Possibly the most "why would you" kind of idea for a tube/tranny hybrid amp

[SK_Caterpilar_SK]

o I have been fooling arround with transistor amps and I had an idea that a DC coupled differential input with triodes in the input instead of transistors could absolutely work.

What I am anticipating is horrid DC drifts and not just as the enviroment changes arround the tube but also with age the tubes degrade in emission. So despite the wacky simulator showing promising numbers where I can compensate the DC offset with the current source of the voltage gain stage In the real world no two triodes will be so well matched as in the simulator BUT...against all the odds...you could have a tube in a transistor power amp that does a lot more than just be a low voltage fed preamplifier.

But considering the reality might just be a total shipwreck I am not suprised why I dont see any documented ideas about this on the internet.

And if you look at the plot of lets say a 12AU7 12AT7 it would be totally doable at these voltages. With probably not bad linearity since the tubes operate in a relatively fixed voltage region with the high gain of the voltage amplifier. So this could be an acceptable scenario where you could use tubes with relatively low voltages. Considering 30V is just about enough for a 12AU7 to conduct a whole 1mA (all though this current doesnt necessarily need to be as high) or the 12AT7 Which can do even 2mA at 30V. So you could avoid high voltages and also use the tube properly (or so to say) (and I dont think a 12AX7 would work at least not with 35V. The model of the tube is supposed to be a 12AX7 but it doesnt copy the plot from datasheets)

Tube officionados gonna hate me but keep in mind this is just an idea that I woke up to from my sleep at 3am, and I will probably never execute it due to the lack of time since im a university student..

(the output emitter follower transistors are already darlingtons with a gain of 1000 just for the test with an artificial bias supply aswell)

[magic]

Tangential nitpick:

This Wilson mirror is not a happy mirror; the left bottom transistor operates basically in saturation. You would need to make the VAS a Darlington pair for it to work well.
Besides, Wilson is not a good choice for this application because once the Vce's of all four transistors are balanced correctly, the two mirror currents are equal and this doesn't leave allowance for base current of the VAS to flow into one branch of the input stage. You will never be able to get the offset to zero (with perfectly matched tubes, as in simulation) and increasing VAS current will only make things worse.
The typical solution used in 99% of opamps* is to have a simple mirror driven by an emitter follower and another identical emitter follower in front of the VAS. See µA741, TL072. With ideal matching, the input pair ends up with exactly equal currents and almost equal collector/anode voltages.

*except those using different topologies altgother, which is frankly a lot of them nowadays, particularly CMOS.

[T3sl4co1l]

Sure, nothing wrong with that.  Just the pitiful performance -- low loop gain, low slew rate, high noise, high input offset...

In particular, the low gain and slew are due to the plate current being most likely under 1mA.  I would guess higher-perveance types are preferable here, like 6DJ8/6922 over 12AX7.

I doubt the offset is actually going to be all that bad, but that said, offsets of like 0.1V would be typical for toobs, and even terribly mismatched BJTs are within something like 50mV.  If nothing else, you'll want to skip true DC operation, like with a beefy coupling cap on the 1k resistor there.

I don't think aging will actually be too bad.  At such low power levels, they'll last forever, and mainly expire from fading cathode emission (likely in many 10k's hr), or suddenly from failed heaters or something (thermal cycling etc.).  Being made at the same time, and operating in the same environment, they'll likely track fairly well considering.

In the transitional era, manufacturers from Philco to Tektronix used hybrid circuits, using, *cough*, well, the best of both worlds instead -- transistors for the low noise, high bandwidth, low voltage, low power, low cost; and tubes basically anywhere that transistors couldn't yet reach, sometimes high power but mostly high voltage.  Which in turn relates to anything driving other tubes, namely televisions and oscilloscopes as prime examples.  You'd see tubes in the HV section, both for power (line output) and rectification (HV/EHT); video/deflection output (high voltage, typically using pentodes with very low plate capacitance, and very high gm), and sometimes in RF or other input sections where transistors weren't quite yet suitable (like your 3HA5 UHF triode or nuvistors, or anything microwave until some time later; and arguably still today in specific microwave applications, like magnetrons).

