Vacuum Rectifiers and SS Diode Emulators

[R.G.]

Here's a quickie.

[Tom Phillips]

Wow! I'd say that the results are are pretty much identical or at least within the probable variation within each device type.

[R.G.]

I was surprised at how much they matched. I checked to see that I hadn't mixed up the screen shots.

[Helmholtz]

For reference, how bad is the simple solution SS diodes + 68R resistor?

[R.G.]

For reference, how bad is the simple solution SS diodes + 68R resistor?

I have in mind adding all of the graphics to a several-page article making these comparisons and the simple solution ((SS+68R) is a good one to do as well. It may take a couple of days to get that all together, as I have some things to do that will hamper my fun time. I appreciate your comments a great deal. You led me to make the graphics that illustrate the spectra. What else would be a good addition?

@ everyone: I'm acutely aware of the golden rule for circuit simulation - simulate, but verify. Modern simulators are very good, as long as the user puts in a decent model to simulate. It's really important to do real-world checks to see that you haven't missed some part of the model. I don't have amps available that use the 5AR4 to do the real-world test of switching between a real 5AR4 and the network, nor a real spectrum analyzer to read it. If you have any urge to do so, please try this out and let me know if I hosed it up.

[Merlinb]

Hmm, doing a few sims I see exceedingly little difference in the 5AR4 versus the simple SS+resistor, even with an unusually low-resistance transformer. I can't really envisage a practical use-case that would justify all those zeners?

[Helmholtz]

This might boil down to the question which of the simulators has the "better" GZ34 model.

[R.G.]

You're right Helmholtz. Something like this is filled with assumptions and depends a lot on what tools you use. All simulators are not equal, and the exact modelling depends on how you use the sim. I have a suite of tricks that I use on my sim to get more believable results. SPICE tricks are a long-running area of study for people who use it; I've been running SPICE since I had to submit punched cards. And the quality of the models is paramount. I would never have bothered with this before I found the equations-based model of vacuum rectifiers that reproduced much of what tube data said. I would have chucked it instantly. And frankly, I won't be happy with the results until I can get measured I-V curves for real rectifier tubes. Simulators can trick you. I have scars in odd places from that experience.

So, Merlin - You're right, this is a tricky place to judge what's good enough. Justified? Not justified? I personally think SS diodes plus sag resistors are good enough for me. Other people might like to experiment further.

We're juggling between people who are fine with SS diodes alone and Golden Ears who claim they can hear the amount of oxygen in copper wires. Ultimately, whether it's worth it or not is not a clear black and white. It depends on the listener, and guitar amp listeners have proven very resistant to people telling them what is good enough. I've tried to be clear that this is a foggy region.

I kind of fall in with you that the returns are diminishing after you snub the transformer, use soft recovery diodes, and introduce some sag with series resistance. But "worth the effort" is very hard to quantify with purely subjective issues like amplifier tone. So the diode-zener-resistor network is something that a purist might like to experiment with. The marginal rate of return is tricky. Real 5AR4s cost US$25 to $50 or so. Maybe a budget purist might like to get most of the vacuum value at 1/5 of the price and never need a replacement tube. Value is in the pocketbook AND mind of the purchaser.

My ideas are that:
1. There's a clearly audible difference between a tube rectifier and bare SS diodes with no softening added. The B+ is higher, and harder, and does not sag as much. Some people describe this as "sterile" or "lifeless" and some like the "punchier bass". Those are both sides of the same coin, and whether one is better or not depends on whether the user prefers muffins or doughnuts.
2. A degree of sag can be put back into SS diode variants by a simple series resistor or two. Whether this is good enough is again a matter of taste. Now that I think about it, this test might separate out people who are listening for sag versus some other ill defined quantity.
3. The complicated diode-zener-resistor network is an exercise in determining whether a listener insists on previously knowing that the rectifier is vacuum vs. solid state. I'm surprised at how very close the I-V curve matched network can match what vacuum rectifiers do. I shouldn't be, I guess. The non-time-specific I-V curve is after all what a device does at a fundamental level. The spectrum from the ripple voltages is pretty telling. If there were funny breakpoint effects and such, it ought to come out there. I did a thought experiment along these lines: assume that vacuum rectifiers do in fact affect the tonal qualities of an amp's sound; if that's true, and it's a result of the laws of physics, not magic, how can that possibly work?

If it's laws-of-physics, it's gotta be either voltage, current, or EM field emissions. EM field is a very dark horse, because of the low frequencies of the rectification process, diode recovery effects considered separately. We're left with volt/amp magnetic or capacitive or conductive effects getting from the rectifiers to the B+ and then into the ... er, tone. Magnetic is heavily wire-routing dependent (and is how I got into this mess, thinking about where the wired need to go), so it's a poor candidate for general tonal effects. Capacitive is possible, but the output of the first filter cap only has about 50Vpk-pk on it. Possible, and frankly the most likely source of "tone interference". Conductive is by far the most likely way to get rectifier-sound into the "tone".

That's the reasoning that led me to look at the spectrum of ripple voltages. I was astounded at the match that I-V curve matching achieved in the spectra.

[pdf64]

I have a Weber copper cap WZ34 somewhere, when the outer tube became detached I recall it was just a pair of silicon diodes and a series pair of 39R power resistors, so 78R total.
The WZ34 has noticeably more sag than a real GZ34, it reduced power output about 20%.

[Helmholtz]

Hmm, doing a few sims I see exceedingly little difference in the 5AR4 versus the simple SS+resistor, even with an unusually low-resistance transformer.

Looking at your table all essential harmonics are the same or even lower level with diode+57R compared to the GZ34.
I understood that R.G. found a different result.
Could you comment on your GZ34 model?

I'm a bit surprized by the odd line harmonics (though quite low value).
What's your explanation?

[Merlinb]

Could you comment on your GZ34 model?

Sure, it's just the one that came with TinaTI. Here is the I/V curve, with the datasheet curve superimposed on top. My model is a bit more conductive than the datasheet. But curvature is what counts.
EDIT: I took the datasheet curve from the 1952 sheet. Apparently it was up-rated in 1958, which is a close match for the spice model.

I was gonna put one on my curve tracer yesterday, until I discovered I only have a GZ30, not a GZ34, d'oh!

I'm a bit surprized by the odd line harmonics (though quite low value). What's your explanation?

I suspect they are a phantom artefact of the FFT calculation, since we're doing the maths on a digitally-sampled waveform of finite duration.

[Helmholtz]

I suspect they are a phantom artefact of the FFT calculation, since we're doing the maths on a digitally-sampled waveform of finite duration.

Makes sense, thank you.

[R.G.]

OK, this is making me nuts. I got some time to do more sim. I found out two things. First, there are a lot of slight differences in modeling vacuum rectifiers, but on the whole, they're going to be OK for all but the ... er.... finest reproductions. That matches my predilections and Merlin's comments as well. Second - hold up on the ripple spectra until I can sort out the differences in my earlier sims and what I'm getting now.

Meanwhile, I ran some new comparisons on rectifier models; two equations based models, the curve-matched diode-zener one, and an SS diode plus a resistor. They're all pretty similar.

The first is the "Leach" model, from Marshall Leach in 1995-96. I believe it's the model in TINA. It's the green trace. They are very close, even in the model statements, differing primarily in the scaling coefficient.
Red is the equation model I used first for the 5AR4. Blue is the diode-zener-resistor model I was using. It's really close to the red trace because I tweaked it to be that way.
Black is a 1N5408 plus a resistor.
They all look closer at lower currents. I was banging the diodes pretty hard in my original sims. These traces are pure I-V, no simulated loading.
Next trace is same models and colors, but with different current per division to see the finer differences at lower currents.




I decided to look at what additional resistance to emulate transformer winding resistance would do. In the traces below, red is the equation based model I first used; green is 1N5408 plus 57 ohms; blue is same as green, but with an added 10 ohms for the transformer winding; the lower red is 1N5408 plus a 100R transformer resistance and a 57 ohm resistor in series.



It seems like transformer resistance has a huge effect. This brings up the question - is it even worth modeling vacuum rectifiers? Maybe not.

Again, it's currently making me nuts.

[Merlinb]

It seems like transformer resistance has a huge effect. This brings up the question - is it even worth modeling vacuum rectifiers? Maybe not

I don't think so, especially the 5AR4 since it has such low internal resistance. The source resistance of a Marshall 50W power transformer is about 120 ohms for example, so recto resistance only accounts for about one third of the total PSU resistance. The curvature of the tube's resistance is only a fraction of that third.

[Helmholtz]

. The source resistance of a Marshall 50W power transformer is about 120 ohms for example, so recto resistance only accounts for about one third of the total PSU resistance.

Which means that almost 70% of the sag is coming from the PT alone.