Pentode Noise Tests

[DaveP]

This is the spec:-
Ideal for FET or vacuum tube input amplifiers
• Wide bandwidth: -3 dB at 2.5 Hz and 90 kHz
• 20 dB of voltage gain with Noise Figure of only 1.5 dB
• Input impedance of 1.4 kΩ for loading loss of 0.9 dB
• High common-mode rejection: 110 dB at 60 Hz

So you were spot on with the noise figure!

Best
DaveP

 

[DaveP]

I have looked at the pentode tests that I've done and the only thing that really stands out is this:-

When as many constants are made as possible, like B+ voltage and the ratio between the plate and grid resistors, a straight line is made passing through zero as this graph shows.  This type of graph is called y=mx where m is a constant.  The constant changes with the ratio of Rp to Rg2.  Because the constant changes, you can't plot the values in a meaningful way if the ratios between Rp and Rg2 are not held constant, they will all be at different angles so will not make a straight line.

This covers the range of plate resistance from 100k to 470K with the Rg2 exactly 4.54 times each value.

This graph uses a different multiplier for Rg2 and the range is only from 220k to 390k but the line is similar.

The final point I can make is that the ratio between the plate current and the g2 current is decided by the spacing of the grid 2 wires (Morgan Jones et al).  This means that your options are limited in terms of reducing noise, because the partition noise is directly affected by this physical constraint.  A pre-war EF12 for example has only a  3:1 ratio, whereas an EF86 has a 5:1 ratio.  Designers learned that they could minimize noise in audio pentodes by increasing the ratio.  Their designs also minimized the current to reduce noise, which is what I have found myself in these tests.

I won't be adding anymore to this thread as I'm working on the V-241 now.  I hope this has helped others on this forum who are still interested in Pentodes.

Best
DaveP

 

[merlin]

For your reference I have attached an article I wrote for Linear Audio, vol 10, about pentode noise.

By the way, if you want to make A-weighted noise measurements and you're not worried about popcorn noise, then you don't need to make a fancy noise shaping filter. You can use a simple, first-order 291Hz to 18172Hz* bandpass filter, with a passband gain of 1.3dB. This has exactly the same noise bandwidth as a 'proper' A-weighting filter and thus gives the same results.
You can do the same for ITU-R468 weighting by using a first-order 2278Hz to 15800Hz* bandpass filter, with a passband gain of 15dB.
*These are the -3dB corner frequencies.

 

[DaveP]

Thanks Merlin,

Very clear easily understandable article which is a great contribution to the subject.

Interesting to note that Fig. X shows lower noise at 1mA than at 2mA which is what I have found too.  There is lower noise despite the gm being lower at 1mA than at 2mA.  This means that the noise cannot be predicted from a simple formula as there are many formulae in action as you have pointed out.  The typical manufacturers set-up of Ia =0.6mA and Ig2 = 0.12 looks to be about optimal.  While a triode obviously has less intrinsic noise than a pentode, the much greater gain of a pentode can be used, via feedback, to lower output resistance and distortion and of course noise.

DaveP

 

[merlin]

Interesting to note that Fig. X shows lower noise at 1mA than at 2mA which is what I have found too.

That PDF was a draft copy so it has no figure numbers. They're all fig. X! If you mean the one on page 4 then the pink noise is lower at 1mA than 2mA, but the white noise is higher. It's a see-saw effect. The formulae do predict this behaviour of course, but the pink noise formula contains an empiracle constant (K) which is annoying.

via feedback, to lower output resistance and distortion and of course noise.

Not forgetting that feedback does not improve the signal to noise ratio.

 

[mjrippe]

The final point I can make is that the ratio between the plate current and the g2 current is decided by the spacing of the grid 2 wires (Morgan Jones et al).  This means that your options are limited in terms of reducing noise, because the partition noise is directly affected by this physical constraint.  A pre-war EF12 for example has only a  3:1 ratio, whereas an EF86 has a 5:1 ratio.  Designers learned that they could minimize noise in audio pentodes by increasing the ratio.  Their designs also minimized the current to reduce noise, which is what I have found myself in these tests.

Hi Dave, I have enjoyed reading this thread but there is one thing that I have not seen you mention.  You talked about the Telefunken V76/V72 being a reference for low noise circuits, then base all of your tests on EF86 tubes.  As I'm sure you know, the V7x used EF804S tubes which are basically EF86 but with different pinout and selected for low noise.  I know the different pinout is supposed to help too, perhaps by reducing partition noise?  Any thoughts on that?

