Thx Hearinspace, this class of idea is neat. Avoids miller C and can have gain higher than mu in triode mode, actually gm * Rload - this is also interesting when looking to save a stage in triode mode.
Hi Merlin and Luckythedog
Thanks for your replies and links, I somehow missed the discussion. I will look into it, it will take a while to pick up all the details 🙂
Nice Sunday! Erik
Hallo Merlin, obviously yo havn`t understood the BestPentode circuit. It does make a big difference if you connect a pentode as a triode or as a pentode - and the BestPentode circuit is still a pentode connection! In triode connection the anode voltage will change dependent of the input signal and amplification factor µ (or µg2g1 in triode strapped pentodes). Therefor you get the miller capacitance. This miller capacitance is removed if the anode voltage is kept on a fixed potential (cascode operation) or if you use the screen grid of a pentode and this screen grid is kept on a fixed potential. Exactly that is done in BestPentode mode - the screen grid potential is fixed. Consequently there is no Miller capacitance! And even better: the transconductance of the pentode is enlarged by the added screen grid current and the partition noise is removed. Of course you could connect the pentode as a triode and cascode it. But then the whole anode + screen grid current has to be supplied by the cascoding transistor. In BestPentode connection it`s the screen grid current only.
All of these facts are written in Linearaudio Vol.0. It`s very likely that you havn`t read it. I can assure you: there are no such things like "advertisements" - despite of that there are true measurements of real world tubes and circuits.
My own BestPentode based phonostage has a THD of -106dB (k2) and -118dB (k3) at 10mVrms input signal / 1kHz and a gain of 46dB. The noise level is lower than that of OPA637 based circuits. You can change the tubes like you want: the gain is almost the same, the noise will differ a bit (+/-3dB worst case) - but on an extremely low level. Try to beat it with your old fashioned ECC83! And please don´t forget to mention, that you need at least 10 pieces of ECC88 to find one having acceptable low noise!
This phonostage has been playing in my home for more than 3 years - without any headaches, but a lot of fun and listening pleasure.
Thats very interesting. Which pentode do you use? And which transistor for best pentode operation? Could you provide schematic?
I have read the article. It is a cascode, albeit a strange-looking cascode, so I can understand why it is not immediately obvious to you. The screen grid is being used as the 'anode' of a triode. If you disconnect the pentode's true anode, you will get the same performance: low noise, no Miller C. The pentode's true anode does nothng very useful in the Best Pentode circuit, it mainly just drains extra current and increases the Miller C slightly.
The OPA637 has about the same level of noise as an average ECC83, so it is not difficult to beat the OP637 with a reasonably quiet triode.
If you're measuring distortion to -106 and - 110 +db presumably your noise floor is lower. This I'd like to see!
Hallo Merlin,
No, that`s wrong. If you disconnect the anode, you can drain the screen grid current only, which is a fraction of the anode current - otherwise you will overload the screen grid and damage the tube. Much lower current means much lower gm (lower "leveraged" signal current per input signal voltage).
The gain of the pentode, cascode and BestPentode circuit equals to:
Vu = gm x Ra (Ra = effective anode load)
=> much reduced gm means much reduced gain!
Please explain why the Miller C is increased when the anode is used in BestPentode operation - and please tell me, how large is this Miller C and how large is the increase of it in absolut values from your point of view!
No, that`s wrong. If you disconnect the anode, you can drain the screen grid current only, which is a fraction of the anode current - otherwise you will overload the screen grid and damage the tube. Much lower current means much lower gm (lower "leveraged" signal current per input signal voltage).
The gain of the pentode, cascode and BestPentode circuit equals to:
Vu = gm x Ra (Ra = effective anode load)
=> much reduced gm means much reduced gain!
Please explain why the Miller C is increased when the anode is used in BestPentode operation - and please tell me, how large is this Miller C and how large is the increase of it in absolut values from your point of view!
Hallo Merlin, sorry I forgot one thing: I wrote my circuit has lower noise than OPA637, that`s true. The OPA637 has an equivalent input noise voltage of about 8nV/SQRT(Hz) @ 1kHz. My BestPentode circuit has an equivalent input noise voltage of ca. 2.0nV/SQRT(Hz) @ 1kHz using the D3a. Using ECC83 you could expect >10nV/SQRT(Hz) typical - some devices have up to 40nV/SQRT(Hz), some about 8nV/SQRT(Hz). ECC83 is definately not a low noise tube - its impedance is just to large.
In the case here with an E180F in straightforward pentode mode, Req wideband spec is only 460R anyway, which is fine in my book in this application. Audio noise performance of the valve will likely be dominated by flicker noise anyway, which will be the same in either mode after adjustment for gain I think. And in simple pentode mode there's ample gain, so although one could operate in cascoded triode mode I don't see why one would, when simple pentode mode seems to fit the bill well here. Besides, when the source is MM phono, there seem little point in scraping the last dB of HD or noise performance I think. Simple pentode mode still seems the elephant in the room to me......
PS Req = 460R in pentode mode equates to 2.7nV/rtHz BTW...........
PS Req = 460R in pentode mode equates to 2.7nV/rtHz BTW...........
Last edited:
True, the gm would be reduced, but I suspect the effect on the signal-to-noise ratio would be small (not least because flicker noise would be reduced).
