See post #7.
Errr, yes, I did see that.
Ok. I've put some time into this response since Qrr is something I've studied and which is of interest to me, and possibly to others.
I do not, alas, have access to Linear Audio magazine, which appears to be print only, and I suspect most here do not such access. Can you summarize how Jone's findings different from Cornell-Dubilier's application notes, such as the one set forth below? Also see Jones's response, noted below and quoted from, which explains, I think, enough of his assumptions that I may properly respond.
To begin, for those trying to understand the background issues raised herein, I located a paper (one of the sources upon which I've previously relied) from Cornell-Dubilier showing placement of the snubber across the diode, and the calculations for the snubber capacitor and resistor:
http://www.cde.com/resources/technical-papers/design.pdf
Design of Snubbers For Power Circuits
By Rudy Severns
28 pages
It is an excellent tutorial and application note.
I am, again, hard pressed to see how the snubber at the transformer secondary can scrub the diode noise after (a) it has passed through the entirety of the supply and (b) been injected into the ground plane by the filter capacitors. The snubber cannot dissipate any Qrr charge until after the return current arrives at the transformer secondary. So we have an entire circuit to modulate using the diode noise as a carrier wave. This seems nonsensical from a noise-reduction standpoint. We try to kill oscillation and noise as soon as it arises, not further downstream after it has corrupted signal.
I did find a letter to the editor of Linear Audio which includes Jones' response (emphasis added):
https://linearaudio.net/sites/linearaudio.net/files/MvdG to MJ V5.pdf
\Mr van de Gevel asks pertinent questions. Why prevent ringing? My experience is that when a circuit rings, that ringing finds its way into other circuits, usually via interwinding capacitance in the mains transformer (a common-mode effect, as Mr van de Gevel points out). We can add filters to block unwanted frequencies, but the best cure is always to prevent interference at its source. The problem is the series resonant circuit made up of transformer leakage inductance, diode depletion capacitance, and loop resistance. If we can reduce this circuit’s Q to < 0.7, we can prevent ringing – making it unnecessary to worry about amplitude and large-signal effects. The traditional solution reduced Q (Q = 1/R sqrt(L/C)) by increasing C – adding a large capacitance across the diode’s depletion capacitance, and making it lossy by adding series resistance. But we could also correct transformer leakage inductance to a resistance, thus reducing L, and that is what the network across the transformer secondary does.
I had not considered the transformer/rectifier as a mixer, with rectifier switching causing double sideband suppressed carrier modulation to a capacitively coupled carrier from mains wiring acting as an aerial. However, the forward resistance of a semiconductor diode is very low, so I don’t see how a rattle capacitor (lovely term) would have reduced modulation. I can’t help feel that the modulation scenario is rather lossy, requiring large signals to be present in the mains wiring, whereas mains transformer primary/secondary capacitance could easily couple a rectifier’s bursts of ringing onto the mains wiring aerial, allowing a nearby radio to demodulate it as hum simply because of the repetition rate.
This reveals Jones' assumptions that
(a) Resonance in the transformer is of primary importance
(b) Effects of modulation/broadcast are minimal.Would you agree on this summary of Jones' position?
While I fully agree with Jones that transforming ringing (a separate topic) is an important issue (I specifically previously noted it as the topic for a separate discussion) I suggest that focusing on this as the sole ringing issue overlooks the importance of the modulation/broadcast effect as well as the return-path distance and the ground plane noise injection. Two snubbers are needed: one for the transformer, and one for the diodes.
I have not yet built one of the Quasimodo transformer ringers, but expect it will demonstrate an opportunity to substantially improve all of my power supplies.
Here's what I previously wrote on the subject of detecting transformer ringing about seven months ago:
###
http://audiokarma.org/forums/index.php?threads/power-supply-ringing./
Mark Johnson over at DIYaudio has designed two automated transformer ringing detectors which allow snubbers to be developed:
Quasimodo and
Cheapomodo.
