Vacuum Rectifiers and SS Diode Emulators

[Mick Bailey]

I had a thought a while back about whether a circuit could be built with selectable I/V profiles to emulate different rectifier types. I didn't get into the details and it may be too involved, but the basic idea was to float a programmable vultage regulator and sample the current (as a voltage) downstream, or maybe a side chain. A microprocessor would monitor the sample and derive a control output for the regulator. A different algorithm for each rectifier type would be selectable from a rotary encoder.

Consideration would need to be given to time constants, hysteresis etc., but the hardware could be quite simple - a regulator, couple of resistors, processor such as an Arduino Nano or RP2040 and a rotary encoder. Maybe $10 in parts. The beauty is in the firmware and flexibility of experimenting without having to subsequently swap components.

[R.G.]

Hmmm. Interesting. Would you just make the regulator impress the ripple on the outgoing voltage directly, or try to have it finesse the voltage/current going into the first cap?

[Helmholtz]

One source suggested that it was more an RF resonance in the wires to and from the diodes themselves.

There might be additional RF resonance by the wiring in the MHz range.
But the large transformer parasitics will cause ringing in the audio range with slow recovery SS diodes.
The measured leakage inductance of 120mH and winding capacitance of 750pF will ring at about 17kHz with a repetition rate of twice the mains frequency.
The ringing/oscillation is driven by the energy stored in the leakage inductance.

Some more reading:
https://dalmura.com.au/static/Power%...ube%20amps.pdf (by MEF member trobbins )
https://www.diyaudio.com/community/t...st-jig.243100/

Does your diode model include reverse recovery?
If so you could add series inductance and parallel capacitance to your ACV source to see the ringing.

[Mick Bailey]

Hmmm. Interesting. Would you just make the regulator impress the ripple on the outgoing voltage directly, or try to have it finesse the voltage/current going into the first cap?

My initial idea was to position the regulator between two caps an use the second as the first B+ node. A conventional adjustable 3-pin linear regulator could be floated. An LM317HV would give more than 50v range, an LM317 more than 30v. The directly addressable programmable regulators can give a wide range and have a very fast response, but I haven't investigated whether they can be floated.

The reference voltage for the regulator is easily generated by filtered PWM (using a simple RC filter). My thought was to effectively create a global I/V feedback loop from the output stage to PSU, with the processor controlling the characteristics.

[R.G.]

But the large transformer parasitics will cause ringing in the audio range with slow recovery SS diodes.
The measured leakage inductance of 120mH and winding capacitance of 750pF will ring at about 17kHz with a repetition rate of twice the mains frequency.
The ringing/oscillation is driven by the energy stored in the leakage inductance.
Does your diode model include reverse recovery?
If so you could add series inductance and parallel capacitance to your ACV source to see the ringing.

Good point; I'll do some checking.

The abrupt slam-off of a standard rectifier could certainly do that. A soft recovery diode would not because its current would taper to zero. This being in series with the leakage inductor would prevent ringing. Snubbing the diode would do much the same, and adding resistance in series with each diode might well do enough damping to prevent ringing.

If that proves out, it's a vote for putting one resistor in series with each diode, instead of one resistor after the diodes. I had wondered why Weber went with two power resistors in the Copper Caps. I put it down to it being easier to get two 10W resistors into the package. Maybe it was for other reasons.

This also makes me consider how a 17kHz damped ring would get into the circuit. 17kHz is not going to do much RF radiation, and any of its rings that re-open the diode will get dumped into the filter cap. That leaves capacitive and inductive emission to get any ringing into the circuit, so it could well make wire layout more critical. Keeping tight loops, or better yet, twisting both primary leads with the CT could prevent much inductive emittsion. The MHz-range ringing can just radiate into the other wiring.

Food for thought.

[Helmholtz]

This also makes me consider how a 17kHz damped ring would get into the circuit.

If the 17kHz ringing actually enters the circuit it would use the same "path" as the ripple voltage harmonics you saw.
I think there's not much chance for either if the filter caps are good, the power tubes are balanced and the layout is reasonable.

[R.G.]

OK, more results. The simulator does model rectifier turn off, as proven by both turn-off currents in the sim and by these nearly disappearing when a fast, soft recovery diode is subbed in for an ordinary diode.

A well-chosen snubber on the standard diode stops all the high frequency stuff, just like the fast soft recovery diodes.

I can still make it ring down near audio by putting in big leakages and self capacitances. I can manipulate the size and frequency of the transformer ringing by selecting leakage inductance and winding capacitance. Snubbing the transformer itself - which is where the ringing comes from - seems to work for the transformer ringing.

The ringing can also be reduced by damping with external resistors, one in series with each diode. The winding resistances don't damp the ringing because the leakage fields are external to the winding resistance (I speculate!) as is the self capacitance, so only a resistance outside the transformer can effectively damp it, and it seems to work better with a resistor in series with each diode, not with a resistor in the CT. I think this is because there are really two windings, each with a self capacitance and also with an interwinding capacitance. That's how I modeled it, anyway.

It seems that there are multiple techniques: (1) snub the diodes (2) use fast, soft recovery diodes (3) use a resistor in series with each diode ( 2 to 4 ohms seems good) (4) snub the transformer. In the real world, the very slow recovery of a vacuum rectifier happens to be so soft that the ringing gets choked out by the recovery itself.

[Helmholtz]

In the real world, the very slow recovery of a vacuum rectifier happens to be so soft that the ringing gets choked out by the recovery itself.

A vacuum rectifier doesn't store a charge like a SS diode, so there are no reverse recovery effects.
Consequently no ringing.

Will you show results?
What's your simulator?

[R.G.]

I did some more research. That's nearly always a good choice.

After tinkering with diodes, snubbers, and so on, I found that a snubber network across the power transformer itself was more effective than tinkering with snubbing the diodes. That led me to research, and that led me to an article by Morgan Jones in LinearAudio on the exact topic. He dug deep into the issues, and at last came to an unexpected conclusion: snubbing the transformer with a 1nF and a 1K was all that is needed.

Hard to believe, so I put those into the sim and .... he's right. I put a 1nF and a 1K across each half of a 330Vac winding and rectified per previous runs. The ringing at turn off of the diodes dropped to about 500u of stuff under about 1V at the biggest peak. This was slightly better than fast/soft recovery diodes. In all of these, no tricks I could play looking for the ringing on the filter cap/load resistor showed any of the ringing. It appears that any transmission of the ringing - and probably the ripple voltage partials too - are almost entirely a capacitance or magnetic field issue in the wiring.
i
I use Multisim. Formatting the results from screen shots and such could take a while. Meanwhile, a 1K and a 1n is incredibly easy to test.

Now back to faking vacuum rectifiers by curve matching.

[soundmasterg]

A lot of people when they talk about rectifiers at least for guitar amps talk about the great sag of a tube rectifier, even though a GZ34 isn't going to have a ton of sag since it is pretty efficient for a tube rectifier. A 5U4 or 5Y3 would obviously have more sag. They also forget about voltage doublers which can give some of the same sag effects, but won't eventually fail like the tube rectifiers. I haven't checked into it, but I would guess the transformers designed for voltage doubler use would probably be cheaper than one meant for use with a tube rectifier? I don't know if the voltage doubler would ring like some of the circuits you have been playing with R.G., but it would be interesting to see the effects in comparison. There are the two diode half wave circuits, and the four diode full wave circuits to consider. I know for myself, I gutted and modded a Bogen CHB100 that used a doubler. I made it a four diode doubler and used my own circuit in the amp, and it sags nicely and has a lot of touch response when being used hard, and I won't have to worry down the road about a failing tube rectifier, which I like of course.

Greg

[R.G.]

Pretty much all transformers will ring due to the parasitics Helmholtz mentioned.

The transformer windings have an effective capacitance from turn to turn. It's a distributed capacitance and not (to me at least) simple to model on a detailed basis. On power transformers, it's usually represented as an end-to-end capacitance across a winding, sometimes as a single capacitance across two windings, as in a CT output. I tried both ways, and also the one-cap-per-winding variant with an overall capacitor across both windings. My intuition said this probably would not matter to the ringing, and it didn't. Leakage inductance is modeled by an external inductor in series with the winding. This is remarkably accurate in sim and in the real world.

Parasitic ringing happens when the current through the leakage inductance(s) gets hard-stopped by the reverse recovery of a rectifier. The section of winding that is then conducting gets its leakage inductor whacked by the diode stopping, and it rings through the connected parasitic capacitors. This happens effectively outside the windings.

So I think a multiplier arrangement would still ring when interrupted by an abrupt diode reverse recovery. From a few runs with the "magic" 1nF/1K snubber, I think it would also be effectively snubbed. I'll do multiplier setup when I can spend some more time on the sim.

I think that much of the sag of a vacuum rectifier comes from the slope of the I-V conductance curve. That sag does scale with a lumped resistor and an SS diode.

As to the multi-breakpoint model, I used a diode - zener - resistor for the breakpoints. There are probably better ways to do this, but this was simple and effective to model. The idea was that each newly-conducting section added its resistor-worth of conductance to the already conducting ones. So the sum of all the conductances comes out to cause the I-V slope of the resistive section of the rectifier being modeled, and the zeners give the offsets to each breakpoint.
First section: a single diode and a resistor; the resistor is large, and models the high-resistance slope of the vacuum rectifier beginning conduction
2nd section: another diode, a zener, and a resistor; The zener prevents conduction until further out the voltage axis; when it conducts, the resistor/conductance in series with it is added to the conductance from the first diode.
3rd through Nth section: another diode, a zener, and a resistor; these add additional conductance at each breakpoint as the zener conducts.
The sum of the conductances add up to the slope of the to-be-modeled vacuum rectifier.

I worried that this would need a large number of sections, but in effect, it was very close with only three sections.

I have not yet sized the zeners and resistors for power rating and so forth. I believe that in practice you only need one of these networks, fed by two SS diodes. There's more work to do to get an effective implementation from real world parts (that is, available from Mouser). I'll take a look at that next, along with some pictures.

[R.G.]

I dug through a sizing for part dissipation and fitting to standard parts. Here's the interim model for a 5AR4.



I think this will do a good job, especially since a diode plus a resistor does a reasonable fake for listening, given comments on the Weber Copper Caps.
Prices are from Mouser, in singles. Making over 10 at a time of this circuit would drop the price 30-50%, even more in higher quantity.

[R.G.]

What's the next most popular vacuum rectifier type?

[Tom Phillips]

What's the next most popular vacuum rectifier type?

Since small low powered guitar amps are becoming so popular, I submit that the 5Y3 is a good candidate.

[Helmholtz]

I'm interested in seeing the ripple voltage harmonics' spectra (measured at the 47µ reservoir cap) with 5AR4, SS diodes with an 68R series resistor and your 5AR4 emulation circuit.