|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
216.254.76.142
And why?call them green and blue...
Follow Ups:
It seems to me that there is a desire for the power transformer's current waveform to be a sine wave, like the voltage waveform. It is my understanding that the only way this can happen is if the transformer's load is non-reactive or resistive only. Of course we can't have that in a filter using caps and inductors....or can we?How about if we make the reactance of the inductor and cap (more or less) equal at 120 hertz and follow that first LC section with another LC section of more normal values? Granted we are going to have some rather high voltages being developed but if the goal is for the current waveform to look like the voltage's does this idea have merit?
"How about if we make the reactance of the inductor and cap (more or less) equal at 120 hertz..."That would put the resonant frequency at 120Hz. Not only would you get no ripple reduction, you might actually end up with MORE ripple! FWIW, Henry's flywheel filter ends up being 'tuned' (sorry, Henry) at something around 120Hz, maybe a little higher, I don't remember exactly. Not sure what that has to do with anything.
Anyway, I'm not sure I understand why we want the current to be sinusoidal in a choke input supply. It seems to me that if we consider an ideal choke input supply with very high inductance and perfect diodes (zero dropout voltage) then the current through the choke would be constant, not sinusoidal.
Could be I'm missing something... If I've got it wrong then I'm happy to be corrected. :)
What if a parallel C was used, tuned for resonance at 120Hz (for you 60Hz mains people) as was often used in Radio TX power supplies. I've been very happy with the results where I've used it, even with the potential adverse side effects. I wonder if anyone has modelled this?
"It seems to me that if we consider an ideal choke input supply with very high inductance and perfect diodes (zero dropout voltage) then the current through the choke would be constant, not sinusoidal."I look at it this way.
The current through the choke in a choke input PS is constant and sinusoidal. Constant only meaning that the current never stops flowing.
On the other hand. Look at how low the amplitude is of this current waveform. That's almost constant. That's what 250Hy will get you.My 2 cents.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Simulate for 26ms after a delay of 1 second.Then try ...say a 3H choke, a 0.47uF cap, and a 2K8 load resistor and see what I(T1) looks like. By adding another normal sized LC filter you can get the ripple reduction and voltage regulation acceptable and still keep the I(T1) looking about the same.
Not that I know if this is a goal worth chasing mind you:)
Hey Russ,I'm not sure that it matters either, but it sure is fun to look at. ;-)
When all of the talk of RF hash being created and broadcast, i thought it might be interesting to look for the source. This is when i noticed that the current waveform for the "real choke input" seemed to have the most HF content.
How this content gets out is a whole new can of worms i'll save for a later discussion. What i do want to point out is that a distorted current drawn through a resistance gives a distorted voltage. luckily we can consider the line voltage to be a very low impedance source so this would seem to minimize the damage, but remember we do need to add the DCR of the windings as series resistors and suddenly we have the mechanism in place for the creation of a distorted voltage.
still not sure what that means in sonic reality, but we are determining merit by simulation dammit and i say the prettiest picture must sound best.
I'll also add that it seems logical that both undistorted current and voltage waveforms are desired when dealing with filter systems.
"This is when i noticed that the current waveform for the "real choke input" seemed to have the most HF content."I don't understand. Are you saying that this will throw more RF than the one in the link below?
There are two "spikes" per cycle in both waveforms and these are much smaller than the spikes without the choke.
Thanks, Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
In this scale, RF content is represented by a vertical line. the more the line approaches a pure vertical, the higher the frequency content. The magnitude of the signals has nothing to do with this. By adding a small choke in front of the first C you end up greatly reducing the narrowness of the current draw "spike" netting what appears to me to be a current draw with much less high frequency content than even a "real" choke input filter.
"By adding a small choke in front of the first C you end up greatly reducing the narrowness of the current draw "spike" netting what appears to me to be a current draw with much less high frequency content than even a "real" choke input filter."In this example, what value choke would be "best".
Thanks, Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
The best choke would be the one chosen by ear, but since we are in sim land you first have to decide what you want the current to look like then decide what looks best. How that translates to what sounds best is still up for discussion.
Anything that adds resistance/impedance between wall power and first cap will help increase conduction angle. Any increase in conduction angle moves you closer to a real choke input supply.Play around....up the transformer DCR...go to FWCT and a 5U4....try an RC filter stage. There is no best here. It is more of a mental exercise and to see if what looks "pretty" on the scope/sim.....and/or.....to see/hear if drawing current at a steady constant level (that reverses each 1/2 cycle) is better than drawing current in sine wave manner. Don't forget to increase load as needed.
Being who and what we are here, I reckon when we see an evil square wave we freak and think "sum of odd order harmonics".
but russ, that Evil R instantly converts distorted current into distorted voltage.
