|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
24.140.192.106
In Reply to: RE: critical L input filter: PSUD2 ... yes or no good? posted by Triode_Kingdom on October 11, 2016 at 20:03:23
One thing I noticed on spectral graphs from the RH test amp was some sidebands coming up at one octave intervals beyond 120Hz (240Hz & 480Hz). The test signal was @ 1KHz.
At the time, I had the 2 LC section flywheel filter installed.
I reviewed some older spectral graphs of a SE active parafeed circuit with the 2 LC section flywheel filter and didn't notice those sidebands coming up.
Follow Ups:
It's difficult to know for certain why that occurred. However, I can tell you that the use of multiple, small-value, Low-DCR chokes and smallish caps creates a lot of undesired responses in the output Z of a supply at low frequencies.
Here's an example (yours looks superb compared to this!):
--------------------------
Buy Chinese. Bury freedom.
TK,
I tried similar 2/3 section LC filters to finals with 320mH/600ma/10 ohm Triad C40X chokes and 50uf caps in the past.
As I recall, the result was not so good due to a mechanical buzzing.
FWIW, I don't think DF/JM ever used or now use those Triad chokes in their LSES filters. Not sure what they do, so I will not speculate.
I'll try running my 3 section flywheel in spice or maybe someone else can run it to see what happens.
Just to elaborate, back when the sole proponent of this nonsense was exclaiming its virtues in nearly every post on this forum, I simulated all the designs using the values he himself described. Every one I tested was absolute garbage, nothing I would ever use as a power supply for audio amplifiers. The problem has to do with output Z, something that particular person doesn't understand and lacks the skill to analyze.
Further on this subject, I've been analyzing power supply performance in some detail over the last few months. It does seem extraordinarily difficult to produce a truly smooth output response when inductance of almost any size or DCR is present in the design. The best that can be hoped for is to push the anomalies to frequencies below 20Hz, but then one has to deal with subsonic sensitivities that might be excited by sources other then the amplifier stage.
One of the best types of supplies I've tested is the HK voltage doubler. Output Z is low and very smooth over frequency, and the contribution this makes to dynamics in real world listening is apparent. Unfortunately, with no inductance to aid smoothing, these designs aren't suitable for SETs. So, the hunt for the perfect low-noise, low-Z supply continues. :)
--------------------------
Buy Chinese. Bury freedom.
I played with your critical L spice model a bit until I got close, then tried my current 3 stage flywheel. Not sure how correct this is but it looks really bad.
The two stage flywheel with lots of filtering at C2 looked a bit better.
dt 667
I'm not sure what resistive values you used for the components in the two-stage. Also, the input termination IMO needs to be larger. 50 Ohms isn't likely to accurately represent the resistive component of the primary and secondary winding of the PT, plus chassis wiring and house wiring. I've been using 150 Ohms, but even that might be optimistic.I wouldn't consider 300uf to be a practical value for the last cap. I've changed it to 100uF in the plot below. Note that Z under 10 Hz rises to a value that eventually equals the cumulative DCR of the series components. There's a resonant peak hidden in there (the one that's visible if the termination is changed to 10 Ohms), but it's pretty well swamped by the resistance.
PS filters also need to be tested for ripple content, of course, and for frequency response looking into the filter from the input side. The frequency response test helps to determine whether the filter will be susceptible to undesired frequencies entering from the line.
--------------------------
Buy Chinese. Bury freedom.
Edits: 10/12/16
I guess the question is... "What does it mean?"
In looking at DDT667's two peaks I'll agree they do not look pretty but how does that actually relate to the actual measured frequency response and the sonic character of the circuit?
I have seen the "ps resonant blips" in measured frequecy response plots but they only seem to show up outside of the passband. In the sim below I would expect to see the peak at 10hz show up as a dip in the rolloff of the low frequencies but assuming a full bandwidth amp, will the 60hz bump also show up?
dave
Variations in output Z don't generally correlate well with the filter's "forward" frequency response. That's why these filters need to be analyzed from both directions. What does it mean? Difficult to quantify precisely. Most amplifier designs are susceptible to modulation of their B+ with various symptoms; distortion, attenuation, degradation of dampening, etc. It's not clear to me how high PS output Z can become before such symptoms are audible, so I simply design supplies to be as low and flat over frequency as possible.