No math, no sliderule, no theory, just connect to the transformer then tweak as one watches the scope until the ringing vanishes. This is the way to eliminate ringing, instead of laboriously solving equations which tend to not work in practice.
Quasimodo is the fancy one (no kits left):
Cheapomodo is the latest incarnation. Not as fancy, but it should more than do the trick. Description:
Some Cheapomodo kits remain available for $13 delivered:
Disclaimer: I have no connection to MJ except as a highly satisfied customer. YMMV.
###
Some final thoughts and then I think this is done. I'm done, at any rate. Going back to finishing up my writeup about capacitance reduction in BaTiO3 in the presence of DC bias.
Linear Audio is sold by single issues, so the cost isn't onerous to access specific articles of interest.
I think the Jones comments I quoted set forth his position without the need to purchase the paper. Did he say anything about RF? From his response to the Letter to the Editor I'd suspect not.
From what I can see, Jones was not looking for Qrr noise and he doesn't address the RF broadcast issue. This is ONE paper, by the way. ONE PAPER by an author who admittedly did not consider the RF issue.
I've seen several dozen papers and application notes on snubber design and the suppression of noise broadcast, because I stopped looking after that point. Now, this issue may be confined to SMPS, but given a tube amplifier's tendency to receive broadcast signal (just look at what happens at the gain stage receiving AM stations) it stands to reason this is a problem which ought to be resolved.
I had an opportunity to test Jones' conclusions when I identified rectifier buzz in my Parsimon amplifier project. After adding the recommended power transformer secondary snubber, that noise was audibly and measurably gone.
First, I believe you observed a substantial reduction in noise. That's the reality of snubbing noise and ringing. I totally and absolutely believe in snubbers, particularly for transformer ringing. That's why I've relayed the information about Quasimodo and Cheapomodo. I also believe in them for diodes for the aforementioned reasons .
But, for the reasons I previously noted about (a) the long path to eliminate the noise (through the entire power supply and amplifier) and (b) the pollution of the ground plane, it is clearly that a transformer snubber cannot remove all (or even much) of the diode noise.
Again, inaudible noise and inaudible oscillation can consume power by amplifying out-of-band (i.e. inaudible) signals. So the Qrr noise still exists, and it spreads everywhere.
Just because we can't hear it doesn't mean that those noise gremlins aren't mucking about in the amplifier. If you go looking for Qrr you can find it.
Here's an easy way to verify that a snubber is working or not working (or not even present): tune a battery-powered radio to the low end of the AM band where no station exists and place it adjacent to the rectifier. If you hear buzz, that's Qrr singing to you. Add the snubber and the noise goes away. Voila! Is this noise/ringing in the audio band? No, but it's using amplifier power to broadcast and that power comes from somewhere.
I will agree that RF emission is likely still present and could conceivably cause audio noise issues in susceptible equipment, but I have no direct experience to confirm that. Solid-state amps would probably be far more susceptible than tube amps if the problem exists at all.
Given the efforts to suppress that noise, particularly in SMPS (a different item, true) I believe it is very real. That is why I devoted time to studying the issue.
Tube amplifiers will receive such noise. Solid-state will happily oscillate, as has been demonstrated in numerous amplifiers. I own a few that require some capacitors to stop this. I can't say which would be worse, as this depends on the design. I'm leaning towards tubes being worse as a class, since the power supplies are simpler and the tubes will happily oscillate at HF, but it's going to be idiosyncratic.
Absent audible buzz, we have to imagine exotic RF intermodulation scenarios in order to justify the cost and effort involved in completely suppressing recovery spikes at the source. To my mind, the use of tube rectifiers falls into that category --- not because of the expense, but because of inferior voltage regulation.
Consider an amplifier which oscillates at 500 kHz. Is this audible? No, of course not. (And it likely will not blow up piezo tweeters, either.) Does it reduce the amplifier's power in the audible range? Yes. Does it cause distortion in the audible range that we can hear? It depends on the diode, the amount of noise, and what the rest of the power supply looks like.
I'd rather use lower-Qrr diodes and add the snubbers to the diodes to remove the noise at the source, than try to do it all with one snubber across the secondary.
