12AU7 Phono Stage

[piano3]

JohanB,

Yes,of course high mu valves but what I was talking about was a higher than published values of Cg-a experienced with high gm RF valves in actual use. Many valves claim 1 point something pF but when you factor this value in when designing the 75 microsecond pole and then listen to the result it is clearly wrong-and I really mean clearly! D3a pentode triode wired specs own up to 3 point something pF as far as I remember. I think MJ found a very large discrepancy with the EC8010 valve.

 

[JohanB]

JohanB,

Yes,of course high mu valves but what I was talking about was a higher than published values of Cg-a experienced with high gm RF valves in actual use. Many valves claim 1 point something pF but when you factor this value in when designing the 75 microsecond pole and then listen to the result it is clearly wrong-and I really mean clearly! D3a pentode triode wired specs own up to 3 point something pF as far as I remember. I think MJ found a very large discrepancy with the EC8010 valve.

OK, I hav'nt checked these high-gm tubes. My stock of NOS 5751's are within specs what i've found out so far. With the D3's you have to encounter the screen grid as well in triode mode. But it is only the Cg(a+g2) that multiplies. All other caps are as is.
Strange that these proffesional "telecom" tubes dont keep the datas as you say. I have to measure this. Should be easy as you just can measure them cold with a socket on your desk.

JohanB

 

[piano3]

I wish I had measuring equiptment that could measure a pF or so reliably! Are you aware of any effects that could cause the Cg-a to be frequency dependant?

 

[DualTriode]

DualTriode,

Have you looked carefully at the Valley and Wallman page you are talking about? If you have the same reprint as I do there are mistakes/printing errors. There are superscripts missing on the rp,gm and mu terms-I was very confused until I figured out what was going on.

Regards,N.A.

I try to avoid first editions.
The copy I am looking at is Boston Technical Publishers, 1964.
There may be an understanding on the way here, I hope.
I do not have a scanner at my home. On Monday I will scan and make pdf’s of the items referenced in my previous posts.
I would rather sort this out. In my mind it will add more value to the thread than letting it hang.
DT
All just for fun!

 

[DualTriode]

Hello,
If you have intrest in the gm' of the common cathode and Cascode circuits they are the same on this page. Google found it. Improving the Cascode's PSRR (page 2)

DT
All just for fun!

 

[piano3]

DT, that was interesting to read, thank you.

JohanB; if you are able to measure accurately Cg-a on triode connected E180F, E280F and E810F that would be very useful information!

 

[JohanB]

DT, that was interesting to read, thank you.

JohanB; if you are able to measure accurately Cg-a on triode connected E180F, E280F and E810F that would be very useful information!

Hi, I'll try to do E180, 280 & 282 next, but during the last hour I've been busy measuring:

Socket A = Noval PC-type brown micanol made by NSF (germany) 1958
Socket B = Noval PC-type White ceramic made by Nuval 1990 in Italy

ECC83 Telefunken NOS orig. plain anodes 60's
ECC82 Telefunken NOS orig. plain anodes 60's
ECC88 Ultron NOS (Germany) 60's
5751 Sylvania NOS 1957

Measure between pin 1 & 2 . Other pins not earted
On the sockets were soldered 2 pcs of L=25mm dia. 0,6mm solid copper wires,
to jack into the meter. This to be able to repeat the exactly position of the tube/socket.

Residual capacitance on meter Escort ELC-131D is a stable 1,1pF that is substracted from the readings below. Resolution on the meter is 0,1pF

Calibration check: a 4,7 pF 2% gives exactly 4,7 on the meter with a Dissipation Factor of 0,005

Socket A alone = 0,7pF D= 0,011
Socket B alone = 0,6pF D= 0,010
ECC83 in socket A = 2,8pF D= 0,012
ECC83 in socket B = 2,7pF D= 0,011
ECC88 in socket A = 3,0pF D= 0,011
ECC88 in socket B = 2,8pF D= 0,007
5751 in socket A = 3,0 pF D= 0,009
5751 in socket B = 2,7pF D= 0,008
ECC82 in socket A = 3,0pF D= = 0,011
ECC82 in socket B = 2,8pF D= 0,010

Calibration check the 4,7pF again........4,7pF D = 0,005

In my 1958 book "Valves for A.F. Amplifiers" by Philips Technical Library, says 1,8pF for the ECC83, but datasheet Philips shows 1,6 and for ECC82 1,5.
My Vade-mecum 1961-1963 says 1,4 for the 5751, 1,6 for ECC83 & 1,5 for ECC82.