I didn't mean best in that way. When I add a small choke it lowers the amplitude but doesn't change the shape. Am I missing something?Thanks, Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
if it lowers the amplitude, it has to change the shape. Assuming the same power into the load, the area under the plots must remain the same independent of filter type, so if you decrease amplitude, you must widen the trace at some point to keep the area constant.
It appears that the total magnitude is higher for green, but green drops off faster/more with increasing frequency. I'm a little hesitant to mindlessly trust the sim results. It would be interesting to build just these two simple supplies and measure the actual noise.
Yes, the FFT plot was an eye opener. I'll take it as gospel but it wasn't what I expected!So if we are going to draw power from a transformer and feed it into reactive parts how do we get some same voltage and current waveforms that we would if said transformer supplied power to a resistive load? I suggested one way but perhaps you have another you'd like to share? Also, have you played around with putting line reactors (chokes) before the power transformer? Maybe if we "contained" all the bad looking stuff to only the PT we are doing fine???
Anyhow, interesting discussion and I am glad to see I am not totally off base.
Russ
P.S. It looked like you were a little "kind" with the cap input filter. Perhaps we could come up with one that had worse spikes and see how that models.
actually the green trace was for a Low L input filter... it sort of takes the edge of the full C input.
Did the choke and cap models include parasitics? I've been playing around quite a bit with Spice sims of supplies and the effects are an eye opener above the first cap self-resonance.
Thermionically addicted.
Even on a picture-only basis, we're looking at less than half. Step Tre's 250H/220uF load and see what happens to the current.
Thermionically addicted.
I think people are getting bit confused here. I am referring to the current through the power transformer primary which represents how current is drawn from the wall.
I see what you mean.It seems to me that the theoretical limit of choke input (infinite inductance choke, etc.) would make the current waveform in the transformer a square wave. Obviously, that's got tons of hf content. Cap input has spikes which also have hf content. If a little inductance at the input helps the cap input then maybe critical inductance would be just right. Spread the cap input current waveform out until it just barely conducts from start to finish.
Unfortunately, that would work at only one average DC current. Change the load and you lose that balance. (Hmmm... I've heard people report that they like swinging chokes....)
It also assumes that the critical size inductor works as well at high frequencies as the deliberately subcritical one. That's a practical issue.
Eh, I suppose I could sim it myself, but it's a really nice day and I just don't feel like it. Maybe later. Maybe. :)
Namely why so many had problems with the current waveform that I assumed to be a choke input supply. It seemed a lot of folks liked the notion of the PT's current waveform looking like its voltage waveform (like sine wave rather than square wave).So I got thinking about how we could make the current waveform look like it was feeding a resistive load and came up with the above notion. If it was followed up with a more normal LC section it seemed possible to get reasonably good ripple reduction and voltage regulation and have the PT's current waveform very nearly sine wave like.
Russ
And yes, as soon as I hit that post button it occured to me the flywheel filter might have been tuned to 120 hertz but for some reason I never bothered to consider that when I was looking at the flywheel concept.
Perhaps I am feeling guilty about picking a path long ago and feeling no reason to second guess it. So on one level a normal choke input filter is the right choice for me and one I had been comfortable with. But I am not willing to disregard what Jeff claims to hear and am intrigued with why Henry would spend the effort. That FFT plot also surprised me but I don't know enough to interpert it properly.
I'm totally hip to taking a fresh look at things we might take for granted.And... if sinusoidal current just means that it's devoid of high frequencies, then I'm hip to that too! :)
Hi Dave!What might even be more educational is to monitor the secondary voltage of the power transformer. This shows quite interesting things.
Whilst we like the advantage of the quite constant current draw of choke input supplies, your simulation shows that there is an instantaneous reversal of this current in the power transformer every cycle. Difficult to believe that this does no harm.
I always tame the spikes generated by this current reversal by a small cap before the first choke (in case of single phase supplies). I monitor the secondary voltage on the scope and size the cap such that the waveform looks like an undamaged sinus again.
This problem with choke input filters is even worse with silicon diodes. I think at the point where the current reverses there is a small amount of time in which both diodes fight against each other.
This is a bit different with three phase supplies. The current is not forced to reverse in the secondary but is handed over from one phase to the next. Still at this hand over point there is a nasty spike while the two diodes fight. While the first one does not "want to let go" the next phase's diode want's to 'take control'. In a three phase supply this can be avoided with a very small series resistor in front of the diodes of each phase. I size this resistor such that each diode cuts off a small time before the next one starts to conduct.
Best regards
I think at the point where the current reverses there is a small amount of time in which both diodes fight against each other.I think it is exactly the opposite: a small amount of time where none of the diodes conduct. The secondary of the PT causes the voltage peak because the current is suddenly cut off.
