|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
50.84.144.34
In Reply to: RE: Thanks & a Question posted by Eli Duttman on February 08, 2016 at 16:55:53
Those are all good ideas. I have to say though, if RF is really a concern, the mains need to pass through a balanced HF filter ahead of the power transformer. None of the three wires at the wall outlet are at RF ground.
--------------------------
Buy Chinese. Bury freedom.
Follow Ups:
T/K it's not RFI I'm specifically after. Large valued cap. I/P filters are associated with tiny conduction angles and very sharp, highly "triangular", ripple waveforms. Per Fourier's Theorem, that waveform contains high order harmonics of the ripple freq. fundamental. The big cap. generates "hash", simply as a consequence of its presence.The fact that the "hash" filter happens to suppress garbage riding on the AC mains is an added benefit. I want to stop trash, regardless of source, from sneaking into the reservoir cap., via the main filter choke's winding capacitance.
Eli D.
Edits: 02/09/16
I see, you're talking about "hash" from the diodes. The chokes I bought are for a different purpose, and I use them before the first cap. This creates a "prefilter" that's very effective at reducing PS ripple. The chokes are much larger than 1.5 mH, but not so large that they create a significant voltage drop. I suppose these chokes also do a good job of cleaning up diode noise, but I see no HF energy of any consequence even without them. That has led me to believe the main filter chokes in a CLC supply do a pretty good job, regardless of their winding capacitance. Maybe this issue isn't so much about circuit topology as it is parts quality. Just out of curiosity, have you ever tried to shield the diodes in the gear you've built or tested? I've been wondering lately if it might be beneficial for reducing noise that's radiated into grid circuits or conducted along ground paths.
--------------------------
Buy Chinese. Bury freedom.
Again, it's not the diodes. Schottky diodes are dead quiet and the "hash" will still occur, when a large valued cap. I/P filter is employed. The unwanted HF energy is a matter of the situation's Physics.
Eli D.
So, you're saying the capacitors create frequencies that aren't otherwise present at the output of the diodes? How could you know that if it only occurs when the diodes are connected to the caps? How do you know it's not the behavior of the diodes themselves?
--------------------------
Buy Chinese. Bury freedom.
Large valued I/P caps. allow the rectifier winding to conduct for only very brief portions of the complete sine wave. That's the tiny conduction angle. The smaller the conduction angle, the larger the charging pulse current and the "sharper" the ripple wave form. The kind of diodes used are irrelevant, in this context.
Any periodic function can be expressed in terms of a sum of sines and cosines of the function's fundamental frequency and the overtones. See Fourier's Theorem. The "sharper" the function's shape, the more important high order harmonics become in completing the sum. Follow the link provided below, where square waves are discussed and a Fourier Series reference is provided. The perfect Fourier Series for a square wave requires inclusion of (sic) infinity. Otherwise, the shoulders are "rounded".
Eli D.
If it's RF doesn't it need to be stopped?
How can it be filtered out?
I mean, it can be filtered out of the line but isn't it broadcasting?
Filtering it out of the line won't help that.
I think of power supplies with cap input filters as transformer, rectifier and cap abusers and also as RF transmitters.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Tre',
Both G_D and the Devil are in the details. A "McShaned" H/K Cit. 2 is a marvelous music reproducer. Adding the "hash" filter makes an improvement to the already very, very, good.
Obviously, a "hash" filter will not help with radiated energy. It does deal with RF crud riding on the B+ rail, regardless of said trash's origin.
With their small stray magnetic fields, toroidal power transformers seem to be an excellent match for cap. I/P filters. I deal with the unwanted, in the power trafo case, wide bandwidth of toroids by the use of ferrite beads on the primary side wires. The beads block garbage from entering or leaving the unit.
Eli D.
"Large valued I/P caps. allow the rectifier winding to conduct for only very brief portions of the complete sine wave. That's the tiny conduction angle. The smaller the conduction angle, the larger the charging pulse current and the "sharper" the ripple wave form."