Conclusion: You have to calculate 3pF as Cag IRL for most designs and all small trodes, both on circuit boards and hardwire. Do not let the g & a leads cross each other on opposite sides of the board.

These measurments is also a little inaccurate, as they dont include the parasitic capacitances through the other electrodes, that was left floating (not connected to ground). So readings could be 0,1-0,2pF higher than the actual capacitance.

The only way to measure exactly, is to put a 47KOhm series resistance input resistor, and find at what frequency the -3dB point is. If the normal MM load resistor is 47K, then the generator impedance is 23,5 KOhm (47/2) and you have to calculate from that.

But 2,5 -3pF seems to be a rule of thumb for almost all small Noval triodes for now.......


JohanB

 

[JohanB]

DT, that was interesting to read, thank you.

JohanB; if you are able to measure accurately Cg-a on triode connected E180F, E280F and E810F that would be very useful information!

Sorry, I don't find my E810F, so I showed the E282F instead.

Hi, these tubes as triode connected and including a ceramic tube socket. All other pins floating:

Telefunken E180F S= 16,5mA/V @ 13mA, u=50 Cg1 to (g2+a) = 4,0pF, to (g2+g3+a)= 4,4pF

Siemens E280F S= 26mA/V @ 20mA , u= 60 Cg1 to (g2+a) = 4,4pF, to (g2+g3+a) = 4,5pF

Siemens E282F S= 26mA/V @ 35mA, u= 27 Cg1 to (g2+a) = 5,3pF, to (g2+g3+a) = 5,6pF

Rp (or Ri) = u/S More data here Frank's electron Tube Data sheets

JohanB

 

[JohanB]

I was a little stupid there. It is of course only the capacitance between g1 and g2 that is of interest. The anode is shielded from g1 by the internal shielding!! New figures!!

Measured in a ceramic PCB socket with 0,3 pf capacitance without tube inserted.

E180F C from g1-g2 = 3,3pF
E280F C from g1-g2 = 3,7pF
E282F C from g1-g2 = 4,5pF

JohanB

 

[piano3]

That is tremendous work JohanB! Many thanks for that and I hope many members read it. I had estimated by ear on 75 microsecond RIAA stage that the values for these triode connected pentodes had to be in this region though I had not suspected quite as high as 5.6 pF!


your last post was while I was writing this-yes that seems more believable!

 

[JohanB]

That is tremendous work JohanB! Many thanks for that and I hope many members read it. I had estimated by ear on 75 microsecond RIAA stage that the values for these triode connected pentodes had to be in this region though I had not suspected quite as high as 5.6 pF!


your last post was while I was writing this-yes that seems more believable!

Yes, the g1 to anode capacitance is only some 30-50 micro-pico-farads when the shildings are grounded, so anode and g3 should not be measured together with g2 for capacitance. It is the stray capacitance between the shildings inside that is left floating, that makes the higher readings when the anodes (and g3) are included.

By the way, do you have problems with microphonics with these tubes?

Many people say that they are, but I suspect there in many cases are VHF self-oscillation, that's causing the exaggerated microphonic rumors......

 

[Salas]

Thanks for the work Johan. They need Rg stopper in most cases, which works against best noise in a head amp to a point. Kinda prone to oscillation sometimes, yes. When cranked for gain with much Miller at play you know how it goes when stray inductances are meeting the grid.

 

[piano3]

I used trioded E180F in a mic preamp and you would certainly get a nice silvery shimmer if you banged the case but in actual use there was never a problem. I am currently using trioded E810F(and I can drop in E180F,E280F or EF184 as I have switchable capacitance in the HF EQ stage) in the second stage of my phono amp without any problems whatsoever-you have to really hit the case near the valves to get any ringing at all. I always use grid stoppers with these kind of valves.

 

[popilin]

Hi DF96

Amazing thread by the way.

No, given the same valve and bias a cascode has exactly the same gm as common-cathode. The difference in gain comes from the different output arrangements. The cascode largely isolates the lower valve from anode feedback, so the effective mu is raised to pentode-like figures. The raw signal current and noise current generated by the lower triode are the same as for common-cathode, but there is no anode feedback to then reduce them. The signal-noise ratio stays the same, except for a little extra noise from the upper triode - usually negligible.

