Yeah, going SS diodes makes life easy humwise, but what to do when you want to have thyratrons.. choke input is must. I know i shot myself in foot, but couldnot resistIt is not the same thing. High L and low C will give high output impedance. High C and low L will give lower impedance, looking from the amp side of the supply.
At this respect, Mr Langford Smith in his "Radiotron" says, at page 1192 of chapter 31:
Please note that it is intended to be a good engineered device, and not necessarily a good "sounding" set, as "audiophiles" say.
Thyratrons in an audio amplifier??? May be, but such a power you want from it‼
I have 6 2D21 small size thyratrons which some time I want to do a delayed relay with them.
I have 6 2D21 small size thyratrons which some time I want to do a delayed relay with them.
I'm almost totally out of the audio game these days, and just stopping by randomly on account of Christmas vacation.
I invented the term "flywheel." Historically, it's a misnomer, based on a misconception early on in my analysis of the concept, but the name stuck.
However, I did not invent the idea of a very small-L choke-input filter. This is something drlowmu came up with. What got me looking at this idea, back around 2005, was the curious claim that with such a filter, there was a value of first capacitor (around 10 uF) that maximized supply output voltage. That seemed counter-intuitive to me, so I investigated using sims.
I found that the phase shift caused by the choke had the effect of extending the duration of the charging current pulse into the first capacitor beyond the peak of the raw AC waveform. The higher the DC current drawn from the supply, the greater this effect. The net result was a considerable reduction in DC supply resistance, perhaps by a factor of two, than would be achieved without the small choke. I demonstrated this both in simulations and in actual circuits.
With this circuit, it is possible to build what you might call a quasi-cap input filter with low DC output resistance, low peak diode currents, and overall low reactance compared to either conventional cap- or choke-input designs.
I documented my findings extensively on AudioAsylum, though I doubt the supporting figures are still online.
Due to my loss of interest in building audio gear, I have never tried this kind of supply in an actual amp, though I have the parts sitting in a closet.
This is in no way an endorsement of drlowmu or his designs, methods, and beliefs. In fact, I quit posting on AudioAsylum years ago, and maybe lost interest in audio in general, largely because I was so disgusted by people like drlowmu.
Still, understanding how the "flywheel" works was one of the most satisfying technical insights I've had in audio, and I daresay just a little bit clever. I am by no means advocating that people should build such supplies, but it would please me if people would understand what the "flywheel effect" is, because I think it's interesting and possibly useful.
Edit: I should add that once upon a time, when I was young and not senile, I went to a university on the west coast that currently has a <5% acceptance rate, and got a piece of paper from it that said "MSEE."
-Henry
Last edited:
Member
Joined 2009
Paid Member
I remember reading those posts on AA with great interest. Like all forums, it attracts some rudeness. At least here there are people with great power to keep people like me under control 😀
I assumed it was something simple but I too never built it and also have parts in my hobby room to suit. I assume that as a diode will only conduct the way we want it to, when forward biassed that the choke acts by developing a reverse-voltage across itself so that the diode may continue to conduct after the peak has started to decline. All this would require from the choke is perhaps a little bit of oscillation ? - your term 'flywheel' seems apt to me. Unfortunately, I haven't given it much study, despite my own 'piece of paper' from a good place there remains some mystery for me in coils of wire...
I remember seeing some gawd awful waveforms on the first choke in simulations. You may need a choke that will take some punishment.
I assumed it was something simple but I too never built it and also have parts in my hobby room to suit. I assume that as a diode will only conduct the way we want it to, when forward biassed that the choke acts by developing a reverse-voltage across itself so that the diode may continue to conduct after the peak has started to decline. All this would require from the choke is perhaps a little bit of oscillation ? - your term 'flywheel' seems apt to me. Unfortunately, I haven't given it much study, despite my own 'piece of paper' from a good place there remains some mystery for me in coils of wire...
I remember seeing some gawd awful waveforms on the first choke in simulations. You may need a choke that will take some punishment.
Last edited:
Turns out I still have some pictures online. The first one is a "macro" view with a 35mH input choke. The red trace is the voltage on the first cap, and the green is the voltage on the second. There is a current step at 2.0 seconds. You can see that when the current increases, the peak voltage on C1 actually increases as well. This is the "flywheel" effect.
Compare to a conventional cap-input filter where there is no such peaking effect:
The next picture is a "micro" view that shows how the input choke changes the timing relationships between AC secondary voltage and input cap current/voltage. In this case the input choke is 320mH which is not in the typical flywheel range, but the picture serves to make the point. Notice that the charging current (red) is delayed beyond the peak of the secondary waveform, and the input cap voltage (green) continues to rise as well and actually exceeds the peak secondary voltage:
This would work much better with the 35mH/10uF first section.
Finally, the same view without the input choke, showing the expected voltage/current relationships of a conventional cap-input filter:
The diodes shut off right at the peak of the secondary waveform, then the voltage on C1 goes down from there until the start of the next cycle.
This concludes your once-every-five-years refresher course on the "flywheel" filter, haha.
Edit: WRT those awful voltages across L1, PSUD sometimes spits out really nasty data but this seems to be an artifact of the program, not real life. I did once build up a breadboard flywheel supply and posted some data tables and scope images. Those pictures are long gone, but I can confirm the circuit worked as expected and there was no bad behavior.
-Henry
An externally hosted image should be here but it was not working when we last tested it.
Compare to a conventional cap-input filter where there is no such peaking effect:
An externally hosted image should be here but it was not working when we last tested it.
The next picture is a "micro" view that shows how the input choke changes the timing relationships between AC secondary voltage and input cap current/voltage. In this case the input choke is 320mH which is not in the typical flywheel range, but the picture serves to make the point. Notice that the charging current (red) is delayed beyond the peak of the secondary waveform, and the input cap voltage (green) continues to rise as well and actually exceeds the peak secondary voltage:
An externally hosted image should be here but it was not working when we last tested it.
This would work much better with the 35mH/10uF first section.
Finally, the same view without the input choke, showing the expected voltage/current relationships of a conventional cap-input filter:
An externally hosted image should be here but it was not working when we last tested it.
The diodes shut off right at the peak of the secondary waveform, then the voltage on C1 goes down from there until the start of the next cycle.
This concludes your once-every-five-years refresher course on the "flywheel" filter, haha.
Edit: WRT those awful voltages across L1, PSUD sometimes spits out really nasty data but this seems to be an artifact of the program, not real life. I did once build up a breadboard flywheel supply and posted some data tables and scope images. Those pictures are long gone, but I can confirm the circuit worked as expected and there was no bad behavior.
-Henry
Last edited:
Very interesting. I have not come across this before.
However, it seems to me that although this circuit trick improves a 'low value reservoir cap' supply it still will have higher output impedance and lower output voltage than a larger value reservoir cap supply i.e. conventional cap input. It is a useful trick if you are limited to low value caps, and might make better use of a vacuum rectifier. So a useful trick in the 1940s, but less relevant today?
I am not knocking it, as I am always fascinated by new counter-intuitive ideas. New to me, that is, as I suspect it would have been known to engineers 70-80 years ago even if they didn't document it.
However, it seems to me that although this circuit trick improves a 'low value reservoir cap' supply it still will have higher output impedance and lower output voltage than a larger value reservoir cap supply i.e. conventional cap input. It is a useful trick if you are limited to low value caps, and might make better use of a vacuum rectifier. So a useful trick in the 1940s, but less relevant today?
I am not knocking it, as I am always fascinated by new counter-intuitive ideas. New to me, that is, as I suspect it would have been known to engineers 70-80 years ago even if they didn't document it.
Member
Joined 2009
Paid Member
If it could be optimized so that small value caps enabled the use of all non-electrolytics we could have a benefit with long term reliability. The diode pulses maybe smoother too, so possibly less noise.
Last edited:
Its fairly simple to implement and since the 1st cap is not large, it can be decent quality film. I use a inexpensive, excellent quality polypropylene DC link for the 1st cap. Subsequent filtering can be done as the designer prefers.. 😉
The upside is that the B+ should not drift over time like so many supplies with electrolytic cap input might do. The downside is the added expense and space requirements of a small choke.
Ian
Last edited:
I just finished replacing all the electrolytics in an Audio-Innovations 2a3 PP amplifier for a friend. It was tedious job, but a complete re-design would be even more effort. I estimate that a re-cap of this amplifier (circa 1990) should have been done at least 10 years ago, and he was running it under sub-optimal conditions for far too long. Electrolytics do periodically need replacing.
Ian
Ian
If I understand it correctly, Morgan Jones-Valve Amplifiers-Fourth Edition, at page 358 covers this issue.
Given the cost and size of film caps and chokes I can see no reason to replace a reservoir electrolytic with a choke and film cap. The only advantage is longer life, at much greater expense. OK if you are building something which must work for decades without repair (e.g. an undersea telephone repeater) but a poor engineering choice for domestic audio IMHO.
Yeah, the weird thing is, with the proper L1/C1 values, the output voltage and impedance are comparable to what you get with a conventional cap-input filter. What is surprising about the circuit is that performance is optimized at a particular value of C1 (typically close to 10uF) and actually gets worse as the value goes up.
Apparently, DC current draw through the filter stores some energy in the input choke. The higher the current, the more the energy. When the diodes switch on, this energy flows into the input cap causing it to charge to a higher voltage than the peak secondary winding voltage. The "flywheel" effect intensifies as DC current goes up. This offsets the drop in the average voltage on C1 that you see in a conventional filter, stiffening the supply. Increasing the L1/C1 time constant slows down or suppresses the flywheel effect.
I, too, would be surprised if no had ever noticed this effect before. I would guess that switching power supply designers would recognize it. It's a nonlinear circuit and I think not easily understood in terms of the basic math you get in undergraduate EE textbooks, therefore not taught. When I was studying EE in the eighties, we weren't even told about choke-input filters. The flywheel input choke, though a low inductance, has to carry a lot of current so is physically large. For most applications a conventional filter will be smaller and cheaper, hence there's little reason for anyone to use this circuit nowadays.
The "flywheel" works bests with low DC resistance in the secondary circuit. This means solid-state diodes and a low-resistance choke. Low DC resistance goes straight the bottom line improving the supply output impedance. Because the total reactance in L1/C1 is small, you get the benefit of low DCR without the associated ringing you would see with an ultra-low DCR choke in the 10H range. You then put on a modest L2/C2 section and you end up with a decent-performing supply with a relatively fast recovery time compared to a regular choke-input filter, and lower peak diode currents compared to a regular cap-input filter. Possibly this could be a good thing. Dunno.
-Henry
Interesting. I guess I'll have to buy the book just for that. I have the older edition.
-Henry
Member
Joined 2009
Paid Member
somebody should simulate this for a solid state amplifier where dc currents are typically much higher.
Thanks. That's the standard choke-input filter analysis and doesn't really apply to the "flywheel" which is not at all a choke-input filter in the traditional sense. The flywheel is probably more correctly labeled a modified cap-input filter.
-Henry
It would work the same but the choke would be larger than the rest of the amplifier. Also, for better or for worse, solid-state amps seem to exist in a parallel universe where the laws of subjectivity that apply to tube amps are irrelevant. I don't know whether that observation should be taken as ironic or not.
-Henry
Argument 1: Electrolytics last decades anyway so why bother?
Remember the Jadis? I was replacing PSU caps in those horribly designed high end amplifiers in less than 5 years after they came out. And that was when electrolytic capacitors were reasonably well made.
Today we have so many fake/knock-off electrolytics from china. An order of Elna caps I had made with a large, well regarded parts distributer for 220uF Elna capacitors arrived last month and not a single one of them was even 200uF. The whole lot was obviously fake/knock-off. The parts distributer just shrugged and offered a return for different (likely faked) parts.
Right now I have little trust in any kind of "standard" electrolytic capacitor. I'm not the only one who has noticed this decline either. Yes, there are high end boutique electrolytics which are often ridiculously overpriced. I have had no choice but to use them periodically in repairs since the alternative is a complete re-design and re-build. For a new build, Its worth investigating alternatives.
Argument 2: The film caps and choke costs too much.
Due to ebikes and ecars, we now have superb quality DC link caps for very reasonable prices. I shopped around online and spent between $2 and $6 apiece for top quality Vishay Films. I have not seen these knocked off yet...
Also, the chokes in question are cheap as well. A very modest one from Hammond can easily fill the bill. Don't forget that these are in the milliHenry range.
Argument 3: Its only suitable for underwater signal repeater and not suitable for home audio use.
Fair enough. I can say from my experience that it is not hard to design or build and it performs as simulated. I could measure no difference wrt distortion either. I have no particular preference though and still use standard choke or cap input from time to time, depending on the task at hand.
Ian
Last edited:
- Status
- Not open for further replies.