--------------------------
Buy Chinese. Bury freedom.
I put in the L resistive values , adjusted R1 to 150 and added ESR for caps C1 & C2 (just a guess at this point) so it would look closer to TK's graph.
L1 is the Triad C56U and L2 is one of drlowmu's designed chokes.
That looks much better, but I need to question whether the impedance rise below 100 Hz might be audible. It would be interesting to SPICE an entire SE amp, feed it with this supply, and look to see what's happening at both the output of the supply and output of the amp. Alternately, one could build this supply, drive an SE amplifier, and then probe the B+ line with a scope while delivering audio power to a load. The question is how much that low frequency rise will allow the B+ to wiggle around...
--------------------------
Buy Chinese. Bury freedom.
This is what I have as a filter with my RH84 project amp.
3 stage LC flywheel to finals.
Add an UF4007 isolation diode + 1K ohm R dropper + C4 20uF//200K bleeder (to driver stage) for the complete filter. The spice Z vs. Hz graph looks quite different.
As a full range amp, perhaps there might be some "psychoacoustic" effect as Chip647 would suggest, but who knows?
I did see some sidebands coming up an octave or two beyond the 120Hz ripple with a different flywheel filter on my spectrum/distortion graph.
Might not even hear anything unusual if you are using bass-shy bookshelf speakers or in a multi-amp system driving some horns, so who knows?.
Funny how Stu Hegeman had the right idea over 50 years ago when practical solid state diodes began to appear. And it still works just as well today. Few will argue the Citation tube amps were some of the best ever designed.So much for these "modern power supply" ideas!
Edits: 10/12/16
HK had had a great affinity for high current (low Z) power supplies for many years. They based their marketing theme on that in the '80s with the introduction of receivers like the HK 590i. To this day, that particular line remains in my memory as one of the best sounding SS amps ever made.
--------------------------
Buy Chinese. Bury freedom.
Hi,
The example that Triode Kingdom posted above, of a LSES or even a flywheel supply, was TOTAL TRASH. He picked values we NEVER use, maybe so he could " prove a point". But all it proved is personal intellectual dishonesty.
I just listened to a 2016 pair of DF amps, on GPA 604s in a MLTL, for three days solid at RMAF, and the amps had no low frequency " resonances " but rather, a GLORIOUS low end.
Also, you show a 3 HY 80 Ohm choke, and not a 3 HY 30 Ohm chokes as TK simulated. 80 Ohms of DCR in a choke positively sucks.
The very FIRST THING Mr. Fulton ever told me about amp design, ( for 2A3s in this case - 1982 ) was, that the chokes have to be 20 Ohms or less. Nothing has changed. Today, I use 10 Ohms or less, and this past weekend in Denver - RMAF, I was listening to LSES with 4 ohm Ls. Was glorious .
Your use of C2 and C3 at 50 uF is arbitrary, not well thought out, and I have showed you better C2 and C3 values, smoother settling. Maybe we need to discuss this privately some !! I suggest that.
Jeff Medwin
drlowmu,
I'll review your PSUD2 graphs and get back to you off forum.
My flywheel filter is about what I could build with the parts I have on hand.
Have a good day,
dt 667
IIRC, you said that was your day job. Do you work in a factory or at a distributor? And I'll bet those light bulbs today are LED or compact fluorescent?Now if at a factory, why not discuss your power supply design with some of your engineers at lunch? After all, an LED or CF lamp contains a switch mode power supply. That is an LC filter on the output.
All the math and theory to design an LED or CF power supply can be re-applied to a linear audio amp supply. And as SMPS design is far more complicated than brute force linear supplies, I'm sure your engineers have lots of superior knowledge on power supply inductor theory.
If I wanted to know more about choke input power supplies, an SMPS design engineer is probably the best source of knowledge you will find today.
Edits: 10/12/16 10/12/16
I'm working in an R&D lab with mechanical and chemical engineers for the last 3 years.
Mostly setting up and running life tests on various products made of composite plastics, rubber compounds and metals. Then I document test results and send data sets to the engineers for analysis.
The company outsources electrical engineering work to Germany.
One of the engineers and the other guys in the lab know I do the tube hobby, but they are clueless to help me out.
Except for the help I get from the forum or online, I am on my own to learn anything new.
Not always such an easy task.
z
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: