Fairchild 670 PSU for tube preamp

[Walter66]

Hello,
having choosen the Fairchild 670 regulated power supply for my next project, a high current consuming tube preamp.
The 670 compressor has unregulated push-pull tube output stages and so I wanted to source the 310V needed for the outputs unregulated, but filtered. The resistors are for channel separation and lowering the output voltages to 310V.

The rest will use the regulated power for line and phono stages, it just needs some stage decoupling resistors and channel separation.
Fairchild don't have real channel separation, but my understanding of this unit is limited. The supply goes directly into the two channels B+.

Can it be done in a proper way to establish two separate good filtered PSU sources with my schemo?
Can the output voltage be 250V with this type of 5651 voltage reference tube? Any additional changes in the regulator needed for this to happen?
Any improvements on my circuit will be appreciated. Thank you.

 

[metalockpick]

Not sure how you'd be using EL34 tubes in a preamp...

If you didn't need EL34s and 400+VDC, I might suggest using a transformer with a 240V secondary, then run a bridge rectifier after that. 230-240VAC, being very commonly available as either an isolation transformer for countries that use 230-240V, or was a step up from 100/120VAC to 230-240.

You'd get ~340VDC in a well filtered supply, and you could do a number of things to regulate the HV, if you choose to. Inductor chokes were cheap and capacitors were expensive back in the day.
A series diode will do more to cut ripple than a poorly tuned choke.
Use good capacitors & proper grounding (Negative of the bridge rectifier should connect directly to the negative of the first brute force filter cap, or with a large diameter wire, as this is where the main noise/ripple will be).

You could run a separate 6.3V filament transformer and a separate HV transformer, which in this case might save you some money, and allow separate AC turn on on and B+ voltage turn on.

 

[Walter66]

Thanks for your suggestions. I use a pair of 807 tubes and need only 310V B+for them, unregulated.

This shown circuit for regulation uses EL34, not my preamp circuit. I designed a passive PSU for the preamp, but a regulated one has better line and load regualtion, lower ripple current. I don't want to use bridge silicon rectifiers in my gear, only tube rectification will do.
This project is not on a budget, this is about to do the best things and to make use of the best circuits for an all tube regulated PSU with proper channel separation.

 

[CJ]

why not use the regulated supply for the whole thing?

it has a a semi wide voltage range, maybe 220 to 270, so getting 250 is no problem.

some say the gas reg tube is noisy but i hear nothing. and it looks cool so put it in a window up front maybe.p.

cross talk should be inaudible at those low power levels required by preamp.

 

[Cqwet Dbdfte]

I have been using HV MosFet + Lm317 + zener etc for 30+ years for low noise regulation. Tube rectifiers (GZ34 Types) only redeeming feature is slow startup, easily done with above MOSFET anyway, use silicon rectifier and gain some voltage, save the heater power for less transformer losses, add power resistors in series with silicon diodes to emulate voltage drop in tube rectifier if you need the first R in RC network, and trade in that 20uF cap for 10x or 20x more capacitance.
Power supply only need to put out low noise DC with minimal load fluctuations.
SiC diodes like vacuum rectifiers have no reverse recovery.
An all tube regulated power supply will be as big as large as the typical power amp. No need for that.

 

[Walter66]

Thanks, do you mind showing a circuit, so that we could better understand?

 

[Cqwet Dbdfte]

You can probably find it looking at my other posts.
Also check out

 

[Walter66]

Did a search for all your 100+ posts, no PSU schemo visible.
You meant something like this one?
I read that LM 317 is very noisy ?

 

[Cqwet Dbdfte]

Those Utube videos pretty much covers it. Your schematic above misses a cap on R1 to ground. C2 could well be 10uF, I use a 7uF film cap, because I have a bunch of them. Not much noise survives after a decent cap on the output.

The typical LM317 has 40 dB gain at 10KHz, most of its noise is very low freq, below 10Hz, just as the typical CLC network being at the mercy of the power company and local loading conditions. After 10KHz noise performance is handled by typical RLC elements.
Define your requirements, how much noise can you tolerate.

The various versions of 317 has different noise performance, again dominated by bypassing the "Adj" pin. You can get better noise if the reference voltage is RC filtered, not possible in the 317's.
Datasheets generalize 317's noise output as "RMS output noise, % of VOUT, 10 Hz ≤ f ≤ 10 kHz, 0.003% ,Typ."
Is it good enough? It depends on how you use it and what your requirements are.

