Member
Joined 2009
Paid Member
With higher current, chokes can often be much smaller in their inductance. The issue is that you need a bigger gap to avoid saturation which cancels some of that benefit. But there might be interest in a power supply that didn't rely on electrolytic caps.
Oh, there's just as much subjectivity in the SS world. Perhaps more than in the tube world, but some folk aren't comfortable crossing from one to the other.
Regarding the 'flywheel' labelled LCLC filter (hpasternack post #166), it is somewhat similar to a CLC filter where C is relatively small, in that it is a transition from 100% choke input to 100% capacitor input filtering, but coming from the other end.
For B+ supplies, I'd think that secondary winding leakage inductance might be getting in to the mH range - I'll see if I have any leakage measurements floating around.
The disadvantage of any choke input filter (even a small choke variant), is the dI/dt management at the diode - choke node, which has only stray capacitance to 0V. Certainly a place for valve diodes, or UF4007 types.
For the post #166 example, each diode conducts in two bursts during its 'on' phase, with the initial current pulse falling back to zero, and then rising from zero again for another go. This resonant behaviour certainly requires a well insulated choke, and perhaps a little more PIV margin on the diodes (as valve diodes are apt at providing).
Reading through this thread did remind me of the Berlin Tech Uni's Project Wildcat, which effectively set up a test between a classic Fender guitar amp that had nominal mains ripple, and a specially modified version without any ripple - guess which version sounded best!
https://www.ampbooks.com/mobile/classic-circuits/class-AB-ripple/
For B+ supplies, I'd think that secondary winding leakage inductance might be getting in to the mH range - I'll see if I have any leakage measurements floating around.
The disadvantage of any choke input filter (even a small choke variant), is the dI/dt management at the diode - choke node, which has only stray capacitance to 0V. Certainly a place for valve diodes, or UF4007 types.
For the post #166 example, each diode conducts in two bursts during its 'on' phase, with the initial current pulse falling back to zero, and then rising from zero again for another go. This resonant behaviour certainly requires a well insulated choke, and perhaps a little more PIV margin on the diodes (as valve diodes are apt at providing).
Reading through this thread did remind me of the Berlin Tech Uni's Project Wildcat, which effectively set up a test between a classic Fender guitar amp that had nominal mains ripple, and a specially modified version without any ripple - guess which version sounded best!
https://www.ampbooks.com/mobile/classic-circuits/class-AB-ripple/
I apologize for my novice way to understant electronics, but I have a question.
Obviously, mains ripple affects sound so, given the application (i.e. hifi), ripple elimination shouldn't be the first consideration when designing a psu? Transient response, reliability and cost do matter as well but we need "clear" DC.
That said, the "flyweel" has to be followed by pi filters with big capacitors or chokes. So, besides that a small reservoir cap can be used, are there any other benefits of this design?
I think DF96 covered that when he said it would be a good trick when only small caps are available (as they once were). Maybe one or two on this thread do that by choice. It is an interesting thread (and thanks hpasternack), but yes I have some larger caps where they are being useful 😉
Member
Joined 2009
Paid Member
The maths taught to EEs ought to be enough to handle this circuit, as it is basically just some calculus possibly combined with simultaneous equations, but I suspect that applying the maths (e.g. choosing a suitable approximation) might be beyond some undergraduate EEs. When I get some time I will have a think about it myself.hpasternack said:
I don't follow your reasoning here. The secondary waveform (yellow) isn't shown rectified. There is one diode conduction period for each secondary peak, negative or positive, and smooth diode current pulses with no signs of resonance in the simulation or real life.
When I built and tested the circuit I used some old HEXFREDs I had in my junkbox. On the scope the waveforms looked exactly like the simulations without parasitic ringing. If ringing were a problem in a particular application, the usual sort of snubbers could be added.
Some people have said that the input choke is overworked in this circuit. I contend that as long as the DC rating of the choke is greater than the peak diode current, the choke won't saturate and should be fine. I used a Triad C56U and it ran quietly with no signs of distress. Of course the choke has to be insulated well enough to handle significantly higher terminal-to-terminal AC voltage than in a conventional filtering application, but the absolute voltages here aren't especially high in the grand scheme of things.
I don't advocate for the use of this circuit; I just think it's interesting. On the other hand, people seem to enjoy trying different things, and the flywheel is different enough maybe to justifying trying it out just for the sake of novelty.
Yes, it's just calculus. Do the integral over the conduction period and calculate the average C1 voltage as a function of DC current. From there you can get the equivalent DC source resistance. It's a bit of a nuisance because the circuit is non-linear so you need to figure out in advance when the diodes turn on and off and adjust the limits of the integral accordingly. Or so I imagine.