Eli, I'd like to explore this a little more, simply because it seems counter-intuitive to me. In an effort to create the effect, I built a simple transformer-powered AC supply in LT Spice (below). What I discovered is that it acts in a manner opposite to what you're describing. Current pulses through the secondary of the transformer become wider when the shunt capacitor is made larger. In a further effort to reconcile this, I also analyzed voltage output of the power supply in FFT mode. The two FFT plots showed no significant difference in high frequency content when the capacitor was changed from 20 uF to 20,000 uF. Only ripple content differs (harmonics of 60 Hz), and it is pretty much all gone by 1 kHz in either case.
Is there a certain way I should modify this circuit to obtain the behavior you would expect?
--------------------------
Buy Chinese. Bury freedom.
1st, simulations are NOT perfect.
Dig out a copy of the 1st or 2nd edition of "Basic Electronics for Scientists" by J.J. Brophy. The text contains an exquisite description of cap. I/P filter behavior and what the ripple waveform looks like.
Without a draw, a cap. I/P filter charges up to the sine wave's peak voltage and the conduction angle drops to 0 o . When a draw is present, the voltage of the cap. drops, over time, and charging pulses begin. A larger cap. yields better regulation, because the draw impacts on the voltage available less. More energy is available to draw on. The charging pulses are of shorter duration. Big caps. reduce the ripple voltage, but increase the ripple current. The waveform of that increasing ripple current gets sharper and sharper, as the capacitance rises. Enter Fourier and his theorem.
Ask Jim McShane or Mike Samra if adding a "hash" filter to a refurbished H/K Cit 2., with its huge doubler stack caps., is beneficial in the background "blackness" dept. No, the difference is not earth shattering, but very good, to begin with, gets a little bit better.
Eli D.
"The waveform of that increasing ripple current gets sharper and sharper, as the capacitance rises. Enter Fourier and his theorem."
But the SPICE simulation indicates that such behavior does not occur. The transient analysis and FFT agree, and both say this logic is wrong. FWIW, I also ran this circuit with a much lighter load (200K Ohm), hoping to see what you're describing. The difference between large and small capacitors only seemed to become more pronounced, with the larger capacitor continuing to produce a wider current pulse in the transformer secondary.
Perhaps what's missing in the theory you're promoting is circuit resistance. Conduction angle notwithstanding, RC considerations would seem to dictate that the pulses become wider with more capacitance. The plots I posted certainly seem to bear this out. In any case, the voltage at the top of the capacitor does not contain more high frequency energy with the larger capacitor. Therefore, whatever benefit the hash filter brings to all of this seems unrelated to the use of large capacitors.
--------------------------
Buy Chinese. Bury freedom.
Please read Brophy's book. What you actually measure, at the bench, is not what software yields. JMO, people rely excessively on software simulations. The simulations get you to the stadium entrance, but they don't get you to the pitcher's mound. Good old cut and try will never be out of fashion.
The bottom line is that equipment employing large valued cap. I/P filters sounds better, when the "hash" filter is installed.
Eli D.
Eli, this discussion has nothing to do with bench testing VS software. I never said you were wrong about claims that amps with the filter sound better. What I disagree with is the statement that the undesired energy you believe to be responsible for this is created in the large capacitor. A mathematical analysis in SPICE - using the same theoretical principles you yourself repeatedly pointed out - appears to confirm my opinion. That means the filter is almost certainly cleaning up energy created somewhere else, and that the need to apply it only to amps with large caps is a misuse of this topology. The origin of the high frequency energy is an important distinction, and it raises the issue of whether a different technique, say a filter at the mains, might perform just as well. Also, once it's realized that we're not dealing with mythological ills within larger capacitors, other options regarding filtering and power supply design might open up.
--------------------------
Buy Chinese. Bury freedom.
TK, I'm a little confused.Is it not true that the larger the input cap value, each diode (in turn) is on for a shorter amount of time and thus the higher the current peak?
If true then the next question is whether or not that creates RF.
Thanks.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Edits: 02/10/16
A larger shunt capacitor will always create greater peak current in the diodes and transformer secondary. However, the larger cap does not always create a smaller pulse width. In addition to capacitor value, pulse width and shape are affected by load resistance, series resistance and the slope of the wave over the period of conduction. The latter is particularly important, because the speed with which the "inrush" of current can take place when a diode first conducts is strictly limited by the slope of the wave. In the end, when all the factors are accounted for, there is no fixed relationship between capacitor size and pulse width or shape, even if the load remains unchanged.
