Fairchild 670 PSU for tube preamp

[Walter66]

B+ power is not reference to just one polarity of the signal, and all noise on B+ is cancelled at the transformer. So there is no power supply noise on the output. So its not considered reference because noise from the supply is not permanently added to the signal.
Look at the fairchild 670 schematics, it has a similar affair going on in its circuits. A lot of the main stream somewhat low end pro audio like API312, LA2A, 610, 1176, 1073 are quick and easy unbalanced circuits. They were cheap to make and later easy to market as a kit.

The main advantage of push-pull designs seem to be a simpler PSU construction because power noise cancels out itself. A clever design this Fairchild.

What I haven't understood: Is the filter function of a serial regulated tube PSU improved when the voltage above the series regulator becomes higher or why do Fairchild put approx. 200V working voltage on the EL34 and have choosen exactly this working point?

Why they didn't came up with the same current, but only 100V above the tube? Does it have to do with using the most linear section of the tube output curves? And do the regulator tube use this section in the same way an amp tube does? I think, they don't do. They should have a stable working point with minimal variations, right?

So why should one choose 200V anode to cathode voltage over 100V or maybe 150V? For better regulation, for better SNR outcome or for more linear behaviour of the series tube when parameters in use are changing?

 

[drtechno]

The main advantage of push-pull designs seem to be a simpler PSU construction because power noise cancels out itself. A clever design this Fairchild. And one with a heavy use of excellent iron, too.

Its not really push-pull in the classic sense because all voltage reference points (negative going signal, Positive going signal and signal crossing) are all referenced to B+ and then its null out by the transformer. Push pull, you actually taking an unbalanced signal then converting it to balanced, but it still retains the unbalanced component. That common noise on the unbalanced component is subtracted along with the unbalanced negative feedback signal.

 

[drtechno]

What I haven't understood: Is the filter function of a serial regulated tube PSU improved when the voltage above the series regulator becomes higher or why do Fairchild put approx. 200V working voltage on the EL34 and have choosen exactly this working point?

Why they didn't came up with the same current, but only 100V above the tube? Does it have to do with using the most linear section of the tube output curves? And do the regulator tube use this section in the same way an amp tube does? I think, they don't do. They should have a stable working point with minimal variations, right?

So why should one choose 200V anode to cathode voltage over 100V or maybe 150V? For better regulation, for better SNR outcome or for more linear behaviour of the series tube when parameters in use are changing?

When I was able to discuss that with someone, they told me the EL34 acts more like a constant voltage source than a regular.

 

[Walter66]

When I was able to discuss that with someone, they told me the EL34 acts more like a constant voltage source than a regular.

Maybe this refers to it's line regulation rather then load regulation, which doesn't seem to be necessary with this kind of circuit (constant load under all conditions)? But the regulator has both implemented, line and load regulation.

 

[Cqwet Dbdfte]

A center tapped transformer feeding power to two tubes does indeed cancel PSU noise, but it also cancels even order harmonics, which if "low enough" would not be a problem but would change the descending even/odd harmonic sequence, leaving mostly odd harmonics. This may be the reason some favor single ended power amplifiers.
Loudness perception is related to this harmonics structure.
A true differential amplifier, input and output is if course a possibility, avoiding any reference to "ground". Very low noise RF oscillators use this idea. "Ground" in the signal context also includes the power supply voltage.
If this is a worthwhile idea to pursue for tube audio I don't know.

 

[Markku01]

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 ?
View attachment 119317

Just look online for a Maida High Voltage regulator. They all have similar schematics to this one. And current limiting is usually provided by a resistor between the two source of the MOSFET and the input of the LM317 or similar voltage regulator. I believe that the calculation of the current limiting assumes that the voltage regulator has a minimum voltage drop. Sometimes it's 3vdc . And the MOSFET threshold might be 3.5vdc . Thus if you place a 9.1V zener for D3 then , 9.1- (3.5+3) =2.6V.
Thus if you place a 39 ohms resistor there , current limiting will occur at apx 66.66 ma . I believe this is the way it's supposed to work.
For the Maida regulator you just need to be sure your MOSFET can handle DC current for the maximum voltage across the MOSFET. You can find that information on the MOSFET curves.
Not all manufacturers will list the DC safe area, so you may need to look for it.