It depends mainly on the pin-out of the valve. If the anode and grid pins are close together, you might have 0.5pF capacitance between them, which will be multiplied by the Miller effect. If the anode is grounded, that capacitance is eliminated. With the EF86 I have seen a 20% reduction in input capacitance by grounding the anode, but other pentodes will be different.
Yes, that is about right for a D3a in triode mode!
And please don´t forget to mention, that you need at least 10 pieces of D3a to find a pair having reasonable matching and acceptably low microphonics! 😉
You must have some very noisy specimens. An average ECC83 is closer to 6nV/SQRT(Hz). Even an ECC81 is not much worse than 8nV/SQRT(Hz), usually (it depends on current, of course).
Last edited:
I think the motivation is that in cascode mode the improvement in SNR is potentially more than just a couple of dB, but several dB. Having said that, you are right, a good pentode is not outrageously noisy when used with an MM cartidge, but it's not great either.
And a MM or MC cartridge is usally more noisy than the amp itself when you play the record
It depends on the record you play if it is of bad quality your amp won't fix that
It depends on the record you play if it is of bad quality your amp won't fix that
Dunno, humble EF184 has wideband noise Req 300R ie about 2.1nV/Hz in pentode mode.........I wouldn't call that noisy by any stretch - secret is in gm@Ia I think. But in audio applications, flicker noise probably dominates device noise and I don't think any of these methods is an escape for that. Microphony is the audible EF184 Achilles point IME, but thermal/flicker noise hardly seems bad at all to me in pentode mode if gm and Rp are good IME. You're right,of course, triode modes are quieter as to wideband thermal noise but in the case of E180F it's 225R plays 460R, and in nV/rtHz terms that's worth about 3dB - if flicker noise doesn't dominate anyway. And in MM phono that should be lost in the wash I think, whereas in MC phono where one is trying to not make a bad situation worse it might matter.
The one I'm listening to now was a little troublesome to set up at first but that was because I used a shunt regulator circuit that I hadn't tested very thoroughly before putting it on the base of the cascode PNP. (I still don't really know how to use spice so have to build and try) Since fixing the regulator the circuit has been stable, quiet and very musical. My guess is that probably the higher the valve gm the better because as the gm goes up the RIAA eq circuit/ouptput impedance can come down.
Hallo Merlin,
Sorry, I could not respond earlier - too much work before Christmas...
Back to the real numbers:
First the D3a and almost all other high-gm pentodes (E180F, E280F, 6SCH11P, EF184, C3g,...) have no such grid-anode pinout! Despite of that the grid1 is typical embedded between 2 cathode pins and located far away from the anode pin. So the real capacitance between anode-grid1 and therefore the Miller capacitance as well is negligible.
And it`s right for BestPentode mode as well! Exactly that is the advantage: you get the low noise of triode mode + the gain of Pentode mode + extra gain of added screen current! Best of two worlds! And very low input capacitance as added benefit!
In my circuits no matching is needed because I get at least 80dB of gain using one D3a and the closed loop gain (which is much lower, of course) is set by means of feedback. Yes, I use lots of feedback and it works great! The circuit is completely DC-coupled, extremely stable and reliable. And it plays music!
Regarding microphonics: That is a common problem with almost all tubes (of course some tubes show extremely microphonics like old balloon types without any mechanical fixation at the top like 27, 37, Siemens Aa and so on). For high gain phonostages one should use damped tube sockets and there is no problem anymore.
My amp is ready for any shootout, in terms of measurement performance as well as listening tests. Come on, bring your classic ECCxx-something amp and we will see - and listen...
Sorry, I could not respond earlier - too much work before Christmas...
Back to the real numbers:
First the D3a and almost all other high-gm pentodes (E180F, E280F, 6SCH11P, EF184, C3g,...) have no such grid-anode pinout! Despite of that the grid1 is typical embedded between 2 cathode pins and located far away from the anode pin. So the real capacitance between anode-grid1 and therefore the Miller capacitance as well is negligible.
And it`s right for BestPentode mode as well! Exactly that is the advantage: you get the low noise of triode mode + the gain of Pentode mode + extra gain of added screen current! Best of two worlds! And very low input capacitance as added benefit!
In my circuits no matching is needed because I get at least 80dB of gain using one D3a and the closed loop gain (which is much lower, of course) is set by means of feedback. Yes, I use lots of feedback and it works great! The circuit is completely DC-coupled, extremely stable and reliable. And it plays music!
Regarding microphonics: That is a common problem with almost all tubes (of course some tubes show extremely microphonics like old balloon types without any mechanical fixation at the top like 27, 37, Siemens Aa and so on). For high gain phonostages one should use damped tube sockets and there is no problem anymore.
My amp is ready for any shootout, in terms of measurement performance as well as listening tests. Come on, bring your classic ECCxx-something amp and we will see - and listen...
Ah, if only that were so in conventional triode mode........only true in pentode mode unfortunately - or if the anode is held at constant potential in triode mode. Besides, parasitic Cag1 can't come from inter-pin capacitance in triode mode, because that is also present in pentode mode where typ Cag1 is 0.018pf for an E180F according to spec.........
- Status
- Not open for further replies.