I think you're doing yourself an unfortunate disservice with that conclusion, and by not reading Jones' article. I quote from his concluding remarks: "We have been chasing a demon that doesn't exist, and then made matters worse by having the wrong parts in the wrong place --- and too many of them."
I will ask around and see if i can scrounge up a copy.
I will ask the question: how could adding an RC snubber to each diode be making "
matters worse". How could a snubber which removes noise make the power supply worse? This might be unnecessary cost, but it does not degrade performance. You're an expert at this sort of thing, so please explain this to me.
(I'm not being snarky, BTW, this is a direct statement of inquiry. I am trying to understand what harm can come from adding snubbers because I have never seen any potential detriment listed. Some things, like the placement of capacitors, wires, and beads on a circuit board matter a great deal, as what appears to be trivial changes in placement have profound effects. But a filter on a diode which is only for high-frequency spikes?)
I do agree that ignoring transformer ringing is a substantial mistake to the detriment of distortion and noise, and focusing solely on diode Qrr instead of transformer ringing would be a misapplication of parts. Yes, that would not be disputable. But to do both? How could that be worse? The cost is minimal.
As the counter argument, I point to the paper I cited by CDE. I think the substantial analysis set forth truly explains the issue and solutions. CDE is hardly a disreputable source of information and as one of the world's largest manufacturers of capacitors I would think the information would be accurate and credible.
The noise charts for diodes of various types, as I set forth, are easily reviewed for correctness. The application notes from the big manufacturers of diodes, capacitors, and SMPS ICs, all address the issue. All have specifically stated that snubbers are necessary, even for conventional linear supplies, particularly ones for low-noise applications.
I again note that Jones' response to the letter to the editor appears to specifically disclaims searching for RF.
The fact the noise must circulate through the entirety of the amplifier before returning to the transformer secondary is indisputable. The question is whether or not this noise is significant. I don't know that I'd rely on Jones (without having reviewed his data) when so many application notes exist to the contrary.
I'll see if i can locate the Linear Audio paper.
I went to review my writeup on the subject and collection of application notes, and discovered I had a copy of an article on diode noise by John Camille which appeared in Sound Practices in . It confirms the design caveats and application notes from CDE and the other major manufacturers. I will attempt to attach the article as it was difficult to locate.
He specifically addresses diode noise:
Development of a 211 Amplifier, Part 3: Reducing Diode Noise
by John Camille (Chimera Labs)
Sound Practices (, Fall)
Done properly, silicon diode supplies can be built that are quieter than untreated vacuum diode designs. One must remember that vacuum diodes also generate a significant amount of white noise that should be corralled in better designs.
Noise reduction for either vacuum or silicon diode rectifiers is a worthwhile undertaking. When silicon diode noise is controlled, the reliability factor and the virtually limitless lifespan of solid state diodes in properly designed supplies points toward the choice of silicon rectifier devices over vacuum tubes.
The primary culprit responsible for the noise generated by good quality silicon junction diodes is the turn-off characteristic. A reverse pulse is generated by the minority carriers crossing the junction after the majority carriers have galloped through. Tremendous strides have been made recently in reducing this effect in diodes designed for use in switched mode power supplies (SMPS). The processes used to create these fast turn-off devices avoid many of the noise and oscillation problems of older diodes.
However, reverse recovery pulses still exist. The energy distribution as a function of time varies with each device but the general trend is down at a rapid rate, as semiconductor designers seek to meet requirements for more efficient SMPS designs.
I have been doing empirical work with the simple-minded idea that the very fast fall time pulse excites the LC resonant circuit presented by the secondary winding of the transformer. The excitation of the LC circuit produces a damped wave burst of RF energy centered on the resonant frequency of the transformer. I have measured the burst frequency fundamental on different transformers at frequencies between 6 kHz and 165 kHz. Of course, the oscillation frequency (f0) is transformer and installation specific.
