Theory, Practice and Sexlife of the "Reflected Plate Amplifier" Hybrid, low Voltage Tube circuit

[Brian Roth]

I'm just saying a cathode resistor without a parallel lytic is indeed local NFB?

 

[Mr. Moscode]

It is a current mirror with gain. A current mirror does not need a second transistor per se, that is only needed to get accuracy.



Yes.



For the Aphex circuit, yes. In absolute terms, no. It is completely possible to make this a non-NFB circuit.



Yes, it is in effect equivalent to 2-Stage RC coupled amplifier, without the RC coupling.

If you use a P-FET with a largeish unbypassed source resistor in the 2nd stage and then a resistive load to the negative rail and sufficiently high voltage on the positive rail (say +48V/-12V) you have two stages no (looped) feedback.

Or, you could use resistive Miller feedback on the second stage and a CCS load for the second stage. Then you have the tube stage open loop and the Solid state par local loop, with low output impedance and low input impedance, forming a virtual cascode for the tube input. We could even do it differential.

Lot's of interesting things to be done on this principle.

Thor

If I read you right, using a p-fet would eliminate the current mirror and the tube would act as single stage triode with a wide voltage swing on the plate, eliminating the "no miller effect" advantage of the circuit.

 

[gyraf]

I'd still simply recommend listening to how it sounds before throwing too much energy its way

 

[JohnRoberts]

I certainly didn't say that, nor write it!
You may want to check your translator, posting here is (almost) strictly english.

+1 there is no rule strictly prohibiting posting in languages other than english but maybe we need one.

I have declined to approve a handful of new member submissions written in languages I could not read.

JR

 

[Mr. Moscode]

I'd still simply recommend listening to how it sounds before throwing too much energy its way

Absolutely. I won't use silicon in the audio path unless it passes a listening test.

 

[thor.zmt]

If I read you right, using a p-fet would eliminate the current mirror and the tube would act as single stage triode with a wide voltage swing on the plate, eliminating the "no miller effect" advantage of the circuit.

Well, a lot here boils down to semantics.

If we use a P-Fet, then we have (say) 3V gate threshold voltage. With (say) an ECC88(6DJ8/6922) running at 44V/0.5mA with ~ -2V Grid Bias, we get an anode load resistor of 6kOhm, which is very low, though not as low as a BJT Base.

In either case the miller effect will be modest. If we use a cathode bias resistor to.a -12V rail, this would be 27k, with a suitable capacitor on the cathode to ground, we get (open loop) a mu of ~28 and around 3kOhm anode impedance.

Effective gain to the P-Fet gate is ~ 18 (25dB) and miller amplified input capacitance ~30pF.

Our P-Fet is actually a bigger problem, if we use IRF9610 we have 170pF input capacitance plus miller amplified 15pF reverse transfer. Let's say we have a 2nd stage gain of 100, we are looking 1.67nF effective capacitance and thus 16kHz open loop bandwidth for 65dB open loop gain. Not terrible, but also not great.

If we use my special "Magic-Fet"(TM) it's only < 600pF for 44kHz open loop bandwidth, or alternatively more loop gain. Add a N-Channel MOSFET as follower, close the loop and we have something close to classic 3-stage low power tube amplification, say for example the pulse technology here:



Of course, we could also operate the tube in a folded cascode. I will review these options more when I get time.

Thor

 

[Mr. Moscode]

Well, a lot here boils down to semantics.

If we use a P-Fet, then we have (say) 3V gate threshold voltage. With (say) an ECC88(6DJ8/6922) running at 44V/0.5mA with ~ -2V Grid Bias, we get an anode load resistor of 6kOhm, which is very low, though not as low as a BJT Base.

In either case the miller effect will be modest. If we use a cathode bias resistor to.a -12V rail, this would be 27k, with a suitable capacitor on the cathode to ground, we get (open loop) a mu of ~28 and around 3kOhm anode impedance.

Effective gain to the P-Fet gate is ~ 18 (25dB) and miller amplified input capacitance ~30pF.