 

[Cqwet Dbdfte]

If the short is long term that scheme will not work well as it would make the voltage reg into constant current reg, putting a big thermal load on the FET, inclusion of a thermistor + NPN shuts off the gate in my scheme.
Thermistor should be bonded to the FET.
The 32K 10W resistor from input to output bypasses the regulator.

The thermal load on the 317 is VERY small, as it only regulates from a low voltage in the FET's gate circuit.
No heatsink needed. The cap on the 317 input reduces its gain and capacity to regulate. The resistor in front of it is not an RC thing, just a current sensing part. There is plenty of mfg. appnotes on "high-volting" the 317, including one using a 4kV vacuum tube.

That scheme above also misses out on using the FET as a cap multiplier, and a cap on the 317 adj. pin. The low values of the voltage divider is likely to satisfy the 317's need for a minimum load, but would reduce the effectiveness of a cap here.

Using silicon rectifiers could include the RC network starting from from AC secondary windings, thus reducing ripple currents into the big caps, and minimizing the excitation of leakage inductance in power transformer.
I'm using plain 1N4007's feeding 1500uF photoflash caps at 500V in a power amp, a CL-60 (current limiter) on the mains transformer primary, which eats the inrush current spike.
Energy storage is about 180 joules.
No PSU HF is visible on the scope.

 

[drtechno]

A center tapped transformer feeding power to two tubes does indeed cancel PSU noise, but it also cancels even order harmonics, which if "low enough" would not be a problem but would change the descending even/odd harmonic sequence, leaving mostly odd harmonics. This may be the reason some favor single ended power amplifiers.
Loudness perception is related to this harmonics structure.
A true differential amplifier, input and output is if course a possibility, avoiding any reference to "ground". Very low noise RF oscillators use this idea. "Ground" in the signal context also includes the power supply voltage.
If this is a worthwhile idea to pursue for tube audio I don't know.

It may look like ground is a component, but only with bias. All of these circuits have one rule though is not to add power supply noise to the signal.
The mic pre was just an example. But the DC power supply ground is not earth bonded in a balanced circuit because some designs will use it as anti-phase feedback. But active negative feedback is what really changes the harmonic output besides transformer construction. I noticed push pull output transformers being associated out there with balanced output. But they are different and the transformers are constructed differently. Those audio output amps, are push pull because they are (non-gaped) lamination and usually all the same type, balanced output are gaped and are not always the same core materials fo you will see a 30%-60% nickle plated or nickle or some use what they call radio lamination. Also, balanced designs will sometimes have constant voltage sources and constant current sink that ensures balanced operation.

 

[Cqwet Dbdfte]

In the DISA amp balanced was more of a design objective, not a true differential, maybe overkill for the job. Input and output tubes each have some common cathode resistance, but not a constant current source, which could have been done with a single resistor returned to negative 150V or some such.
The resistors on the input trafo looks like a load for it, clearly the transformer windings puts only a few (hundred) ohms to ground.
The input trafo could have been between grids only, so differential CMRR was not much of an issue.
Unbalanced DC currents in a transformer benefits from a gap to increase the threshold for saturation.
Negative feedback is indeed linked to harmonic profile.
https://www.passlabs.com/technical_article/audio-distortion-and-feedback/

 

[Walter66]

What's the use of C2 in this Maida voltage regulator, please?
Does it function like a cap multiplier in combination with the MOSFET? If not, what's the purpose of Q1 anyway?
I calculated R6 for an output voltage of 350V as to 280K Ohm and want to give this a try.

 

[Cqwet Dbdfte]

Not a cap multiplier, but a better ground reference for the LT317's reference. Improves it a great deal, as covered in mfg data sheets/app notes.
As I stated previously, adding a cap to ground on R1/D3 would make Q1 into a common gate amplifier, (with controllable risetime) with U1 working on the source, hence any cap here would slow it down.
If you want to lighten the work for Q1/U1 put an RC network before it.

 

[Walter66]

Thanks, got one more. That's a cap multiplier in front of the Maida regulator, isn't it?
Why not use it this way? What makes the previous circuit without cap multiplier preferable (we can delete C8, if you don't like it).