Tim

[SK_Caterpilar_SK]

Tangential nitpick:

This Wilson mirror is not a happy mirror; the left bottom transistor operates basically in saturation. You would need to make the VAS a Darlington pair for it to work well.
Besides, Wilson is not a good choice for this application because once the Vce's of all four transistors are balanced correctly, the two mirror currents are equal and this doesn't leave allowance for base current of the VAS to flow into one branch of the input stage. You will never be able to get the offset to zero (with perfectly matched tubes, as in simulation) and increasing VAS current will only make things worse.
The typical solution used in 99% of opamps* is to have a simple mirror driven by an emitter follower and another identical emitter follower in front of the VAS. See µA741, TL072. With ideal matching, the input pair ends up with exactly equal currents and almost equal collector/anode voltages.

*except those using different topologies altgother, which is frankly a lot of them nowadays, particularly CMOS.

Youre absolutely right. Damn I wish to reach your level of knowledge at some point right now its just putting lego together for me .

Yes indeed the bottom left transistor was in saturation (hovering over the transistor shows in the info that it was indeed in saturatio during the entire sine wave signal)

I have checked the TL07xx datasheet and upon redoing the mirror as in the datasheet all of a sudden made the simulator run considerably smoother .

However I ran into another problem with the DC offset I will soon discuss on Tim's reply.

[SK_Caterpilar_SK]

Sure, nothing wrong with that.  Just the pitiful performance -- low loop gain, low slew rate, high noise, high input offset...

In particular, the low gain and slew are due to the plate current being most likely under 1mA.  I would guess higher-perveance types are preferable here, like 6DJ8/6922 over 12AX7.
I doubt the offset is actually going to be all that bad, but that said, offsets of like 0.1V would be typical for toobs, and even terribly mismatched BJTs are within something like 50mV.  If nothing else, you'll want to skip true DC operation, like with a beefy coupling cap on the 1k resistor there.

I don't think aging will actually be too bad.  At such low power levels, they'll last forever, and mainly expire from fading cathode emission (likely in many 10k's hr), or suddenly from failed heaters or something (thermal cycling etc.).  Being made at the same time, and operating in the same environment, they'll likely track fairly well considering.

Yes you are absolutely correcct in all of the above however. The12AX7 woudnt allow for just a moderate amplifier but you would need at least 50V on the plate of the 12AX7. While I was right about being able to operate vacuum tubes like this I totally forgot about grid current. Turns out those 50uA are the problem here. It creates a big enough negative voltage across the feedback resistor to become a problem. Running higher cathode currents would induce more grid current as now the grid is positive to the cathode..I am sure you know what I mean by this.

12AT7s are also a very "meh" candidates for the same reason all though much better option than a 12AX7 however the 12AU7 is probablly still okay if I run it at 1mA each triode section could still be out of the grid current zone

In the transitional era, manufacturers from Philco to Tektronix used hybrid circuits, using, *cough*, well, the best of both worlds instead -- transistors for the low noise, high bandwidth, low voltage, low power, low cost; and tubes basically anywhere that transistors couldn't yet reach, sometimes high power but mostly high voltage.  Which in turn relates to anything driving other tubes, namely televisions and oscilloscopes as prime examples.  You'd see tubes in the HV section, both for power (line output) and rectification (HV/EHT); video/deflection output (high voltage, typically using pentodes with very low plate capacitance, and very high gm), and sometimes in RF or other input sections where transistors weren't quite yet suitable (like your 3HA5 UHF triode or nuvistors, or anything microwave until some time later; and arguably still today in specific microwave applications, like magnetrons).

Yes precisely in fact I can relate with my old TESLA BM510 oscilloscope. It had a ECC85 in the input for high input impedance the rest of the signal processing was transistorised and then it hit the horisontal amplifier and vertical amplifier where all the signal processing was done with transistors and EL86 were used as output devices. I very much like these mixed bag technologies. Fun to repair just dont forget to discharge the DC blocking cappacitor of your test gear when you go from tube to transistor. The timebase was built on transistors too.

[Terry Bites]

Have a look at tubecad https://www.tubecad.com/2017/05/blog0380.htm
I dont know much about tubes I was born in the 50's ! Even if the tube currents are the same, won't there be a horrible offset if the gm's aren't identical?
Lest I forget: The ZDS1009 is a ready made mirror, and they're dirt cheap. (120v and up to 1A!)