Our P-Fet is actually a bigger problem, if we use IRF9610 we have 170pF input capacitance plus miller amplified 15pF reverse transfer. Let's say we have a 2nd stage gain of 100, we are looking 1.67nF effective capacitance and thus 16kHz open loop bandwidth for 65dB open loop gain. Not terrible, but also not great.

If we use my special "Magic-Fet"(TM) it's only < 600pF for 44kHz open loop bandwidth, or alternatively more loop gain. Add a N-Channel MOSFET as follower, close the loop and we have something close to classic 3-stage low power tube amplification, say for example the pulse technology here:

View attachment 109546

Of course, we could also operate the tube in a folded cascode. I will review these options more when I get time.

Thor

Thanks for your thoughtful reply. Don't mean to take you off your usual duties - beer and the beach!

Your P-fet is actually a mosfet not a J fet?
What's your "Magic Fet"?
Is that schematic your design?

Opinion part, not aimed at anyone:
According to the Aphex paper the idea is supposed to outperform a standard tube design like the one you posted but it's 80% solid state. Outside of a newbie project for low voltage tube fun, I don't see the point except getting rid of the output cap with a servo. But then again, as you know, I'm a tube guy. A 12ax7 bypassed K gets you 35 DB of gain as in your posted schematic. 2 in a row is 70, (which in my experience is too much if you're crunching the overall gain down to 20 db with nfb. It will sound tight or congested. Use a 12AU7 instead. I did this with a Thor preamp. Owner was thrilled! )

BTW, I used a mosfet follower in my Minuet in A preamps and they work well and sound good. B+ was 335, regulated.

 

[thor.zmt]

Your P-fet is actually a mosfet not a J fet?

BSS84

Is that schematic your design?

Which one? What I described in the text? Yes. I will draw it and demonstrate later, when the weather becomes too bad for the main job (Monsoon season is coming in Thailand).

According to the Aphex paper the idea is supposed to outperform a standard tube design like the one you posted but it's 80% solid state.

Well, there is a bit of hyperbole, but I have used my circuitry derived from the first principles shown in many products, invariably to substantial critical and customer acclaim.

Outside of a newbie project for low voltage tube fun, I don't see the point except getting rid of the output cap with a servo. But then again, as you know, I'm a tube guy. A 12ax7 bypassed K gets you 35 DB of gain as in your posted schematic. 2 in a row is 70, (which in my experience is too much if you're crunching the overall gain down to 20 db with nfb. It will sound tight or congested. Use a 12AU7 instead. I did this with a Thor preamp. Owner was thrilled! )

Well, I find making tubes run and sound good on SELV interesting from a commercial standpoint.
If you take the All in One I designed (Speaker with Wifi, Bluetooth and hybrid Tube + Class D Amp and a Speaker arrangement for wall-to-wall stereo from a 60cm wide box) wich uses a commercial pre-certified 12V supply, one set of certification headaches are removed by staying below 50V DC.

BTW, I used a mosfet follower in my Minuet in A preamps and they work well and sound good. B+ was 335, regulated.

Yes, they can work very well, just watch capacitances.

BTW, using a P-MOS as folded cascode, we could make a "pure" Tube 2-Stage + follower Amp running from +48V/-12V, DC coupled in to out to boot. Not that I see this as inherently superior design to Tube -> PMOS -> NMOS follower, but why not?

Thor

 

[Mr. Moscode]

BSS84

Yes, a mosfet.

Which one? What I described in the text? Yes. I will draw it and demonstrate later, when the weather becomes too bad for the main job (Monsoon season is coming in Thailand).

The one in the post is what I referred to. 12AX7 w/AU7 follower.

Well, there is a bit of hyperbole, but I have used my circuitry derived from the first principles shown in many products, invariably to substantial critical and customer acclaim.

The low noise claim would be valid because the tubes in a lo Z environment.

