As far as I can see Langford-Smith is very careful to distinguish between mutual characteristics and dynamic characteristics (e.g. see p25). By definition, gm is the transconductance - as I said, the ratio between output current into a short circuit and input voltage. In the case of the cascode his calculated gm' for the circuit is quite correct - this gives the short circuit output current. For a common-cathode he would say gm'=gm, but as this is a trivial statement he probably has no need to spell it out.
Where does he say gm=mu/(rp+R) (or gm'=mu/(rp+R) )? If he does say it, then he is wrong! Truth is not established by quoting a guru - even Langford-Smith (although I accept that he is generally reliable).
It does make a difference when the needle touches the vinyl, if the amplifier was designed by someone who thought that a cascode would be quieter than a common-cathode input because he was confused about the definition of gm and the effect this has on shot noise.
Where does he say gm=mu/(rp+R) (or gm'=mu/(rp+R) )? If he does say it, then he is wrong! Truth is not established by quoting a guru - even Langford-Smith (although I accept that he is generally reliable).
It does make a difference when the needle touches the vinyl, if the amplifier was designed by someone who thought that a cascode would be quieter than a common-cathode input because he was confused about the definition of gm and the effect this has on shot noise.
The Radiotron Designer’s Handbook does not contain that formula. So I suppose that Langford Smith is not the person that is wrong. That formula can be found in the MIT Radiation Laboratory Series, Vacuum Tube Amplifiers 18 section 11.7 Equation (4).
Confused? No. Transconductance is well defined in any of the texts.
DT
All just for fun!
A better SN ratio, because in a preamp (low current) it is very easy to make a power supply that is essentially noise free.
You'll have exactly the same input noise in a grounded cathode and a cascode input stage in a low frequency amp. Gm is also the same. Only voltage amplification can be higher, if you put a higher value resistor, due to the high anode impedance of the upper tube.
To build a power supply that is "essientally noise free" is not enough, and the ground reference of this supply must return to the physical point (cathode or cathode capacitor) of the input tube. Not to an other stabilizing board in an other place in the amp. Things get complicated and out of hands.
You don'tn need to, when there are so many other excellent circuit designs, to use in a RIAA amp input stage!!
The art of designing sweet souning low noisetube amps, is to make elegant & simple solutions were you still can control and evaluate the few components. Not to build "SS amp-clones" with tubes and complicated transistor powersupplies.
I of course I know of, and have used the Wallman cascade in Radio Frequency amplifiers for TV & radios, where they got a lower noise figures, due to the elemination of the miller cap in the lower tube. The neutralization is then easier to make, and gives less losses. This is for HF
The secondary tuned circuit in the upper Hi-Z anode can be of high Q type to acheive much gain and selectivity, and will be shielded from feedback from the input circuitry by the grounded grid stage.
But now it is LF amps we're talking about, and you can achieve the same noise performance at lower THD at much better PSRR with:
a) An ordinary high Gm input triode with a high value resistor & high voltage B+
b) Use a CC (constant currens source) in the anode.
c) Use the higher triode as a CF (cathode follower) and use a bootstrap capacitor from the output to between the split anode resistors.
d) Connect the 2 tubes as a in SRPP (but give only 6dB PSRR)
e) Connect the tubes in a my-follower configuration.
All these configuratins referes the output signal back to the signal ground, where it shoul be, and you will not listening to your transistor powersupply added noise/hum in your speakers.
You must understand that with cascode (and pentode) stages, the powersupply comes in series with the signal.
JohanB
Last edited:
Jeez, why is everyone on this forum so relentlessly eager to suck the fun out of literally everything?
If someone is lucky enough to have a noiseless power supply then a cascode certainly IS a valid option (and makes a whole lot more sense than the SRPP, which is nothing but trendy). The cascode as a low-noise input stage is an ancient idea, not an "SS amp clone". It has also been pointed out by others that you can make the cascode have excellent PSRR by feeding PSU noise to the upper grid. Ultralinear operation is also a fun option (oops, fun = bad).
Remember, not everyone is fortunate enough to have access to arbitrarily high, gm special quality valves. Some of us have to work with what we have got. You should not disguise your opinion as gospel.
Last edited:
Yes, we oldtimers are really boring and like to suck the fun out just everything (we don't like or agree).......
.......