What all this means to the experimenter is that there are one or more transmitters buried in your amplifier. These transmitters produce 120 harmonically-rich pulses each second with a fundamental frequency (Q for each diode rectified supply. These pulses are radiated and conducted to other parts of the amplifier where they are detected and amplified along with the desired signal. Those beautiful wiring harnesses of old are real sonic killers for this reason. What you get is "diode grunge" that rides on the audio signal.
A more insidious problem is that these diode created bursts are also coupled back into the AC mains where they can affect unprotected low level stages elsewhere in the system.
...
Afterwards, I rationalized the cure, thinking that Schottky diodes have relatively few minority carriers, thus they provide little kick to the LC resonant circuit formed by the secondary winding. Since then, I have routinely replaced all of the pn diodes with Schottky diodes when I rebuild and recalibrate instruments for my shop.
I discovered the same pn burst problem during early work on the 211 amplifier. The attitude at the time was, "If I could see an artifact on the scope, it would be audible". In went the Schottkys on all low voltage and bias supplies. The high voltage supply for the 211 was another problem, however. A suitable bridge for the V power supply would require around a hundred 90V devices in the stackl Enter brute force techniques ...
...
The transformer-rectifier-filter interface must be short and sweet! I am wary of leads (antennas) over one inch long. My "new construction" supplies are fully shielded per a future article in SP. VHF RF construction techniques will make even the quietest amplifier quieter and sweeter! If you must bundle wires, use triaxial coax with proper grounding techniques. Think RF!
-
211_Amplifier_Part_3_Diode_Noise_by_Camille_Sound_Practices___Fall.pdf
Empasis added to this:
The author never measured any pulses, or oscillations RF or otherwise anywhere in the amp he only cited their existance, and only at the diode or the transformer. He didn't bother to note the size of any "artifact" ("If I can see it on the scope, I can hear it.") and didn't report finding them at the critical points in the circuit like the driver grids, filtered B+, and B++, output plates, output secondaries. He didn't bother to isolate the noise under conditions of no input signal. But he did go on to the scare talk about transmitters buried in amps. This paper is pure ...._____ (fill in the blank with anything you like) I say gibberish.
You're calling "pure gibberish" on John Camille? For reals? Do you now who he was, what he published, and what he built? Or the esteem in which he was held by the DIY community?
I suppose you'll also call "pure bullshit" on Hagerman:
Jones, in the Letter to the Editor I above cited, quoted, and linked to, clearly stated that he did not consider RF mixing as a possible factor.:
I had not considered the transformer/rectifier as a mixer, with rectifier switching causing double sideband suppressed carrier modulation to a capacitively coupled carrier from mains wiring acting as an aerial. That means he missed out on exactly what I described, and what was described the sources I cited.
Modern diodes have very low Qrr to the point that there is less to snub, but it still exists. The Jones approach reduces the ringing of the transformer and possibly the diodes after the noise has been circulated through the entire amplifier.
I was taught—notably by old-school engineers who built incredible analog circuitry including using Philbrick tube opamps and Burr-Brown opamp modules—to reduce or eliminate noise at the source, instead of using filters downstream, and to use less noisy components wherever possible. My prior post about the differences in Qrr for various diodes accurately and adequately explains that.
I have set forth links to FOUR primary sources (CDE, Camille, Hagerman and Jones' own writing!) to explain the engineering and justify my position . Anyone may read them. You have set forth exactly what? A critique of a short article in an audio journal which was not intended to be fully dispositive of the issue.
I have seen no published work to the contrary only a reference to a paper whose author later explained he did not address the phenomenon being discussed. If I can find the paper I'm sure I'll find it confirms what the author subsequently wrote about his own work.
I've set forth enough information for anyone to draw their own conclusion.
Further discussion is re-circulating and pointless.
I herein include a FIFTH original source detailing and explaining how diode Qrr stimulates transformer ringing. (It appears I saved these files in several different places and am discovering more data.)