Well, I find making tubes run and sound good on SELV interesting from a commercial standpoint.
If you take the All in One I designed (Speaker with Wifi, Bluetooth and hybrid Tube + Class D Amp and a Speaker arrangement for wall-to-wall stereo from a 60cm wide box) wich uses a commercial pre-certified 12V supply, one set of certification headaches are removed by staying below 50V DC.

Yup.

Yes, they can work very well, just watch capacitances.

BTW, using a P-MOS as folded cascode, we could make a "pure" Tube 2-Stage + follower Amp running from +48V/-12V, DC coupled in to out to boot. Not that I see this as inherently superior design to Tube -> PMOS -> NMOS follower, but why not?

Playing fast and loose with the term "pure tube" eh?

Thor

You're in Thailand? I thought you were in the Netherlands. An assumption on my part.
Weather the storms well!

 

[thor.zmt]

Playing fast and loose with the term "pure tube" eh?

Well, if we use a MOSFET as folded cascode, is that REALLY different from using a resistor anode load and a coupling capacitor (other than cascode connection)?

The output current will be essentially 100% modulated by the tube (ok, a bit of noise much lower than from our tube).

You're in Thailand?

Yup.

Weather the storms well!

I have been in south east asia for over a decade and half. Back in Hongkers we used to get proper Typhoons, the weather at the Thai Coast so far proved much more sedate.

Thor

 

[Mr. Moscode]

I just saw the schematic in the earlier post is a Pultech. Duh.

 

[thor.zmt]

I just saw the schematic in the earlier post is a Pultech. Duh.

And I propose that my ECC88 (or ECC83) input, BSS84 2nd stage and stn1hnk60 follower is conceptional and performance wise very parallel to Pultec, but runs with +48V/-12V and is DC coupled (except input).



R11 adjusts open loop gain and thus overall HD levels and spectrum. R58 is adjusted for 0V at the cathode to allow noise free gain switching.

As shown here, open loop gain is 70dB with >20kHz open loop bandwidth using 5670 or 6DJ8 or similar tubes. The tube operates at ~ 42V/0.5mA, with 5.7V Heater.

With me, normally the circuit would operate from 12V with an inverting Buck-Boost switcher providing -12V and a Boost switcher with charge pump providing +48V. All parts except the tube would be SMD.

Thor

 

[Mr. Moscode]

And I propose that my ECC88 (or ECC83) input, BSS84 2nd stage and stn1hnk60 follower is conceptional and performance wise very parallel to Pultec, but runs with +48V/-12V and is DC coupled (except input).

View attachment 109664

R11 adjusts open loop gain and thus overall HD levels and spectrum. R58 is adjusted for 0V at the cathode to allow noise free gain switching.

As shown here, open loop gain is 70dB with >20kHz open loop bandwidth using 5670 or 6DJ8 or similar tubes. The tube operates at ~ 42V/0.5mA, with 5.7V Heater.

With me, normally the circuit would operate from 12V with an inverting Buck-Boost switcher providing -12V and a Boost switcher with charge pump providing +48V. All parts except the tube would be SMD.

Thor

Pretty cool. The one thing it won't do that the P-Tech can do is wide voltage swings without clipping. 250 V B+ can move a lot of electrons.

Have you built it or is this just on paper?

 

[thor.zmt]

Pretty cool. The one thing it won't do that the P-Tech can do is wide voltage swings without clipping. 250 V B+ can move a lot of electrons.

Have you built it or is this just on paper?

Somewhat different (mostly less gain, additional features) but fundamentally highly similar circuit, yes and had several 1,000 sold.

Review: iFi Audio iTube2 - Sound Perfection Reviews



Production ended when stocks of NOS GE 5670 ran out. It could have been continued using 6N3 from russkii or Chinese stocks, but these sounded enough worse to cancel the product.

Thor

 

[Mr. Moscode]

Nice. You have 70db of gain in a box that uses 9 db? How do you do bass boost? Passive eq, or NFB boost?

You probably know about our Super It phono pre. Up to 70 db gain, hybrid tube/fet. 3 tubes, but should be 4. Would you want to make it and sell it? I know you probably have your own version. Highly repected.