Never fix a thing that's alrady working
But why not make a new idea of how to use the LF-cascode better. I have an idea you still can use it and get rid of the PS noise here.
Just make a good high Z (As many M Ohms as possible!!), low noise/low capacitance CCS and put in the upper cascode tube anode to B+. Then a load reistor from the anode to ground, to determine the stage gain.
The CCS must of course feed current to both the cascode and the resistor, but no problems there, only a few mA's if you want high gain with the resistor at 47-100KOhm.
Just use a low noise resistor like MF or WW (not carbon composit os CF)
That will give a perfekt isolation to B+ noise, and with a PSRR of many many 10th of dB's.
JohanB
.......Never fix a thing that's alrady working
But why not make a new idea of how to use the LF-cascode better. I have an idea you still can use it and get rid of the PS noise here.
Just make a good high Z (As many M Ohms as possible!!), low noise/low capacitance CCS and put in the upper cascode tube anode to B+. Then a load reistor from the anode to ground, to determine the stage gain.
The CCS must of course feed current to both the cascode and the resistor, but no problems there, only a few mA's if you want high gain with the resistor at 47-100KOhm.
Just use a low noise resistor like MF or WW (not carbon composit os CF)
That will give a perfekt isolation to B+ noise, and with a PSRR of many many 10th of dB's.
JohanB
I too,believed for a long time that a cascode was the ultimate low noise input stage(because of reading Crowhurst etc.) and planned to incorporate one into my first RIAA preamp. However when I heard a few using 12AY7's in commercial preamps I was not too impressed and indeed they sounded quite hissy. Nevertheless, I will build one sometime as I am curious- perhaps incorporating some of the clever tricks suggested above. By the way, Merlinb, there are loads of really cheap high gm,low microphonic valves around. I highly recommend,inter alia, Russian single triode 6S3P-EV if you have not already tried it.They will require gridstoppers right on the socket pins.Price is usually between £1-£2.
I have been mulling over (i.e. the stage before design) a phono preamp for some time. I originally intended to start with a 12AY7 cascode, but I was put off by the poor PSRR and the need for a fairly high HT so that grid bias can be sufficiently negative to avoid grid current. This thread has been very interesting.
You make reference to Wallman low noise RF circuits and dismiss the cascode for low frequency service as it is only applicable to RF. Different from your approach I assume you are referring to the circuits in the MIT Series vol 18 edited by Valley and Wallman. It is interesting that the cascode circuit is prominent in Chapter 10 Low-Frequency … and Chapter 11 DC Coupled circuits. Also interesting is the number of other circuits that appear in Chapter 11 that populate popular audio designs; common cathode, Aikido, SRPP, CCS cathode follower, CCS common cathode, mu follower.
From my point of view all these circuits all have merit and are worth playing with. One focus of the evaluation is noise. It not a search for the best, simplest or the like. This is a hobby; I do it for the fun.
I think that is curious and inconsistent that you reject the cascode out of hand. You present option as fact with nothing of substance to back it up. I can imagine you pounding on the table with your fist to make your point. You claim the gm for the common cathode is equal to the cascode. This where I am coming from:
Chapter 11 in Wallman
Eq. (34) the current gain (gm) of the cascade = mu/(rp+((rp+Rp)/(mu+1))))
Eq. (4) the current gain (gm) of the common cathode = mu/(rp + Rp)
For the common cathode the plate resistor is in series with the internal plate resistance of the valve resulting in a reduced gm , an order of magnitude, depending on the value of the plate resistor.
Provide some substance; show how the equations in MIT vol. 18 are wrong. Or is it un-supported opinion?
DT
All just for fun!
You can filter down the line with a local capacitance multiplier to the first stage's feed. Won't need more than one or two high hfe low cob BJTs. A cascode is much wider bandwidth and if quietly fed has a very interesting subjectively open tone. In 30-40dB gain, a triode tube is Miller infested indeed. Closes in. Limits the MM carts loading variation choices (200pF in cables are there already) and/or fights some input Tx secondary optimum square wave termination. We can't dismiss it out of hand if we shoot for best just because it needs care. In another frame of mind, for simpler PSU and good enough all rounder, including noise, I prefer the Mu follower there. Something can be judged only given the brief, the targets, the lengths to be willingly met or not.