 

[DaveP]

You talked about the Telefunken V76/V72 being a reference for low noise circuits, then base all of your tests on EF86 tubes.  As I'm sure you know, the V7x used EF804S tubes which are basically EF86 but with different pinout and selected for low noise.  I know the different pinout is supposed to help too, perhaps by reducing partition noise?  Any thoughts on that?

Glad you enjoyed the thread.
Well, it's simple economics, the EF804S is getting silly money and I don't think it's worth it.  I think it has improved cathode material to give more consistent low noise results.  I think the pin-out change would have been more important for ac heaters but I have always used DC heating.  As I said earlier, the partition noise is related to the amount of g2 current, so the lower the  g2 current the better, this is decided by the physical grid wire spacing,  the EF86 is pretty good in this respect.

When I said the V76 / V7x series were a benchmark for low noise, I actually meant the circuits themselves, although good tubes contribute too.  You might like to follow the V-241 from scratch thread in Pre-amps as I will be combining some V76 circuitry in this project.

DaveP

 

[PRR]

> All valves show this basic 'seesaw' behavior between flicker and shot noise.

"Seesaw"... love it.

The Tharma curves show that the 1/gm rule given in such as Terman may work above 10KHz, but 1/f-type hiss rise makes audio-range different.

Semiconductors also have a noise corner but recent decades have generally pushed this well below 1Khz, sometimes far below 100Hz.

> To make a fair comparison between triode mode, pentode mode, and 'BestPentode, it is essential that the valve always operates under the same conditions of cathode current and bias voltage (and therefore the same total gm).

As an abstract study of tubes, same V and I reduces variables.

Designing a preamp, the Designer would use whatEVER current and voltage work best for the tube and topology. Of course "seesaw" hints that we can't stray far from what the tube (and application) want.

An obvious comment on the "BestPentode": if you can get a BJT, you can put it under the tube cathode, wind a lower-Z transformer, get far lower hiss and probably wider bandwidth.

Tharma's article, Wireless World, Sept 1963, 7MB PDF file.

 

[Tubetec]

I have almost all of the wireless world magazines from 1947-1970  ,If anyone wants a copy of specific articles they have found in references that arent available online I'm happy to make scans and post them up .I was given a copy of the radio designers handbook ,WW magazines, tube reference manuals ,books on building valve amps from all the great Uk valve companies as well as loads of components and transformers by a kind old gentleman by the name of Rob Reidy who was in the RAF.While leafing through the magazines I even found a slip asking him to explain why he once went awol from one of the bases. Im sure Rob has long since passed at this stage as it was almost 25 years ago now he replied to my add looking for reference material about valve amps and he was already in his 80's at that time. 

 

[PRR]

Most issues of Wireless World 1911-1987 are downloadable at:
http://www.americanradiohistory.com/Wireless_World_Magazine.htm

 

[Tubetec]

Im just looking at a C3G pentode  I made as baxandall tone stage ,works very well but Im struggling to get the last bit of mains induced hum out of it
I have a heavily choke smoothed ht(five stages RCRCLCRCLC and a regulated heater supply in a seperate enclosure , Ive also run the heaters off battery ,
Even with the ht switched to standby and the heaters running off batteries I still get a small amount of residual hum as soon as the power supply is energised ,Ive tried different grounding arrangements such as seperate signal earth back to power supply ,seperate ht- ,and seperate chassis . I was thinking of bringing the centre tapped resistor on the heater supply on its own wire back and grounding it close to signal ground in the preamp instead of in the power supply , also maybe considering a humdinger pot. Maybe I need to reorientate the chokes for minimum induced hum . The hum is more or less completely masked in normal use by resistor noise from the input , its certainly not visible on a scope ,but I can hear  it when the gain is cranked right up and I hold the hold the headphones pressed into the side of my head ,which probably also boosts the low end due to coupling .
Worth mentioning here as well the noise I'm looking at is below 1mV so my multimeter just registers 0 Ac between HT, LT, signal output and ground. Next I might just disconnect the LT bridge from the transformer and see if that makes any difference while running on batteries.I might change the reg from 7805 with a couple of diodes under the ground pin to a 317 which is a bit quieter.
.Maybe I'm just knitpicking here trying to bannish any residual trace of hum from the noise floor ,any ideas welcome.