 

[thor.zmt]

Nice. You have 70db of gain in a box that uses 9 db?

Actually in this box the resistor that sets OLG is switchable, so feedback levels and switchable from 60db to less than 30dB. But the output stage still gets the full OLG for linearisattion.

How do you do bass boost? Passive eq, or NFB boost?

NFB Boost.

You probably know about our Super It phono pre. Up to 70 db gain, hybrid tube/fet. 3 tubes, but should be 4. Would you want to make it and sell it? I know you probably have your own version. Highly repected.

I have my own designs for this. For example:

Abbingdon Music Research PH-77 Phono Equaliser

While in production, Stereophile Class A without advertising, FWIW.

I no longer have any connection with if/AMR, BTW.

Thor

 

[Mr. Moscode]

Actually in this box the resistor that sets OLG is switchable, so feedback levels and switchable from 60db to less than 30dB. But the output stage still gets the full OLG for linearisattion.



NFB Boost.



I have my own designs for this. For example:

Abbingdon Music Research PH-77 Phono Equaliser

While in production, Stereophile Class A without advertising, FWIW.

I no longer have any connection with if/AMR, BTW.

Thor

I would bet a 12au7 would be good choice for this. And replace the follower fet with the other half of the 12au7. Of coourse you're up to 2 tubes now but tubey-er, what ever that means. On this forum they are considered filters and not all that transparent.

Nice work. You were a hired gun? Did you license them out? Or employed? Too personal?

 

[thor.zmt]

I would bet a 12au7 would be good choice for this.

Framed Grid "High Perveance" tubes are a better choice at low voltage. Hence the 5670. A 5687 would be a much better choice than 12AU7 (or 6SN7).

I'm not a great fan of the 12AX/12AU7 (or 6SL7/6SN7) range.

The AMP PH-77 used four parallel 2SK170 in a folded cascode in the first stage, load being a tapped resistor chain for fine gain adjustment and implementing the 75uS EQ, followed by a buffered grounded grid second stage using 6072A (or 12AT7 variants like 5965 etc) followed by the 3180/318uS EQ and a 5687WB Mu-Follower.

It would have not made STP CA with 12AX7 and 12AU7 instead of 6072A & 5687WB, if only because you cannot ship NOS GE tubes as standard with gear using these tubes.

Nice work. You were a hired gun? Did you license them out? Or employed? Too personal?

Company co-founder. Eventually we had too many disagreements.

The Show Must Go On (click link to read)

Thor

 

[gyraf]

Aah, the scope of this discussion is much clearer to me now - it's intended / aimed for a hifi-audience?

Note that they hold beliefs and aesthetics quite different to those of pro audio: There is some overlap, but not at all as much as you'd think..

And yes, I agree that this way of low-voltage-using real tubes is a great idea for the hifi segment, and listening to what it does - in that context - really makes sense now

/Jakob E.

 

[thor.zmt]

Aah, the scope of this discussion is much clearer to me now - it's intended / aimed for a hifi-audience?

No, it can be applied just as much in pro audio. The fun is, that by adjusting operation conditions for the tube and open loop gain a wide range of sound characters can be obtained, adjustable/switchable even.

And yes, I agree that this way of low-voltage-using real tubes is a great idea for the hifi segment, and listening to what it does - in that context - really makes sense now

I think your view is excessively cynical, Re "hifi".

Relatively low voltage tube circuits are sonically valid, in any context, as they do not run "starved anode", are excellent for DIY, due to the lack of dangerous voltages, more so for beginners.

As long as 48V are present why not a DIY API 500 rack tube mic pre, or gain make-up Amp for a passive summing mixer?

Tube sound does not come from using high voltages.

So I'd be happy to (for example) discuss a mic pre with classic Sziklai differential Frontend as "electronic transformer" and after 26dB of gain there, two stages of "hybrid tube" gain.

Will it sound like a TAB V76? Not really, though it could be tuned if desired to sound s lot like a TAB V76. But if you really want a TAB V76, buy the original, no DIY copy in 2023 will be accurate enough.

Thor