Assuming you are quoting accurately, then they too are wrong! Transconductance gm is defined in terms of the output current into a short circuit (or constant output voltage = same thing), not the output current into a specified resistance. See any book on valve theory. I keep saying this, and you keep finding people who apparently don't understand it.
Current gain is not gm. (The current gain of a valve is huge, assuming little grid current flows.) Output current at some random impedance vs. input voltage is not gm. Adding an anode resistor does not affect gm, and so does not affect the noise performance of the valve. It does reduce voltage gain, by stealing some of the available signal current.
I think it's pretty obvious that DualTriode is referring to the 'effective transconductance' available to drive the specified load, not just the gm of the tube alone. But perhaps he should have used gm', to avoid any mix up.
In either example the current gain is practically infinite, which is no use to anyone, so it is still reasonable to talk about 'effective gm' instead, i.e., voltage gain divided by load impedance.
Last edited:
But the "effective transconductance" is the same as the actual valve mutual transconductance, for the common-cathode - that is the whole point! Let us measure mutual transconductance of the circuit, including the anode resistor. To do this we need to keep the output of the circuit at a constant voltage (that is part of the definition of mutual transconductance, after all). Constant voltage at the output means that no signal current flows into the anode load, it all goes out the output into our short circuit. Thus gm'=gm.
Anything else is not mutual transconductance, effective or otherwise. You can't simply redefine terms like this, otherwise language ceases to have any meaning. The mutual transconductance for a circuit is defined in exactly the same way as the mutual transconductance for a valve. The whole valve mutual transconductance is available to drive the load. This is not true for all circuits (e.g. cascode has a small correction term), but it is true for the common-cathode (and the cathode-follower?).
We seem to be going round in circles, so I am not going to argue this point any further. I have said what I have said, and I stand by it.
Anything else is not mutual transconductance, effective or otherwise. You can't simply redefine terms like this, otherwise language ceases to have any meaning. The mutual transconductance for a circuit is defined in exactly the same way as the mutual transconductance for a valve. The whole valve mutual transconductance is available to drive the load. This is not true for all circuits (e.g. cascode has a small correction term), but it is true for the common-cathode (and the cathode-follower?).
We seem to be going round in circles, so I am not going to argue this point any further. I have said what I have said, and I stand by it.
Salas makes a very good point about the Miller capacitance of the input tube which is much worse for the high gm types and the Cg-a figures in the datasheets are not reliable(usually underestimated for some reason??).This was the clinching reason for making my RIAA amp differential(halving Miller C) which is a much bigger undertaking than trying to get a cascode right.
Oh dear,it looks like I have to improve my understanding of Transconductance-back to the textbooks-sigh...
Oh dear,it looks like I have to improve my understanding of Transconductance-back to the textbooks-sigh...
Last edited:
S = Gm = Mutual Condutance, expressed here in mA / V
u = mu = Amplification factor
Ri = Tube internal Plate Resistance
RL = Anode Load Resistance
In my world and many ohers here on the forum, the formula is:
S= u/Rp
And when you add a load resistor, you have to write the formula:
u = S x RpxRL/(Rp +RL)
When you express your supplied formula in red above, that the load resistor RL comes in series to the plate resistance Rp and change the Gm (S) of a tube, yes, I get upset and bang my fist in the table.
Gm is a "tube constant" and not dependable of the load resistor in a CC stage
If you add a non decoupled cathode resistor, the transconductance for the circuit decreases, But Gm for the tube is still the same.
The formula yhou supplied above and i marked in blue, I can not follow at all.....
......I don't have your book...Yes, an intresting discussion in this thread, and I may be I should have put in more IMO's not to upset anybody.
Of course we do it here for fun, but it has also been my profession all my life.
JohanB
Last edited:
Miller capacitance can be a problem with high mu, rather than high gm. Cg-a in the datasheet does not include the wiring, and sometimes does not include the valveholder. I would always add a pF or two to the stated figure. Differential certainly helps, but you gain 3dB of noise. Like all electronics, it is a matter of balancing one thing against another.
Hope you mean "high u"-types (not High Gm)
I use the 5751 (in a mu-follower low current), that is claimed to have between 1,2 to 1,4pF in Cga. U = 70 in a mu-follower, so 1,5 x 70 gives around 100 pF, witch is tolerable. A ECC83 with 1,6 pF & u = 100, will give 160pF. a little over but still OK.
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.
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.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.