See the attached paper from . Note that Johnson addresses the Jones technique of adding a snubber at the transformer secondary as, "
We’ve also learned how to overdamp the secondary to absolutely prevent it from ringing, even when stimulated." This explains what Jones observed. Johnson provides the explanation: "
The answer is: rectifier diode turn-off."
Your attention is called to Johnson's suggestion that Hex FRED diodes be used for low-noise performance to minimize this transforming ringing via diode excitation.
At this point I think the issue has been proven, via actual science and engineering, to be both real and easily resolved. Jones' approach to transformer damping is a good one, but insufficient to fully resolve the issue as it fails to address either broadcast or diode stimulation of the transformer.
Note: Italics and bold in original.
Simple, No-Math Transformer Snubber using “Quasimodo” Test Jig
by Mark Johnson
7 Sept,
The bell-ringer (stimulus) that causes oscillation in a real supply: diode turn-off.
Thus far we’ve simply observed that the power transformer’s secondary
can exhibit oscillatory ringing,
if stimulated. We’ve also learned how to overdamp the secondary to absolutely prevent it from ringing, even when stimulated. But until now, we haven’t discussed what the stimulus might be. The answer is: rectifier diode turn-off.
The transformer secondary’s sinusoidal output voltage is applied to the rectifier(s) + filter capacitor(s). When the secondary voltage exceeds Vcap + Vdiode, the rectifier turns on and the transformer recharges the filter capacitor. After the sinewave crests and the secondary voltage begins to fall, eventually it falls below (Vcap + Vdiode) and the rectifier turns off. Unfortunately, some rectifier diodes have uncontrolled or poorly-controlled turn-off characteristics, and they turn off extremely rapidly. Particularly troublesome are rectifiers that turn off “abruptly”, i.e., with very large dI/dt.
Abrupt diode turn-off with extremely large dI/dt is the stimulus that causes transformer secondary circuits to begin oscillatory ringing. Large dI/dt immediately manifests as a large voltage across the leakage inductance (recall: V = L dI/dt), and this large voltage-step is the start of the oscillatory ringing waveform.
“Aha!” people have exclaimed, “then let’s find and use rectifier diodes whose turn-off is guaranteed to have small dI/dt! Our transformer bell will never ring, because it will never be struck!” Such rectifier diodes do exist, and are guaranteed by their manufacturer’s data sheets to have small dI/dt at turn-off. They are called “Soft Recovery” diodes, and the newest models have a datasheet specification named “softness factor” (Tb / Ta) which quantifies just how low their dI/dt actually is. Beware, soft recovery diodes often have very large forward voltage drop -- sometimes higher than 2.5 volts -- so calculate your supply voltage headroom (margin) very carefully. Also calculate your rectifier power dissipation; you may need a significantly larger heatsink for soft recovery diodes than for standard (or Schottky) diodes with low Vfwd.
Philosophically, I prefer to employ a belt-and-suspenders approach: prevent transformer secondary ringing, two different ways. First, snub (overdamp) the secondary so it cannot possibly ring, even if stimulated. Second, use soft recovery rectifiers so the secondary cannot possibly be stimulated. Especially in a DIY piece of equipment where the cost of a CRC snubber is completely negligible, I think it makes no sense at all to ever omit snubbers. Fairchild “Stealth” soft recovery rectifiers are slightly more expensive than garden variety silicon bridge rectifier assemblies, and Vishay “Hex FRED” diodes are yet more expensive. But even these are only $2.28 in quantity ten; to purchase 8 of them for a pair of fully independent bridge rectifiers on the rails of a DIY power amp, would cost only $18.25. Less than two admission tickets to the cinema. DIY project budgets can afford this.
In the competitive arena of commercial products designed to a price point, I think snubbers (made of low-tech passive components) are significantly cheaper to source and implement, than new, modern, soft recovery rectifiers. So if it were my task to build and sell power amps at less than $2.00 per RMS watt, I’d probably put the snubbers in and leave the soft recovery diodes out.
-
Simple,_No-Math_Transformer_Snubber_using_Quasimodo_Test_Jig_by_Johnson_,_Sept_7.pdf
All Comments ( 0 )