[TimFox]

A rule of thumb for small-signal triodes is that high-mu triodes should not be operated at "low" plate voltages.  To first approximation, the cut-off grid voltage for a triode is the plate voltage divided by mu.  With only 50 V on the plate and mu = 100, that is a measly -0.5 V on the grid, which gives no "grid base" and probably involves highish grid current.
When operating at plate voltages below, say, 100 V, a medium-mu triode is generally a better choice.
(My favorite is the 7586 Nuvistor, but it doesn't come in duals.)

[Terry Bites]

I wish I still had that 60s edition of Scientific American with the nuvistor add. " Do you really need to pay for miniaturisation?" RCA desparately sticking it to the stupid mainframes made with transistors- You'd be mad to go solid state, obviously....

[TimFox]

Interesting:  I'm not aware that vacuum-tube digital computers ever used Nuvistors, which were more expensive than the usual glass-miniature tubes.  They were originally used for RF applications (low parasitics and high transconductance).
My favorite Nuvistor ads were those that highlighted their radiation resistance, even against neutron flux in reactors.  See page 4 of  https://worldradiohistory.com/Archive-Catalogs/RCA/RCA-Nuvistors-Reference-1965.pdf

[magic]

There is a crude solution to input bias currents (grid, base, gate, etc): if the currents are equal on both sides, make the resistances driving each input equal too and then the errors cancel out. Right now, the feedback side sees 1kΩ in parallel with 15kΩ, but the input sees 0Ω from the signal source - not good. If this is equalized than the offset should disappear in simulation (and decrease IRL).

[TimFox]

With properly-biased triodes (say, with grid-cathode voltage at least 1.5 V negative), the grid current is small, but it can get quite high with, say, -0.5 V bias.  The grid-cathode diode can conduct at small negative bias since the electrons emitted from the hot cathode have some kinetic energy.  For a JFET, substantial input current normally requires positive bias on the gate-source diode.  Of course, with BJTs the base-emitter diode is always forward biased.
Matched source resistances can help, but only if the two input currents are well matched.  For some BJT op amps, the "offset current" is roughly 10% of the "bias current", since the input transistors are well matched.  For many JFET amplifiers, the bias currents (leakage current of reverse-biased drain-gate diode) are low and the offset current is not much smaller.  For the TL07x JFET series, TI's typical spec on the offset current is half the bias current, but the guaranteed values are equal.  For an LM741 BJT amplifier, TI's spec has the offset current typically 1/4 the bias current, with a 2.5x ratio guaranteed.
From my own data, with a few mA of plate current at reasonable plate and grid voltages, grid currents are typically around 10 nA (between 2 and 15 nA depending on tube) after warm-up and outgassing for small-signal triodes.

[T3sl4co1l]

FYI, grid current is due to the balance of electrons emitted from the cathode (at thermal energies), and from the grid (from being heated by the cathode).  Also whatever ions are around, usually not much but significant obviously in a gassy tube.

The effect is, grid voltage typically idles around -2 to -3V, when open circuited.  (This is true even in the absence of plate current; a vacuum tube is a heat engine, if a poor one -- a typical one generates merely a fraction of a microwatt!)  Current increases rapidly as voltage rises, indeed exponentially at low currents, softening to a V^(3/2) power law at higher currents.  (In a triode, the exact curve of Ig vs. Vg will depend on Va as well, of course; this characteristic is most obvious for diodes.)

On the upside, these circuits work just fine with BJTs, and you should have no problem also running a vacuum tube at hFE > 50 say, even at such paltry plate voltages.  Yes, applying hFE to a field-effect device, so what?

Tim

[TimFox]

Under normal operating conditions, almost all of the cathode current electrons go to the plate in a thermionic triode.  For a given plate voltage, the curve of grid current vs. grid-cathode voltage has a zero-crossing at what is sometimes called the contact potential (misnomer) due to the net current of electrons intercepted from the cathode, electrons emitted by the warm grid, ion current from residual gas, and other effects such as secondary emission and photo electrons. At less negative and positive grid-cathode voltages, the current increases dramatically, following Child’s Law for a space-charge-limited (forward biased) thermionic diode.  Sometimes, a small-signal triode amplifier is biased to exploit the negative grid bias from the electrons flowing to the grid and back to the grounded cathode through a few megohm resistor.