Thanks for the excellent link to WW also PRR,

 

[DaveP]

Tubetec.
I'm not sure if your circuit is within a chassis yet, but I can recommend Merlin's layout to reduce hum.
Available here; http://www.valvewizard.co.uk/Grounding.html

The chassis is grounded to the earth wire at the IEC socket and the power supply and amp grounded at the input socket.  This is counter intuitive to me but it works.  I ground the screen wires in screened cable to the chassis at one end, not the amp ground.

On 50Hz mains, it will be 50Hz hum from ground loops and 150Hz from magnetic radiation from transformers.  I only use toroids now in my builds.

I suspect there is a small ground loop in your circuit somewhere.

Good luck!

DaveP

 

[Tubetec]

I found on battery heater power with either terminal grounded to chassis I got a good reduction in 50hz hum.I'll try a humdinger with the regulated supply next to see if varying the ground potential of the heater helps ,
Now the 50hz is out of the way its revealed other harmonics on the HT , maybe I best increase the size of the bypass caps from .01 uf to 1-2uf along the LCRC network to more effectively shunt midrange noise to ground.The time constants  of the filter sections are subsonic so maybe arent as effective at supressing higher harmonics from the line as they could be.

In the end I rotated a magnetically screened choke 90 degrees and and all the noise vanished , there was coupling between the mains transformer and the choke . ,which explains why even with the HT bridge  disconected the noise remained .

I might investigate varying the screen voltage now to see how it effects noise levels within the valve itself ,there might be a sweet spot .

 

[bluebird]

In the end I rotated a magnetically screened choke 90 degrees and and all the noise vanished , there was coupling between the mains transformer and the choke .

Ha! how good does it feel when hum finally vanishes...awesome.

While reading the initial problem, it sounded like coupling hum usually common when an unshielded inductor in an eq is close to a power transformer. Many people have had that problem with racked Neve channels in close proximity to other gear. But it was confusing when you said you had the power supply in a different chassis.

You could probably just do away with the choke. Sounds like you have enough RC stages that you don't need it.

 

[Tubetec]

Hmmm hum is back as soon as I bolted the inductor to the chassis ,I might have to find a way to isolate it ,shouldnt be too difficult.

 

[Tubetec]

I like having the choke in the final decoupling stage just before the anode load resistor  ,I just think the extra energy stored in the core is instantly available at moments of peak signal .In this case its a 1H 70 ohm mumetal screened ,and anode and screen current of the C3g is no more than 15 mA .

 

[buildafriend]

Glad you figured it out tubetec! Maybe you should make a new thread in the lab about your troubleshooting

 

[DaveP]

I abandoned the use of chokes in the power supply for the reasons you have just found.

It is now possible to get much larger compact capacitors with 12,000 hour life from Rubycon, for example.
I used 5 in series for my recent V241 mic pre project.

A choke is ideal for rejecting ac for dc but then it gives you the radiation problem instead
A 180uF cap will have all the reserve voltage you need.

The only problem with all that capacitance is that it takes longer for your bleed resistance to drain them after switch off

DaveP

 

[Tubetec]

Yeah Im a big fan of large electrolytics on stuff like this , it ensures a nice steady power up and down .
RC smoothing is fine for low current , but the choke despite the extra cost and complications keeps the Ht requirements down to reasonable numbers when higher current is needed.Of course it also gives very low frequency time constants when used with large lytics so any low audio frequencies are better handled with less sag. 2,4,6,8uf were common back in the old days with choke supplies now 220 or even 330 uf is readily available .In this case the screen grid is fed via a 75kohm resistor followed by a 100uf to ground and also bypassed to the cathode with foils ,it takes about two minutes after power up to reach a stable voltage,I can hear how the noise(hiss) in the pentode varies as the screen voltage builds up. One suggested opperating point is 220v anode 150 screen ,but I found a lower screen voltage definately decreases the noise

 

[scott2000]

the power supply and amp grounded at the input socket.. 

I'm having a bit of difficulty understanding this. So, basically all the power supply caps  should be floating and their negative leads tied to chassis at the input jack?? I have an older power supply that's just grounded through the can capacitor's shield right to the chassis so, after removing this can and installing regular electrolytics individually, I'd just tie the negative leads together then off to the input jack shield to chassis???? Non isolated jacks.....

What about an AC heater center tap hum pot.....This ground to them as well?????

I ground the screen wires in screened cable to the chassis at one end, not the amp ground.

So the power amp is grounded at the input jack , when using screened wires, where is this location of grounding? Just somewhere else on the chassis aside from the ground at the input jack?????????


Thanks!! I've read the Aiken amps article on grounding as well as the valve wizard and I'm not convinced I understand well enough.....

Sorry to derail.......