficient bandwidth to take proper measurements, now you are saying that the bandwidth must be restricted with a filter, I don't think you can have it both ways.

Did I? Please re-read. Particularly my answer #4 "You've just neglected one factor: noise measured by Dave is not limited to 0-20kHz. " I attracted your attention to the fact that the BW may be inappropriate, meaning it could be either too narrow or too wide.

You can't include Ian and Merlin in this conversation, Ian has made a more positive response.

Ian had the same concerns as me about the 100k grid resistor and the instrumentation.

Merlin a one liner.

Which does not prevent it to be pertinent and in accordance with my opinion.

I included a grid resistor in order to apply a signal  to measure the gain and subsequently calculate the gm.  This kind of work is very sensitive as to positioning of leads and components and even where I stand to take the measurements (Like Heisenberg's Uncertainty Principle).  A series of noise measurements with different components soldered in and out must not take too long or baseline drift will set in.  In these circumstances I decided it was better to leave the 100k grid resistor in place rather than keep disturbing the components with soldering.

You are sacrificing scientific rigour in favour of practicality.

  Simply shorting out the resistor with a croc clip does not help because the loop of wire acts like an antenna.

And you think a resistor does not act like an antenna?

I am not reacting to your criticism out of defiance, I have spent most of my 68 years in solid state chemistry observing similar disciplines to that which you uphold.  After so many years on the bench one develops a feel for experimentation, what corners can be cut and which can't.

I have spent most of my 69 years in audio design. After so many years one develops a knowledge of all the pitfalls that lay in experimentation, particularly to those of whom it is not the discipline of election. If I endeavoured experimentation in chemistry, I would very humbly listen to criticisms coming from more experienced people and put aside any pride and preconceived ideas. What would you say if I did esterification in a copper pot?

I am really disappointed you take that as a personal attack, but if you choose to ignore competent opinions, it's your right. I won't fight it, but as a moderator, it is my duty to make sure that assertions not supported by complete objective scientific data do not reach the status of indisputable truth.
I believe we have reached a point of utter non-productivity, so I will limit myself to this one and only role.

But it won't happen because the common belief is that pentodes are passe', Frank Blöhbaum’s work for Linear Audio was a rare exception, generally, only triodes are deemed worthy of investigation nowadays. 

Now it looks you're in "Dave P. against the world" mode; not the best way to approach the subject with serenity, rigor and openmindedness...

 

[DaveP]

I agree this is going nowhere.
The English are less direct than the French in general and cultural differences have a lot to do with it.

What makes you think I was ever offering my work as undisputed truth, I'm not that arrogant!  I posted it for discussion and to be helpful.

Look Abbey, I have enormous respect for you and would never fall out with you, but you are not  God, so don't expect me to defer to you on every occasion.  Part of this is me, I will admit that over the years I have come to lose respect for experts in many fields.  I have seen architects tear down communities and house them in tower blocks and wonder why social cohesion vanished.  I have seen experts advise the government to rip up the UK railway system because they thought everyone would be travelling in helicopters in the future.  I have seen doctors prescribe opoids for minor ailments and destroy people with addiction.  I have had to deal with PhD's who submitted to me electrically unbalanced chemical formulae for manufacture because they never took the time to check.  I am not saying you fall into that category, but it is the reason I don't accept everything that is tossed my way.

You are sacrificing scientific rigour in favour of practicality.

Exactly, because in this incidence it was appropriate to do so.  If I had said to myself "don't try this because you don't have  the equipment or the training", then we would be none the wiser.  If people are concerned about my methodology they can choose to repeat my experiment for themselves or ignore it, it is just out there, but not as a statement of fact.

We can agree to differ.

Best
DaveP

 

[merlin]

So in effect you are saying that that no-one can take noise measurements without a psophometer and the appropriate filter, because the noise may simply move to another part of the spectrum,  I give you that point. 

Your DMM probably has a measurement bandwidth that is well within the useful audible range (10kHz? less?), and even if it is not true RMS, I expect you will get measurements that are good enough for your own relative comparisons. Not good enough for publication, but good enough for you.

The problem is the 100k grid-ground resistor. It is simply too large.  Its noise will swamp the effect of the tube under many operating conditions, leading you to false conclusions even for your own practical purposes. If you think your real-world source impedance will be 20k then by all means use 20k (although I prefer zero ohms), but 100k is too much. It is not representative of many real-word circuits at all.

 

[moamps]

The problem is the 100k grid-ground resistor. It is simply too large.  Its noise will swamp the effect of the tube under many operating conditions, leading you to false conclusions even for your own practical purposes. If you think your real-world source impedance will be 20k then by all means use 20k (although I prefer zero ohms), but 100k is too much. It is not representative of many real-word circuits at all.

