Amusing hybrid circuit

Oops, I thought of an improvement maybe you could bootstrap the op amp's ground connection with a PNP and the signal to the 6AK5 grid.

Depends on the value of the plate resistor.

Nah, need to replace the cathode follower with MOSFET follower, plus MOSFET current- sink pulldown. Good combo of technologies.

--
 Thanks, 
    - Win

I've done that, cascode signals out of the power rails of an opamp. That can create an essentially perfect class-AB power amp with a microscopic crossover zone.

But those tubes will never make 100 MHz. Even 30 MHz was tough with

6AK5s.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

The op amp supply rail is bouncing around as the plate current increases and decreases, at low frequency the PSRR handles that so long as the supply to output compliance is maintained (or just doesn't try to cross over in the case of a RR output amp.) but at high frequency PSRR gets worse and it looks like will degenerate the gain, here.

Aw shucks only 30? The op amp does all the power gain and there's no Miller effect going on here with the tubes that I see. Well I'll sim it and see what it's like there at least I have those tube models such as they are.

The 5654 is an industrial/military 6AK5 with a different pinout I think

Yep, an op amp is a five-terminal device, and the trick is valid. Not sure how it'll work for you, though, because the op amp bias currents aren't always well-controlled. You'll possibly have a few milliamps quiescent current at your cathode, is that OK?

Similar schemes are useful for making low-dropout regulators.

If you run the negative current from the op amp through a PNP current mirror, that outputs the 'error' bias current (i.e. the positive-rail current that doesn't flow through the resistor on the op amp output).

If push-pull is wanted, another op amp can have the output resistor-tied to its positive rail... and the negative-rail current mirror can run a second cascode.

One key figure of merit for a tube is Gm/Cp. Given some plate capacitance Cp, you need a very low plate load resistor to get the time constant down. At 100 MHz, the plate resistor will be so small that there's no voltage gain. Inductive peaking helps some.

You can model the tube as a Gm (VCCS, g in Spice) with added plate capacitance to ground, maybe some plate resistance and input/miller capacitances too.

Tubes had millisiemens (millimhos) of transconductance; pitiful. Mosfets and phemts and such have whole digits of Siemens.

The later frame-grid tubes were better than the older stuff for wideband amps.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

It will need a lot of mA to get the Gm of the tube up. Maybe add some resistance across the opamp rails.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Rbias sets the quiescent current thru the tube which flows down thru the op amp positive rail and out thru Rbias. The quiescent current thru the tube creates a drop across the plate load, that's then divided down and filtered to a DC bias voltage fed to the non-inverting input, which then acts as self-bias to set the current thru Rbias and all should work nicely for DC.

At least for signal coming into the op amp creating a varying supply current down thru the op amp positive rail the tube is just functioning as a current buffer I don't see where trans-conductance is coming into play there, all the signal current that flows into the op amp supply terminal must flow thru the 6AK5 plate resistor what else can it do?

You'd still want a high bias current to get good transconductance because negative feedback is being sent to the 6AK5 grid out the 12AT7 cathode but the movement of the 6AK5 plate and its grid are in-phase, so the Miller effect isn't happening there.

Aside from the screen that is but the screen current should be some small fraction of the plate current and approximately constant in normal operation

No, the tube's "transconductance" is 1 / Rbias,** and the VAS gain, or voltage-amplifying-stage, is Rplate / Rbias. To get much gain you'll need a high Rplate. The gain bandwidth will be limited by all the capacitances, tube and wiring.

** As we point out in AoE, this signal-out-the opamp's Vcc pin trick was taught by the famed Robert Widlar, about 45 years ago. The tube is in cascode mode.

--
 Thanks, 
    - Win

It also sets the effective transconductance of the opamp.

The quiescent current thru the

The impedance looking up into the cathode is 1/Gm. That looks, to the opamp V+ rail, like a resistor and a bit of capacitance. Fast opamps won't like seeing that as a supply. If Gm is 5000 uS, it looks like

200 ohms.