Confession time: I suffered from math phobia in high school and college. I did well in my education but compensated for less than stellar math skill by learning to think intuitively about how circuits worked. The equations on their own never really spoke to me. I wasn't completely incompetent at math, but I definitely had to work harder on account of this weakness.
Since leaving school and taking up a career in software development, my math skills have deteriorated badly. I still have that intuition, but couldn't solve an integral if my life depended on it.
It turns out that whatever fascination I had with software way back when has almost completely evaporated. I despise writing software for a living now and wish I could do real engineering, but I'm not even close to qualified at this point. It's ironic to be very competent, and well compensated, at performing a job that one can't stand. Oh, well. Had I pursued an electrical engineering career I might be saying the same thing now and wishing I had gone into software. You only live once.
-Henry
Hi y'all, can you take a pic and scan it and post it here? There is no problem with posting it here,
for the discussions sake, just cite the book, I'll do that here so you can just copy and paste.
NOTE: For the third edition, look at the tittle page, change the date and then
put 3rd ed. where they 4th ed. is.
In what chapter of the VA book does this fall? I have 2nd ed here.
Jones, M. (2012). Valve Amplifiers, 4th ed. Newnes, Oxford, England.
We are able to use small amounts of published work under fair use for the
sake of discussion, learning, humor, etc. And, we are well under the minimum
of 5 percent.
Happy New Year Henry, nice to see you floating around.
for the discussions sake, just cite the book, I'll do that here so you can just copy and paste.
NOTE: For the third edition, look at the tittle page, change the date and then
put 3rd ed. where they 4th ed. is.
In what chapter of the VA book does this fall? I have 2nd ed here.
Jones, M. (2012). Valve Amplifiers, 4th ed. Newnes, Oxford, England.
We are able to use small amounts of published work under fair use for the
sake of discussion, learning, humor, etc. And, we are well under the minimum
of 5 percent.
Happy New Year Henry, nice to see you floating around.
Last edited:
Back in the day, a fellow named Jeff Medwin (drlowmu, who made an appearance earlier in this thread), introduced the concept that later became the "flywheel" on the AudioAsylum Tube DIY forum. Jeff's thinking, such as it is, has evolved over time. His original contention was that the defining and essential characteristic of a high-fidelity power supply was ultra-low DC resistance. He was quite emphatic about that.
Later on, "Low DCR" evolved into "Low Storage Energy Supply" (LSES, or sometimes parodied as LIES) -- which means having the absolute minimum stored energy, e.g., reactance, in the supply filter. LSES was the invention of another fellow named Dennis Faker, sorry, Fraker, the proprietor of a boutique company called "Serious Stereo" that manufactures very expensive but otherwise ordinary SE 2A3 amplifiers designed according to a dogma that Medwin and Fraker take very seriously but that most technically-minded people have a hard time understanding or seeing what makes it special.
It was at the time, and may still be, part of the gospel of tweak tube audio design that cap-input filters are noisy and nasty and dramatically inferior to choke-input filters. There are some valid technical reasons to believe this. However, one major disadvantage of choke-input filters, which has been known for decades, is their almost unavoidable tendency to ring or "bounce." This is on account of the very large value of input choke required. The cure for choke-input filter bounce is to increase the resistance in the secondary circuit through the use of vacuum tube diodes and relatively high DCR chokes. Also, if you make the filter capacitor following the input choke very large, in combination with the series resistance, the result will be the LC resonance turns into a pair of LC and LR low-pass filters that doesn't ring. However, the filter will be very "slow" and have relatively high DCR which directly goes against the LSES design methodology.
The claim behind LSES is that some ripple is tolerable, but that very fast power supply recovery time, and non-resonant transient response, is what's needed to make a good-sounding amplifier. Therefore, only a modest second filter stage is used in this type of filter. The more filter sections you add, and the bigger you make the chokes, the more likely you are to end up with an underdamped supply with a nasty impedance curve and a tendency to ring.
There was never any real analysis offered for the design choices that went into LSES, nor any convincing responses to valid technical technical criticism of the approach. The design evolved by the application of what Fraker called "cowboy logic" -- uninformed intuition and iterative tweaking and listening. Years ago I used to find this all very unsatisfying and annoying. I don't really worry much about it now though you can probably sense a touch of the old attitude in my current posts.
As I said earlier, I was intrigued by a comment someone made on AudioAsylum that the value of C1 should be chosen to maximize the DC output voltage of the supply. I initially assumed this was the result of some kind of resonant peaking, hence my choosing the name "flywheel," which suggests harmonic behavior. Later I realized there was no resonance, but I couldn't think of a catchy alternative name so "flywheel" stuck.