Dave started the thread with "After the thread on the Lorenz mic pre recently". 
Lorenz mic pre has 1:18 transformer, most known German preamps have  input transformers with 1:20 ratio and in some variant 1:30, so the 100k is much closer than 20k.
All of them use pentode with Ia smaller than 1mA at the input, IIRC.

 

[merlin]

most known German preamps have  input transformers with 1:20 ratio and in some variant 1:30, so the 100k is much closer than 20k.

OK, good example!

 

[abbey road d enfer]

Your DMM probably has a measurement bandwidth that is well within the useful audible range (10kHz? less?),and even if it is not true RMS, I expect you will get measurements that are good enough for your own relative comparisons. 

Estimating the HF response of  most DMM's is kinda hieroglyphic, but OK, assuming pinkish noise spectrum, comparisons can be made. However, the LF response is an issue; including subsonics in the measurements can be a major error factor, since it is proven (by you in particular)  that shifting the bias point concurrently shifts the point where flicker noise starts to dominate shot noise.

 

[DaveP]

Merlin,

Well, I don't mind repeating it with say a 20k grid resistor if you think that will be more convincing?

Where I'm coming from, the V72  mic pre had grid resistors of 250k and 400k depending on late or early model, so 100k seemed somewhat modest.  The V-241 which started all this has a 2M grid resistor.

Nevertheless, out of respect of all the help your articles have given me down the years, I'll redo the noise section.

DaveP

 

[merlin]

the V72  mic pre had grid resistors of 250k and 400k depending on late or early model

But that resistance was shunted by something else, no? An input transformer or other signal source?

However, the LF response is an issue; including subsonics in the measurements can be a major error factor,

I think we can assume the DMM bandwidth goes down at least to 50Hz mains frequency, and presumably lower in practice. OK maybe not 20Hz, but probably low enough for Dave's needs.

 

[abbey road d enfer]

Dave started the thread with "After the thread on the Lorenz mic pre recently". 
Lorenz mic pre has 1:18 transformer, most known German preamps have  input transformers with 1:20 ratio and in some variant 1:30, so the 100k is much closer than 20k.

That is correct, of course, but the practical probability of using a transformer greater than 1:10 is so small that it has been instinctively eliminated. Which is confirmed by the fact that Dave P. mentioned the use of a 1:10 Jensen xfmr.

 

[DaveP]

Yes, that's true, but I can't attach the transformer because without a custom enclosure and serious shielding the noise picked up swamped my results,  With the 100k it at least was stable enough for me to decide if the DMM stayed longest on 1.9mV or 1.8mV.

Thinking about it, because I have already done the gain and gm measurements, I could just solder the grid to earth now and repeat the test.  But it still will only be relative measurements.

DaveP

 

[DaveP]

I think we can assume the DMM bandwidth goes down at least to 50Hz mains frequency, and presumably lower in practice. OK maybe not 20Hz, but probably low enough for Dave's needs.

Ok, I can settle the DMM issue, I've just tested it with my sig gen.

10Hz 500mV, 100Hz 500mV 1kHz 500mV 10kHz 300mV, 20kHz 100mV 30k 6mV

Does this constitute a useable meter in your opinion?

DaveP

 

[abbey road d enfer]

I think we can assume the DMM bandwidth goes down at least to 50Hz mains frequency, and presumably lower in practice. OK maybe not 20Hz, but probably low enough for Dave's needs.

I have made tests on my $50 DMM. HF response is +6(!)dB at 200 kHz, LF is -1dB @20Hz, -6dB@5Hz and -10dB@2Hz. Otherwise relatively flat from 20 to 20k. Indeed, the double-ramp conversion process interferes with LF measurements, so these figures are very approximate. It is definitely unsuitable for noise tests, even comparative, because it takes subsonics and ultrasonics into account.

 

[bluebird]

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.

 

[abbey road d enfer]

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.

That would be having the finger in more than one pie. Noise mechanisms in xfmrs are well known. Their Johnson contribution is perfectly modelled by the reflected resistance at the secondary. Their other contribution to noise is magnetic susceptibility to external fields.
Combining both would end up adding an additional uncorrelated noise source that would just make reading the measurements more difficult and teach nothing.

 

[buildafriend]

Dave,

Can you post a schematic of your test setup?

I have experimented much more with triodes than pentodes. Being that the EF86 is considered such a great tube for low noise characteristics I have been thinking of experimenting with one.

And thank you for your contributions!

All the best,