 

[Cqwet Dbdfte]

C7 behaves like a cap multiplier, I'd use a ~5 - 10uF film cap, with R5 min 100k, maybe 500k to get slow ramp HV.
Adding more cap multiplier could work, I use hundreds of uF with 300-ish resistor RC input network.
C8 is BS.
C6 can be 10uF in series with 100 - 300 ohm for lower noise.
The 317 is an amplifier, not a resistor.

 

[Walter66]

So this last circuit doesn't need the input FET to improve ground reference for the 317?
To use a cap multiplier vs. the circuit using IRF840, which one is the better idea, technically speaking?
I'd like the idea to have a cap multiplier wipe away all dirt before the regulator and I suppose the IRF isn't supposed to do this job.

Of course, one can install an LC or RC network in front of the regulator, but this would rob much of the elegance of this design, doesn't it?
A big choke is expensive and space occupying, resistors need additional voltage for the PSU. Not the best solutions IMHO.

 

[Cqwet Dbdfte]

Your first sentence makes no sense.
The input 800V Fet is not related to the 1.25V reference.
In my opinion using a LT317 with a HV Mosfet is better than a cap multiplier, because it also regulates the voltage.
There are more variables, I would not use the 317+FET for a power amp. What is the load, constant current or a spiky load? Microvolt signals or millivolt signals, switching artifacts?
The 3 youtube videos in the thread covers it nicely.
Use LTspice to simulate your circuit then get your soldering iron warm.
Technically speaking, I only speak technical.

 

[Walter66]

A friend of mine lately compared different PSU for his SET (20mA load) tube preamp. He talked me into using this Maida version because he likes it best.
My own tube preamp will consume approx. 50mA constantly. Phono MM input stage, approx. 8V max output voltage.
Unfortunate I'm not a LT spice user, it would take some time to work into this material and after getting experience, its all theoretical. I try to understand those circuit, but in the end, the practical outcome means sound is most important.

 

[Cqwet Dbdfte]

"Sound" is secondary to the PSU circuit design.
Power supplies have no "sound". Zero AC is the target, and the AC voltage on the output is the figure of merit.
Easy to measure. Nothing subjective here.

 

[Wilhelm]

Thanks, got one more. That's a cap multiplier in front of the Maida regulator, isn't it?
Why not use it this way? What makes the previous circuit without cap multiplier preferable (we can delete C8, if you don't like it).
View attachment 122809

High voltage supplies based on transistors, ic's etc, often need some protective measures added..The function of D4 , C7, R5 and M2 , is to limit the voltage across the LT 317, during start-up and turn-off. Best regards

 

[Cqwet Dbdfte]

I ran a sim on the regulator above, it showed a 100dB rejection of 120 Hz ripple voltage, using a ~10uF cap on C7 improved about 20 dB. No C8.
A cap multiplier is a simple device, whereas an amplifier w/reference has much more effect, LT317 has over 80dB rejection, gradually dropping to 70dB at 10kHz.

The 6.2V zener may not be enough for using a HV FET depending on its threshold voltage. The 317 needs to dropout min ~2.5V , 5V better, plus the threshold of the FET, commonly around 3V, so a 10V or so zener would work.

A cap multiplier is a simple bandaid we can leave out in a discussion about robust voltage regulators.

 

[Wilhelm]

In fact, introducing the C7 will be somewhat detrimental to the operation of this part of the circuitry. The mosfet is in there to handle the steep voltage-rises and drops at turn-on and turn-off. Attaining close to zero ripple, before voltage-regulation, should be fairly easy with the compact electrolytics available now-a-days. If that's a concern, the easiest way to get less ripple in the proposed "Maida" schematic, before the regulator, would be to use a normal bridge-rectifier ( with a different, higher voltage transformer ), instead of the voltage-dobbler arrangement, which was probaly only used because a relevant transformer was not at hand that day Mr. Maida made the thing ( ! ) Voltage-dobbler psu's will require ridiculously more Electrolytics/capacitance than the most commonly used bridge-rectifier psu's, to attain the same ripple-levels. Best Regards