The first section of the flywheel isn't so much a filter as it is, I dunno, what you might call a DC voltage accumulator. I'm repeating myself now, but the flywheel effect is the increase of the peak C1 voltage as DC current goes up. This is what gives the filter lower DC output resistance than would be observed without the input choke. The second section is mostly responsible for knocking down the ripple. Care has to be taken in the choice of second section component values to minimize ringing (the 35mH/10uF first section is pretty much optimal and shouldn't be changed). Solid-state diodes are mandatory to get the maximum flywheel effect and to keep the overall DC resistance low.
I think there are enough examples of fine-sounding amplifiers that use regular old cap-input filters with big electrolytic capacitors to discredit Medwin's claim that anything but an "LSES" filter is sonic junk. On the other hand, there's enough weird second- and third-order stuff going on in audio circuits that I cannot honestly claim there is nothing of value going on here. So I don't really have an opinion. My desire all along, and it's been more than ten years, has simply been to share my technical findings, to try to add a little bit of objectivity to what has otherwise been an almost technically barren controversy spawned by Medwin and Fraker. It's amazing the discussion is still ongoing. But lots of people would probably say it's amazing people still build tube amplifiers.
-Henry
Thank you Henry for a very informative post.
I have been reading a lot of the conversation about this here and on the asylum.
i have yet try it but am quite tempted out of pure need to know curiousity
Cheers
Ed
I have been reading a lot of the conversation about this here and on the asylum.
i have yet try it but am quite tempted out of pure need to know curiousity
Cheers
Ed
I agree.
IMHO, the tube amp resurgence that began maybe in the early nineties has played out. Anyone can build a really nice amplifier, push-pull or single-ended depending on your taste, from off-the-shelf components and schematics. There are fanatics who, rightly or wrongly, still claim to be pushing boundaries and making progress. I won't explicitly say they are fooling themselves, but I do think we're at the point of diminishing returns. So now the challenge maybe isn't to do anything revolutionarily better, but to have fun trying quirky, unorthodox things just to see what happens. The flywheel seems to fall in that category.
I've had a pair of magnificent Magnequest outputs sitting in a box since 1992. Perhaps ten years ago I had a pair of matching power transformers custom built. I never made amplifiers out of these parts because I was reluctant to build anything less than a technological tour-de-force and I kept putting it off. At some point before I die I will eventually compromise and build a pair of competent amplifiers that will work very well, I am sure. I will almost certainly use a flywheel supply. For the moment though I am spending my time and money -- vast, incomprehensible sums of money -- restoring vintage BMW motorcycles.
Too many projects, too little time.
-Henry
I just let them be fanatics on here. Its more entertaining...
...." flywheel " circuit lol I waiting for the "turbo" circuit rotflmao!!!!
Sorry, I was referring to the first psud2 sim circuit in post #166, with 35mH/100uF/320mH/100uF - that circuit response shows the diodes with a double peak - with a 50Hz supply - but not with a 60Hz supply.
Your #166 post referred to the latter psud2 circuit showing the diode conduction response as being non-typical of a 'flywheel' approach.
Ciao, Tim
Last edited:
Sorry, maybe I'm being dense but I don't see a double peak. Can you elaborate? The second-to-last graph in post #166 is "atypical" in that the input choke is 320mH. The too-large choke diminishes the flywheel effect, but the flywheel mechanism is evident: peak diode current is reduced, and diode current continues to flow after the secondary voltage peak. There are two diode pulses per AC cycle because it's full-wave rectification. No resonance or "double peaking" involved.
If you sim the first circuit in psud2, and change the freq to 50hz, and look at the diode current waveform, it has a double pulse. I can't upload a screengrab for a few days.
Interesting. I'm lazy as hell, but if I can find a windows computer somewhere I might try it myself. I am curious what you're seeing.
Just pm'd you.
Interesting, at 60Hz, if the first filter cap value is reduced then it shows a double peak for cap between 4 and 7uF. At 50Hz, the double peak shows up from about 7 to 10uF.
There is obviously a resonant voltage swing on that first filter cap that can push the diodes back in to conduction by swinging low faster than the mains waveform.
Interesting, at 60Hz, if the first filter cap value is reduced then it shows a double peak for cap between 4 and 7uF. At 50Hz, the double peak shows up from about 7 to 10uF.
There is obviously a resonant voltage swing on that first filter cap that can push the diodes back in to conduction by swinging low faster than the mains waveform.
You may be right. I never saw such a thing in my investigations, but it was a long time ago. I can tell you that in my practical experiment I didn't see this behavior, but I'm not doubting you here. Just curious. Sometimes PSUD gives odd results, though. I'm curious now, will give it a try in a bit and let you know what I find.
- Status
- Not open for further replies.