This is evident when the circuit is analyzed using SPICE. In the circuit above, the 20,000 uF shunt capacitor creates current pulse widths in the diodes of approximately 4.3 mS. Contrary to the theory promoted in this discussion, reducing the capacitor to 20 uF creates a smaller pulse width of about 2.8 mS. However, when the capacitor is made even smaller, say 2 uF, pulse width increases again to about 5.3 mS. Lack of a definitive relationship between capacitor size and high frequency content is further emphasized when FFT plots are created using these values. The power supply with a 20,000 uF capacitor simply does not contain increased high frequency energy.
Finally, there's the fact that at 20 kHz, a 20 uF shunt cap exhibits 0.4 ohms to ground, whereas a 20,000 uF cap presents 0.0004 ohms. In this context, the idea that the larger cap would pass more high frequency energy into the amplifier looks really silly.
--------------------------
Buy Chinese. Bury freedom.
"However, the larger cap does not always create a smaller pulse width. In addition to capacitor value, pulse width and shape are affected by load resistance, series resistance and the slope of the wave over the period of conduction. The latter is particularly important, because the speed with which the "inrush" of current can take place when a diode first conducts is strictly limited by the slope of the wave."
I can see that. Doesn't R1 have the largest affect?
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
I think R1, R2, and R4 have a similar effect. They're all in series with the current into the capacitor.
--------------------------
Buy Chinese. Bury freedom.
OK, so is the following true?Theoretically, in the absence of R, the larger the cap value the more HF/RF.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Edits: 02/11/16
"Theoretically, in the absence of R, the larger the cap value the more HF/RF."With no R in the circuit (except a fraction of an ohm at the 60 Hz source so SPICE will run, and the load resistor), increasing capacitance appears to consistently create narrower pulse widths in transformer and diode current. However, even then, HF energy does not increase at the output of the supply. FFT analysis indicates decreased HF energy with larger value capacitors. My hypothesis is that the reduced impedance from larger capacitors is much more effective at attenuating HF energy than the narrowing of the pulses is at generating it.
Also, note that the series resistance after (and in) the secondary of the transformer has to be absolutely zero for the conduction angle to consistently decrease. Even five or 10 ohms in the winding defeats the theory. We can't get even close to that in real world amplifiers.
--------------------------
Buy Chinese. Bury freedom.
Edits: 02/11/16
Thank you for that analysis.
The last thing I want to understand is whether or not the RF can be broadcast through the air to the circuit even though it's being attenuating on the line by the cap.
This is something Lynn Olsen has spoken of.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Tre', an FFT plot of diode current pulses does not reveal any RF. Energy even at 20 kHz is 70 dB below the 60 Hz spike. I really think this issue has been blown way out of proportion by audiophiles looking for a way to explain what they hear. With a 20,000 uF capacitor in the circuit I posted, current pulses through the diodes are still nearly 1mS wide. In addition, their corners aren't sharp, probably due to the slope I mentioned previously. This just isn't RF territory.
--------------------------
Buy Chinese. Bury freedom.
OK, I wonder what Lynn was getting at?
Scroll down to "Power Supplies and Noise Spectra"
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
If I were to re-title that section, I would call it "Deficiencies of Poorly Designed Power Supplies." In my opinion - and I say this with considerable respect for much of Lynn's work - this particular lecture borders on fear-mongering. Nearly every caveat and doomsday scenario is based on one small area of the piece, specifically the following:
"...both spikes [wide and narrow] yield a comb spectra going out to at least 100 kHz or more, depending on the residual inductance of the first power-supply capacitor."
Just to be clear, Lynn is saying that undesired energy of consequence will extend to 100 kHz even if the first capacitor exhibits zero inductance. That is categorically untrue. I'm sure it's possible to create a supply that performs so poorly in this regard, but modern parts and general audio design principles make it highly unlikely that anyone would. FWIW, I've also analyzed many power supplies, and I've done so with Lynn's favorite instrument - spectrum analyzers. You can believe that if any of the typical analog supplies behaved in this manner, I'd be blowing the horn on this topic myself.
--------------------------
Buy Chinese. Bury freedom.
TK, Thanks for all your time on this.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
What Eli said! The small second filter of the choke and cap makes a noticeable improvement. I know Mikey agrees too because we've talked about it on occasion.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: