SY said:
Very funny. Balanced opamp circuits with psrr better than 60db throughout the audible range show similar sensitivity to rectification. How can a tube circuit be fixed?
Apparently it's possible to train your brain to completely ignore any raw data the ears provide as long as it's incompatible with basic engineering principles.
Mercury vapor rectfiers?????????
Is that done to lower the on-state voltage drop?
More RF / transmitter type stuff?
Is that done to lower the on-state voltage drop?
More RF / transmitter type stuff?
To take a more holistic view, we are running into the same difficulties here that are unavoidably encountered universally when specifics are not isolated/analysed/measured/whatever. It keeps an aura of mysticism and snake oil mincing in through the cracks.
Rdf's examples are valid - but why? Firstly (and I regrettably did not read Bruno's thread), to choose caps for an "optimal resistance" at pulse frequency components can be explained by the influence of its optimal R - I have had the same experience with a very wide band amplifier. But the problem was not that perfect power supplies were under suspicion; rather that my circuit (including layout) showed certain characteristics at r.f. frequencies. (With all respect to Bruno, and if I understand correctly, it is risky to rely on a capacitor's internal resistance to solve a problem. It is not a specified parameter and will differ from brand to brand.)
Then re regulation of a pre-amp supply. Define regulation. The "regulation" (meaning low impedance) of a set of RC stages is excellent as the frequency rises, which is the requirement here! It is only the d.c. regulation that is poor and which is not a prime requirement with class A stages.
And so on. I would venture to say that proper analysis will always yield an acceptable explanation - it has so far for me; maybe I was just singularly lucky (I know am not that clever!) And we can also never ignore that subjective observations have been shown to differ. At the danger of boring folks: I always think of the very good example, where HF News & RR years ago published the personal choices of some 6 seasoned reviewers (the likes of Martin Colloms, Ken Kessler, etc.) of hi-fi components. I recall that amplifier-wise, the "best" choice of some were not even on the short list of others! And these were people experienced in the detection of small differences!
My (bad) experiences regarding ss rectifiers are limited, but objections could mostly be identified by some artifact showing up in a spectrum analysis, and swamping caps or small series resistors (or sometimes changing the layout) could always cure the same. I would be the first to learn and apply any findings that would expand the above, but until a better solution comes along I will stick with these. I have read that folks do not like the 5AR4 - but then the GZ34 is used copiously. (In my data sheets they are equivalent.) I respect all subjective experiences - I was not there to disagree. But they will gain in prestige if supported by blind tests.
In the meanwhile, there is a lot of solution in Poohbah's advice, at least regarding (usually) simple audio amplifier supplies.
Regards.
Rdf's examples are valid - but why? Firstly (and I regrettably did not read Bruno's thread), to choose caps for an "optimal resistance" at pulse frequency components can be explained by the influence of its optimal R - I have had the same experience with a very wide band amplifier. But the problem was not that perfect power supplies were under suspicion; rather that my circuit (including layout) showed certain characteristics at r.f. frequencies. (With all respect to Bruno, and if I understand correctly, it is risky to rely on a capacitor's internal resistance to solve a problem. It is not a specified parameter and will differ from brand to brand.)
Then re regulation of a pre-amp supply. Define regulation. The "regulation" (meaning low impedance) of a set of RC stages is excellent as the frequency rises, which is the requirement here! It is only the d.c. regulation that is poor and which is not a prime requirement with class A stages.
And so on. I would venture to say that proper analysis will always yield an acceptable explanation - it has so far for me; maybe I was just singularly lucky (I know am not that clever!) And we can also never ignore that subjective observations have been shown to differ. At the danger of boring folks: I always think of the very good example, where HF News & RR years ago published the personal choices of some 6 seasoned reviewers (the likes of Martin Colloms, Ken Kessler, etc.) of hi-fi components. I recall that amplifier-wise, the "best" choice of some were not even on the short list of others! And these were people experienced in the detection of small differences!
My (bad) experiences regarding ss rectifiers are limited, but objections could mostly be identified by some artifact showing up in a spectrum analysis, and swamping caps or small series resistors (or sometimes changing the layout) could always cure the same. I would be the first to learn and apply any findings that would expand the above, but until a better solution comes along I will stick with these. I have read that folks do not like the 5AR4 - but then the GZ34 is used copiously. (In my data sheets they are equivalent.) I respect all subjective experiences - I was not there to disagree. But they will gain in prestige if supported by blind tests.
In the meanwhile, there is a lot of solution in Poohbah's advice, at least regarding (usually) simple audio amplifier supplies.
Regards.
rdf said:
Maybe this is ignorance speaking. Then inform me. But doesn't "RF hash" relate to the 100/120 hz ripple? And since the chokes take care of that, as far as I understand, shouldn't ss hash be a no-issue?
The large inductors used in PSU's can conduct a certain amount switching noise through via their interwinding capacitance. A little inductor or ferrite is all that's needed to kill it.
I guess I am surprised that I don't see common mode chokes and low impedance caps being used in the tail end of PSU's.
With all you spend to put together a tube rig... and the sand you wrap around the bottles... why not save a few bucks (and Joules) and use SS rectifiers?
I guess I am surprised that I don't see common mode chokes and low impedance caps being used in the tail end of PSU's.
With all you spend to put together a tube rig... and the sand you wrap around the bottles... why not save a few bucks (and Joules) and use SS rectifiers?
A number of us are posting simultaneously, so pardon the re-entry.
Sorry, analog_sa, but do you then support the subjectivist manifesto "that objective measurements are unimpoprtant compared with the subjective impression gained by listening tests. Where the two contradict the objective results may be dismissed." ?? (Quote from an article "science vs. Subjectivism in Audio Engineering" by Douglas Self).
If so, you ignore the wealth of acousto-medical research done over decades, proving ad nausiam that the ear is an extremely sensitive detecting sense but a poor and non-repeatable measuring device. You are also asking me to accept that 2+3 can be =8 if any human sense tells me so.
(But it is rumoured that mathematicians have discovered that 2+2 can be 5 - for very large values of 2.)
Regards.
analog_sa said:
Sorry, analog_sa, but do you then support the subjectivist manifesto "that objective measurements are unimpoprtant compared with the subjective impression gained by listening tests. Where the two contradict the objective results may be dismissed." ?? (Quote from an article "science vs. Subjectivism in Audio Engineering" by Douglas Self).
If so, you ignore the wealth of acousto-medical research done over decades, proving ad nausiam that the ear is an extremely sensitive detecting sense but a poor and non-repeatable measuring device. You are also asking me to accept that 2+3 can be =8 if any human sense tells me so.
(But it is rumoured that mathematicians have discovered that 2+2 can be 5 - for very large values of 2.)
Regards.
rdf said:
Nope. The Schottky effect is not based on minority carrier conduction. As such there are no minority carriers when the junction gets reversely biased and no current spike due to minority carriers being swept out of the junction. I guess 'zero recovery' is a fair description.
phn said:
Not related at all. We are talking many MHz here and chokes conduct at these frequencies as well as wire 🙂
An interesting experiment is to compare RF hash from a silicon diode and a mercury rectifier and try to figure out why one sounds pleasant and the other doesn't. Never seen hard data on mercury.
As said, this is beyond my competence and my previous post may be all wrong. But as I see it, either hash is the result of poor design or this quote is rubbish (it cannot be both ways):
"As can be plainly seen the only difference is the rectifier. OK, so there are a couple of capacitors and resistors too. But let me 'splain! The resistors are there to reduce high frequency transients and as current limiters. The capacitors across the diodes are to reduce the reverse current switching. SS diodes switch off fast. This creates a spike. The caps reduce the size of the spike. They are generally about 0.1 µF, 1000 volts or better. The voltage level is because when the AC wave on the other side of the diode goes negative, it adds to the positive voltage stored in the capacitor across the diode. So if the filter cap has 350 volts, then the diode can see upwards of over 700 volts across it. The filter choke takes care of it the rest of the way. In fact, the spike is of a far greater frequency than the 120 hertz ripple, so it does a much better job of removing the so-called solid state hash that so many complain about. So, if they merely use a choke, it will be gone and so shoud one of the biggest complaints about SS rectifiers. Hmmm (no, not hummmm). Let's prove this by math:
A 3 henry choke at 120 hertz has a reactance of about 2300 ohms. The same choke at 10 to 20 kilohertz, the region where hash likely is, will have a reactance of:
Xc=2PiFL=6.28x10 000x3=188400
A whopping 188 Kilohms! How the heck can hash get past that! So, a solid state rectifier can be as good if not better, due to its not dropping as much voltage hence not losing that extra power, than a tube rectifier."
"As can be plainly seen the only difference is the rectifier. OK, so there are a couple of capacitors and resistors too. But let me 'splain! The resistors are there to reduce high frequency transients and as current limiters. The capacitors across the diodes are to reduce the reverse current switching. SS diodes switch off fast. This creates a spike. The caps reduce the size of the spike. They are generally about 0.1 µF, 1000 volts or better. The voltage level is because when the AC wave on the other side of the diode goes negative, it adds to the positive voltage stored in the capacitor across the diode. So if the filter cap has 350 volts, then the diode can see upwards of over 700 volts across it. The filter choke takes care of it the rest of the way. In fact, the spike is of a far greater frequency than the 120 hertz ripple, so it does a much better job of removing the so-called solid state hash that so many complain about. So, if they merely use a choke, it will be gone and so shoud one of the biggest complaints about SS rectifiers. Hmmm (no, not hummmm). Let's prove this by math:
A 3 henry choke at 120 hertz has a reactance of about 2300 ohms. The same choke at 10 to 20 kilohertz, the region where hash likely is, will have a reactance of:
Xc=2PiFL=6.28x10 000x3=188400
A whopping 188 Kilohms! How the heck can hash get past that! So, a solid state rectifier can be as good if not better, due to its not dropping as much voltage hence not losing that extra power, than a tube rectifier."
Johan Potgieter said:
I like to leave more room for mysticism (or more charitably, 'unknowns') than many but essentially agree in most all particulars. No resolution without defining particulars. Incidentally, Bruno did exactly that. Specified a brand and value. In fairness it's probably not that sensitive a parameter, as long as R is greater than some value. I thought it an elegant engineering solution.
To take up the example of RC filtering, nothing you don't already know but even though regulation can be excellent at audio frequencies PS rejection above the audio band might not be if the normal audio engineering path of big electrolytics is followed. My current approach, possibly voodoo, has been to use the normal large inductors with enough capacitance for the circuit, the latter of a type with low impedance at the highest frequencies possible, (which usually means large poly-oil motor runs as the most expedient at the voltages required) and to follow poobah's advice (the horror! 🙂) of using high frequency micro-henry inductors follwed with small film caps to maximize rejection through the low MHz range. Spice simulations suggest the benefits are huge, if RF rejection is consdiered important, at low cost and almost zero loss in the audio band.
There's one important aspect of any PS discussion that hasn't been brought forward yet. When the audio classic textbooks were written power lines were used solely for power. Safe bet an electrical service in an urban area that isn't laden with RF junk from computers, home automation, industry, etc. is a rare find these days. Some countries have even approved broadband over power lines. The old answers, well engineered though they were, potentially target a set of conditions which no longer apply.
analog_sa said:
I understand but thought the immediate and sharp turn-off of a Schottky, though better for lacking a recovery 'bounce', was still more of an RF noise generator than soft recovery diode which didn't transition as hard. Again, more guesswork looking for clarification on my part.
phn said:
A 3 henry choke probably has a self-resonance below a 100 kHz, above which it becomes a poor capacitor. Most large, high voltage electrolytics still resonate well below that. PS rejection essentially bottoms out at that point and gets worse with increasing frequency until the next set of effects take over (lead capacitance to chassis, the inductance of traces and internal wiring, or whatever they may be.) I guess it all hinges on whether PS rejection in the huge band above audio is considered an important factor. For the minimal investment in parts and design I prefer to remain paranoid.
Good point rdf (the horror! 🙂 ),
Switching noise caused by diodes may be mute in consideration of noise already on the mains. Issues of diode spectra aside... we are not even dealing with sinewaves at our inputs anymore. And as legislation pushes for more PFC loads this comes at the expense of still more hash on the power lines.
I've been putting common-mode chokes and little ferrite beads on lotsa stuff lately... amazing how many fewer glitches I have to deal with.
😀
Switching noise caused by diodes may be mute in consideration of noise already on the mains. Issues of diode spectra aside... we are not even dealing with sinewaves at our inputs anymore. And as legislation pushes for more PFC loads this comes at the expense of still more hash on the power lines.
I've been putting common-mode chokes and little ferrite beads on lotsa stuff lately... amazing how many fewer glitches I have to deal with.
😀
phn said:
You don't mention the source of the quote, but yes, it is rubbish. A typical PS choke has very substantial parasitic capacitance (10nF? 20nF?). And why is the hash 'likely' at 10-20kHz?
Yeah... bad article... "rubbish", quoting analog_sa.
Best to learn from a teacher and not another student...
😀 😀
Best to learn from a teacher and not another student...
😀 😀
poobah said:
OK, horror should have been in "scare quotes" to match all that fur. If I recall your background is/was in switching power supplies. Any pointers? A few designs have been posted here employing common mode chokes in B+ and AC filament supplies, usually Kuei Yang Wang's. I like to put a ~hundred microhenry inductor / small film stage early in the supply, but that neglects to possibility of ground contamination. Hirsute suggestions greatly appreciated in advance.