As usual I'm a bit confused about noise, especially that produced by the upper triode of the cascode, since its common grid connection, I'm confused about how to refered noise, triode noise usualy is refered to grid.
The gain of the upper triode is

G2=(μ+1)RL/(rp+RL)

For the bottom triode, I have an expresion that is unwritable, because of index notation (Herr Vogel), but clearly

│G2│>│G1│

So, what about upper triode's noise ?
Why is usually negligible ?

I'm struggling with a hybrid cascode, and can not find answers, neither Morgan Jones says anything about it.

Thanks in advance.

 

[Merlinb]

So, what about upper triode's noise ?
Why is usually negligible ?

I'm struggling with a hybrid cascode, and can not find answers, neither

It is easier to solve noise problems like this by considering the noise current generated in the tube, rather than trying to convert it to EIN and then working out different voltage gains and adding etc etc... Having said that, the noise formulae for a cascode are fairly involved whichever way you do it, so simplify:

Assuming the two device both identical and generate the same noise current in the anode, for the upper device to contribute negligibly to the total, the lower triode needs a voltage gain of perhaps x3 or more. Provided this is the case, the lower triode contributes three times more (or more!) noise than the upper. And we all know that when adding uncorrelated noise sources, the dominant sounce is always VERY dominant:
sqrt(3² + 1²) = 3.16 (i.e. only 0.5dB worse)

The EIN is therefore roughly equal to the EIN of the lower triode on its own. Of course, if the lower triode happens to have less than x3 gain then you can't make this simplification any more, and things get rather awkward.

The lower triode works into the internal cathode impedance of the upper triode, which is
rk = (Ra+ra) / mu
You can use this figure to work out the voltage gain of the lower triode using the usual formula for a common cathode gain stage.

With a hybrid cascode it gets even more tricky, because the two devices are unlikely to generate equal noise currents. If the lower device is a FET then it will probably generate less noise than the upper (triode), in which case the total noise may well end up worse for the cascode than if the FET were arranged as a common-source amplifier with plenty of gain. For example, if the lower device generates 4 times less noise current than the upper, then it will need to acheive a voltage gain of 4 x 3 = 12 to dominate the total noise. Possible with a BJT, but perhaps not with a JFET?

 

[Salas]

Merlinb, excellent summary.

 

[popilin]

Hi Merlinb, thanks for your reply !

For more clarity, unifying notation, lets asume that the upper triode is 2 and the lower triode is 1, asuming also both are identical, for the unbypassed version of the cascode, voltage gains are

G2=(μ+1)Ra/(ra+Ra)

G1= - μ rk/[ra+rk+(μ+1)Rk]

With

rk=(ra+Ra)/(μ+1)

Clearly

│G2│>│G1│

So I don't understand your numbers │G2│=1, │G1│=3

Please bear with me, I think slowly...
Thanks again.

 

[Merlinb]

G2=(μ+1)Ra/(ra+Ra)

G1= - μ rk/[ra+rk+(μ+1)Rk]

With

rk=(ra+Ra)/(μ+1)

Clearly

│G2│>│G1│

So I don't understand your numbers │G2│=1, │G1│=3

The numbers I wrote are not the gains of the triodes, it was just meant to illustrate that if one noise source is three times larger than another, then the total is only 0.5dB worse than with the noisiest source alone.

Another way to think of it is to imagine that the EIN the triodes is, say, 1uV each. If the lower triode has a gain of 3 then its EIN is amplified to produce 3uV of noise at its anode. The upper triode then adds its 1uV to the mix, making 3.16uV (basically the same as the lower triode alone).

 

[Salas]

Also at least 1/3 the overall gain must be coming from its head amp to dominate any full preamp's noise performance.

 

[popilin]

The numbers I wrote are not the gains of the triodes, it was just meant to illustrate that if one noise source is three times larger than another, then the total is only 0.5dB worse than with the noisiest source alone.

Another way to think of it is to imagine that the EIN the triodes is, say, 1uV each. If the lower triode has a gain of 3 then its EIN is amplified to produce 3uV of noise at its anode. The upper triode then adds its 1uV to the mix, making 3.16uV (basically the same as the lower triode alone).

Ahh, now I understand (I hope), thanks for clarify.
That means the common grid stage only contributes its own noise regardless its gain.(?)

In addition that I think slowly, I'm a dumb.

Thanks a lot