Use a high transconductance low noise type like the D3A(7721) or E810F(7788) and it should work quite well based on my very, very limited experience. Input capacitance will be quite low with just a little care.
Another option is a cascode built around the ubquitous ECC88, Russian 6S3P or 6S4P, or the 5842/417A - any of these will also give you low miller capacitance as the lower tube in the cascode will have relatively low effective mu.
Gains of 46 - 50dB or more are achievable in the first stage with care.
Another option is a cascode built around the ubquitous ECC88, Russian 6S3P or 6S4P, or the 5842/417A - any of these will also give you low miller capacitance as the lower tube in the cascode will have relatively low effective mu.
Gains of 46 - 50dB or more are achievable in the first stage with care.
Hi Gluca,
No, I actually mean pentode connected D3A, and 7788.. Or alternately triode cascodes which could incidentally include a D3A in triode connection.
I made the mistake of using the D3A in triode in the design prior to my current muscovite design (analog source) and the total miller capacitance and strays at more than 300pF is a killer.. This is not the best sounding phono stage I have designed by a long shot, how it interacts with the cartridge is a big problem eliminating most MM from consideration..
No, I actually mean pentode connected D3A, and 7788.. Or alternately triode cascodes which could incidentally include a D3A in triode connection.
I made the mistake of using the D3A in triode in the design prior to my current muscovite design (analog source) and the total miller capacitance and strays at more than 300pF is a killer.. This is not the best sounding phono stage I have designed by a long shot, how it interacts with the cartridge is a big problem eliminating most MM from consideration..
I guess the reason you're asking is noise. With triodes, to minimise noise you need to operate at the lowest possible anode current you are happy with. High gm helps too, but anode current is the main thing to think about.
With pentodes there is additional noise introduced due to screen current partition. To minimise that you need to minimise screen current relative to anode current. In other words, you need to run with low anode current, very low screen current, and high gm is a bonus if you can get it.
Having said that, pentodes are probably not as bad as some would have you believe. For example, the equivalent input noise of an EF86 over the audio band is only about 2uV, whereas a reasonable triode might be 1.5uV. (OK, a really well chosen triode might achieve 0.5uV, but be prepared to pay for Telefunkens or something!)
You could of course use a triode at the input, but pentodes for later stages where the noise is no longer important, but then you may have to think about linearity instead.
With pentodes there is additional noise introduced due to screen current partition. To minimise that you need to minimise screen current relative to anode current. In other words, you need to run with low anode current, very low screen current, and high gm is a bonus if you can get it.
Having said that, pentodes are probably not as bad as some would have you believe. For example, the equivalent input noise of an EF86 over the audio band is only about 2uV, whereas a reasonable triode might be 1.5uV. (OK, a really well chosen triode might achieve 0.5uV, but be prepared to pay for Telefunkens or something!)
You could of course use a triode at the input, but pentodes for later stages where the noise is no longer important, but then you may have to think about linearity instead.
Last edited:
TANSTAAFL will be with us forever. Name your poison: pentode partition noise or triode Miller capacitance. Cascodes are interesting because CMiller is low, along with avoiding partition noise. The "gotcha" with a cascode is abysmal PSRR. IMO, hybrid cascodes of either permutation deserve consideration. Allen Wright successfully stacked a high gm triode on top of SS, but a high voltage MOSFET stacked over a high gm triode bears looking into too.
A high μ, high gm, low RP, low CMiller triode to keep in mind when scoping a RIAA preamp out is the 6GK5. A low noise pentode to consider is the Octal 6AC7. Don't forget the silicone rubber damping rings and DC heater supply.
A high μ, high gm, low RP, low CMiller triode to keep in mind when scoping a RIAA preamp out is the 6GK5. A low noise pentode to consider is the Octal 6AC7. Don't forget the silicone rubber damping rings and DC heater supply.
Yes, I saw his presentation at ETF and read the article. What's interesting (to me) is that he hasn't really overcome the partition noise issues in pentodes. His circuit, at essence, is a cascode with a bipolar on top, triode on bottom with the screen used as a plate. The plate in his pentode has hardly any effect on the circuit- I built one on the bench, disconnected the plate, and got almost identical results.
This isn't that surprising since using the screen of a pentode as a plate in low noise circuits is a long-known trick.
No, that's the same topology, though with a MOSFET on top rather than a bipolar. If you do the AC signal analysis, you'll see that the plate hardly enters into it and that (with an adjustment to the DC current), it can be disconnected with the circuit still operating the same way.
If you can tell me the plate voltage on the circuit you drew, I'll walk through the calculations.
edit: One more question- in his ETF lecture and the Linear Audio article, Frank used a resistor load. Has this changed or was the CCS your addition? It would seem to make the gain and output impedance crazy-high.
If you can tell me the plate voltage on the circuit you drew, I'll walk through the calculations.
edit: One more question- in his ETF lecture and the Linear Audio article, Frank used a resistor load. Has this changed or was the CCS your addition? It would seem to make the gain and output impedance crazy-high.
And yet one more question... what is the impedance looking in to the source of the MOSFET (since you have your simulator out and running)? I did my original analysis and measurements with the resistor plate load and bipolar driving the screen, as in Frank's original presentation, where the calculations were much easier!
)? I did my original analysis and measurements with the resistor plate load and bipolar driving the screen, as in Frank's original presentation, where the calculations were much easier!The anode/plate does play a role: it collects electrons. If you disconnect it then g2 has to collect all of them, which in some cases might cause it to overheat. Of course, this won't be a problem if you are running at low current.
I think this pentode/cascode hybrid was invented too late. It is of limited value for audio because of 1/f noise, which it does not improve. It might have been useful for HF radio front-ends, where partition noise does matter, although it is unclear to me what advantage it has over a normal cascode.
I think this pentode/cascode hybrid was invented too late. It is of limited value for audio because of 1/f noise, which it does not improve. It might have been useful for HF radio front-ends, where partition noise does matter, although it is unclear to me what advantage it has over a normal cascode.
Thus my point about DC. If a D3a is running 20mA plate current, then g2 current is going to be about 6mA. So disconnect the plate, run 6mA to the screen from the top cascode device, and the performance is about the same. No overheating.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.