Rdf,
The modern computer pollution is indeed a valid point often overlooked. I take care of that at the power input. Quite a filter network may be required in more hostile situations - where I live it is mercifully relatively "quiet".
I would also support more attention to the lowly h.t. line; this is a tube thread, and classic amplifier designs were not so kosher there because they did not need to at the time. One needs to watch with at least a scope and look way above the audio band on the amp signal points.
By the way, if I was not clear, my remarks about Bruno's solution was very conditional and general. If he indeed did the necessary specs, fine.
As an aside (but perhaps of value to some), your comments about where chokes self-resonate must also be kept in mind regarding output transformers. It is generally regarded that leakage reactance is the important parameter at high frequency, but very often it is rather the internal capacitance. I have measured a transformer with -3 dB frequency at over 160 KHz as a result of leakage only, but worth not much above 30 KHz as a result of interwinding/section capacitance - and the latter is often not even specified.
Regards (off for the night with me).
The modern computer pollution is indeed a valid point often overlooked. I take care of that at the power input. Quite a filter network may be required in more hostile situations - where I live it is mercifully relatively "quiet".
I would also support more attention to the lowly h.t. line; this is a tube thread, and classic amplifier designs were not so kosher there because they did not need to at the time. One needs to watch with at least a scope and look way above the audio band on the amp signal points.
By the way, if I was not clear, my remarks about Bruno's solution was very conditional and general. If he indeed did the necessary specs, fine.
As an aside (but perhaps of value to some), your comments about where chokes self-resonate must also be kept in mind regarding output transformers. It is generally regarded that leakage reactance is the important parameter at high frequency, but very often it is rather the internal capacitance. I have measured a transformer with -3 dB frequency at over 160 KHz as a result of leakage only, but worth not much above 30 KHz as a result of interwinding/section capacitance - and the latter is often not even specified.
Regards (off for the night with me).
Hey rdf,
I'm usually working on some sort of energy system... be it conversion, or monitoring. I've done some 350 kW 60 hz sine amps... class D if you will. The noise in those big boys is just horrendous. Keeping the junk out of the controls is always a pain.
This thread is good as an exercise in thought, keeping in mind that even a "perfect" theoretical diode model implies a noise spectrum of it's own... because of the sharp "corners" at turn-on and turn-off. The only time a perfect diode WOULDN'T generate some spectra is when the power supply is unloaded and the caps are charged at the peak. It stands to reason that a tube rectifier with a "soft" characteristic might indeed be the quiestest thing going.
But when you consider all the junk on the power lines, which should shoot through a tube rectifier as well as anything, it seems that part, if not all, of the advantage of a tube is lost.
You should kill 2 birds with 1 stone if you could. I think a common mode choke in the PSU would do wonders for line-borne noise and at the same time make the advantage of a tube rectifier needless. Maybe something like a L/C/common-mode-choke/split R/C type arrangement might work well. Maybe work well enough to make diode choice unimportant.
Of course you lose the slow start feature of the tube diode. But that's easy enough with a FET.
Thoughts?
I'm usually working on some sort of energy system... be it conversion, or monitoring. I've done some 350 kW 60 hz sine amps... class D if you will. The noise in those big boys is just horrendous. Keeping the junk out of the controls is always a pain.
This thread is good as an exercise in thought, keeping in mind that even a "perfect" theoretical diode model implies a noise spectrum of it's own... because of the sharp "corners" at turn-on and turn-off. The only time a perfect diode WOULDN'T generate some spectra is when the power supply is unloaded and the caps are charged at the peak. It stands to reason that a tube rectifier with a "soft" characteristic might indeed be the quiestest thing going.
But when you consider all the junk on the power lines, which should shoot through a tube rectifier as well as anything, it seems that part, if not all, of the advantage of a tube is lost.
You should kill 2 birds with 1 stone if you could. I think a common mode choke in the PSU would do wonders for line-borne noise and at the same time make the advantage of a tube rectifier needless. Maybe something like a L/C/common-mode-choke/split R/C type arrangement might work well. Maybe work well enough to make diode choice unimportant.
Of course you lose the slow start feature of the tube diode. But that's easy enough with a FET.
Thoughts?
Hi
Don't you think high on-resistance vacuum rectifiers have much influance on power compression characteristics? That's probably why some guiterists prefer their sound... hmmm?
regards
Adam
Don't you think high on-resistance vacuum rectifiers have much influance on power compression characteristics? That's probably why some guiterists prefer their sound... hmmm?
regards
Adam
- Status
- Not open for further replies.