Given that the grid of the 6AK5 is not at AC ground, there is still Miller capacitance. That will affect the feedback loop from the 12AT7.

The circuit is amusing but not useful.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

No not many circuits involving 6AK5s are in 2019. The Chinese equivalent still seems to be made in large quantity and you can get 'em 2 for $3 delivered...from China. Actually I guess they're Russian by way of China, 6Zh1P-E.

They seem to be well-made and work fine I wonder if they use them for anything military/industrial over there, still, or if it's just for the audio market.

More like 5MHz, maybe 10 or 20 with pitiful voltage gain and peaking.

The opamp end can be arbitrarily fast, but no matter what you do, Miller or not, the couple 4pF of plate capacitance, plus g-a of the follower, loads the gain node. For typical loads of 5-20k let's say, that's a cutoff of

2-8MHz. With Gm around 5mS max, loads below 2kohm don't do any good.

Doing it with a planar triode (2C39, etc.) isn't too bad, but it's still not a particularly efficient use of electricity. A handful of BJTs easily outperforms!

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/

I think that's pessimistic, and my claim was optimistic. The pentode "cascode" stage coupled thru the power supply has a lot of low-frequency gain. The negative feedback to the grid trades that off for bandwidth.

I ran some sims with the best tube models I could find, with the LT1223 as the op amp, 100 volts on the plates 75 for the screen, can get 6 dB of broadband gain into the 10s of MHz, it drops to 3dB looks like at about 40MHz. Will surely be worse IRL. So I think better but nothing miraculous

As the op amp should be able to do that no problem by itself, so only good for "level shifting"

Is it pessimistic if it's measured personally? Is it optimistic if it's unrealistic? :^)

Note that the negative feedback isn't an argument that defeats the node capacitance * load resistance. It's not even a bootstrapped circuit!

If it were a mu follower for example, with... gosh, I wonder if you could even bootstrap a screen to any practical effect? The Miller effect would be worse, but then, you avoid that with cascode, so you could have a pentode in bootstrapped triode mode, as it were. Voltage gain is lower too, because screen is an input, but maybe a combination of cutoff frequencies could be adjusted to give a modest gain (2ish?) at useful frequencies (40MHz?). Still can't do anything about follower input capacitance.

But yeah, that would help, but that's about as much as you can possibly get. Shades of Phil's favored bootstrap-everything strategy, except you don't have a good enough follower here to yield much (triode CF gain = 1 - 1/mu even for CCS load).

Anyway, the more fundamental quantity is the gain-bandwidth product. You can have tons of gain through the power supply, absolutely -- but you need a huge resistor (or CCS) to do this, and then it rolls off really low (100s kHz?). Same thing about dominant pole opamps, innit. You can apply NFB, or you can employ lower resistances, but you still get the same cutoff either way -- it seems you can't fool the GBW more than a small amount at a time, and with great effort (peaking, distributed amp).

This applies also to narrowband (tuned) circuits, where the bandwidth, rather than the baseband cutoff, is limited by the exact same relationship (load resistance and stray capacitance).

It does start to fall apart for diffusive devices. For the 2N7002, there is a dominant rolloff in the low 10s MHz, but it does keep on going -- at lower and lower system impedances and gains -- into the 100s MHz, along a roughly diffusive (sqrt(f)) path. Very fast devices that match to system impedance effortlessly (MMICs?) seem to go on much further than their packaging, or low-frequency-equivalent specs, would suggest.

Tube models are weird. I'm not aware of any that are truly good. Most are very quirky with respect to, say, screen voltage. If you got the right bias voltages, it probably worked out okay.

I've been meaning to write some myself, based on virtual cathode theory, but it's a very low priority so I haven't started.

FWIW, 6V6/6AQ5 is impressively close in specs to 6AK5, 5702 and such -- adjust capacitances as needed, of course.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/

A triode might be better instead of a pentode. When the plate voltage drops below screen voltage, the cathode current goes to the screen supply, breaking the signal path. In the small pentodes, there is little risk of overloading the screen, but with larger tubes even that needs to be taken into account.

--

-TV