 

[Walter66]

Good video, thank you.
Why you got rid of C8 in the video circuit? Please elaborate. The video says, it's an improvement on the circuit.
I wonder whats the best place for the cap multiplier, in front of the regulator or behind it (to remove all the rest of the hum after the reglulator)?

 

[Cqwet Dbdfte]

You can try yourself. The schematic above has problems, LTspice defaults voltage sources as "voltage sources" which is theoretically correct, but it ignores reality. All LTspice examples (10,000's) have this error. It has bitten me a few times.
To repeat, 0 ohm voltage sources is a software guy's wet dream.
With a typical voltage source, with resistance, in my sim like a vacuum rectifier, it makes more sense to have C7 large, as makes the FET a capacitance multiplier, and as a slow ramp as you want because the gate current is effectively zero.
Also using low resistance values for R6, needed with no load stability, reduces effectiveness of C6.
C8 will make it harder for U2 to regulate M1's source.
Also notice the AC voltage injected into my voltage sources. It reflects DC with AC components from an actual power supply. I also included some inductance, for kicks, to remind me of leakage inductance. Diode models may not accurately describe the noise effect of some diodes. Use SiC diodes for minimum reverse recovery time.

 

[Walter66]

Thank you. I saw in Shindo Labo preamps, that he put the cap multiplier always after the tube regulator. To me, this makes sense. Shouldn't the voltage be regulated and then filtered for best SNR, not the other way around? Do you see your method as favorable and if so, why?

 

[Cqwet Dbdfte]

You could add more RC filtering after a regulator, but once you hit like -120dB noise rejection it loses meaning. The 317 type regulators only has so much gain to counter noise, they run out of gas fast after 10KHz. Then the caps provide low impedance to ground. Prefiltering for the reg makes its life easier. The first RC eats a lot of garbage and reduces peak currents from rectifier and caps.
A nice OP amp plus a low noise reference would give better performance, the 317 is just a convenience.
If your load is constant current, like mine, you already have better power supply rejection than a plain resistance load or a cathode follower, (unless the two use the same value for plate and cathode resistors.)

 

[Brian Roth]

Just my 2 cents re. capacitance multipliers, which are excellent in many applications. Beware of the fact that the simplest versions are NOT short-circuit proof, as drawn in post #10. Post #9 seems to include protection for the "pass" transistor.

A typical over-current or short at the output of the simplest cap. multiplier usually causes the pass transistor to fail with an internal short from input to output.

Don't ask me how I learned that...<g>.

Bri

 

[Cqwet Dbdfte]

...also the peak current in your C8, (ESR and capacitance dependent) could make life hard on the FET. The NPN current limiter chokes off any excessive currents fast, and any long term shorts would heat up the thermistor shutting down the FET.
The Zener voltage and the cold temp value of the thermistor sets the current limit.

 

[Walter66]

Why not use a TIP112 Darlington instead of the FET? I have those at hand and want to try it out. Safety diode included.

 

[Cqwet Dbdfte]

Not an option. The pass device has to handle rectifier peak input voltage from 0V. Caps are 0V at startup.
Use at least 500V devices.
Bipolar has no distinct advantage over MOSFETs.
Been using 800V+ parts since 1990.
100V parts is no-go.
Use fully insulated TO220 for simplicity, unless you are burning off gobs of power into a heatsink, where you need to optimize thermal transfer.

 

[Walter66]

Thanks for your advise.
Your MOSFET is stated as obsolete and very hard to source (long lead time). Can I use

AOTF9N70 instead? Will the circuit have to change for different type MOSFET?​

What use is MPSA42, please? I just want to add a simple cap multiplier in front of the regulator..would like to simplify a bit.

​

 

[Cqwet Dbdfte]

The NPN is a current limiter. Do not omit.
Use the 6 ohm thermistor.
The FET and C2 is the capacitance multiplier. Don't put any cap between the 317 and the FET.
It would reduce the effectiveness of the 317.
Any HV FET with low gate capacitance/gate charge should work. Try running the SIM with the FET of your choice, if you can get the SPICE model.
C1 and C2 I would use HV film caps, 5 - 10 uF is ok.
This is as simple as it gets.
R1 is higher than the typical data sheet value, as it is used to establish minimum load.

Not sure if this link works
https://www.digikey.com/en/products...IcHWc8QgwvZ3JIJXwlSKWC0xJYtnUUDAGi1kFClDMZiAA
I'd try the STF10LN80K5.
Check your results with a scope, and don't get zapped!

 

[Walter66]

Thanks for your advise, much appreciated!