Hi!That may well be. I wrote "I think" because I'm not sure. It's been a while since I played with this since I switched to three phase years ago.
Should be easily verifyable by monitoring the volatge across both diodes and see if there is a overlap or a fragment when none conducts.
Then there would a period in which the voltage across both diodes is above the "knee voltage" at the same time.I thought a mechanism like this is in play here: The choke tends to keep the current flowing. Thus the voltage at the end connected to the diode is pulled lower and lower, so the diode keeps conducting. When you scope the voltage at the input side of the choke (or after the rectifier) you will see that it is not just a rectified sinus going from 0 to the crest voltage, but it also sags below zero, which would indicate that the diode would still conduct even if it's input voltage approaches 0. But at a certain point the other diode will start to conduct too.
A scope picture of the secondary voltage points to this direction. It seems as near the zero crossing, the voltage "sticks" for a while. I assumed that is the periode when both diodes are conducting.
But again. I'm just wildly guessing here.
This certainly is the mechanism with three phase supplies. There the insertion of the resistor helps. This causes one diode to stop conduction just before the next begins. Although with this "fix" the current also ceases flowing for a moment, no voltage spike arises.
Best regards
All this talk about trees, how are you implementing three phase?!?
Thermionically addicted.
Hi!There is an article in Sound Practices 17. This issue was only released on the archive CD.
There is nothing special about it. Three phase schemes can be found in many electronics books. For example in the first chapters of the RCA transmitting tube manuals.
There's also an article in wikipedia. This is the scheme I typically use:
http://en.wikipedia.org/wiki/Image:Three-phase_bridge_rectifier.jpg
only TV dampers instead of the silicon diodes.
Let me know if you have specific questions.
Thansk Thomas. I understand the use of 3-phase and its advantages but are you using direct 3-phase service in the home or synthesizing it from single phase?
Thermionically addicted.
Hi!3 phase service right out of the fuse box available over here :-)
I remember discussing the 3-phase supplies with you quite some time ago. Short of running a rotary converter it isn't an option for me. But I might have a chance to pick up a bunch of SLA batteries from a UPS that is being replaced. That would give me a 480VDC supply (but I am a little concerned about safety!).So have you ever tried a battery (only) supply for B+ and if so, how did it compare?
Hi!No I haven't. I plan to run a phonostage off batteries. But that's a future project. A friend of mine is using a battery powered phonostage and likes it very much (he uses 3 phase for most of his system tto)
just to see what happens. I'll post the results when I'm done.
Naz
Thomas,
I'm sort of with the "there is a small amount of time where the diodes fight each other" school.Actually what happens is that this:
A diode when conducting has a depletion region of a certain width with a electrostatic field across it - that is it has a capacitance. When reversing the applied voltage you have to discharge that capacitance (the physics bods talk about sweeping out the minority carriers but its easier just to think in terms of discharging the cap). This results in a sharp peak of current FROM the power supply capacitors BACK INTO the diode before it turns off. The size of this "splat" of current depends upon the size of that intrinsic capacitor or more exactly the amount of charge in that capacitor which must be swept out to turn the diode off.
This charge is called Qrr the "reverse recovery charge" and it will not be a surprise to most of you that typical values are:
standard silicon diode - 500nC
Ultrafast Soft Recovery - 100nC
Schottky Diodes - 50nC
Silicon Carbide Diodes - less than 20nC
Tube Rectifier - 0nC
Thomas> > When you scope the voltage at the input side of the choke
> > (or after the rectifier) you will see that it is not just
> > a rectified sinus going from 0 to the crest voltage, but
> > it also sags below zero, which would indicate that the diode
> > would still conduct even if it's input voltage approaches 0.When the diode stops conducting it doesn't care
what the choke wants - it isn't conducting.
Now you have the inductor drawing current that
the diode isn't supplying so then you get the
lowering of the choke's supply side voltage.
The choke is sucking every available electron from
the wire and causing a voltage spike which only
discourages the diode from conducting all the more.
...add a resistor ACROSS the secondary winding of the traffo (before any diodes) so it's current isn't completely cut off when the diodes go into their "dead band" near zero crossing. It's the sudden change from some current to no current that creates the voltage spike. It helps on both cap and choke input supplies.I size this R by looking at an unconnected low voltage winding (heaters etc) on the same traffo with the delayed sweep on the scope, and decrease the R untill the switch-on and switch-off spike is vanished, or at least majorly reduced and turned into a soft ripple rather than a spike.
Sure this R uses some power - but with your wall of rectifier tubes I think you wouldn'd notice the extra heat!
I also tried using the same resistor on the HT but it uses MV rectifiers. I couldn't really hear any improvement so did away with it but on the LT (a seperate transformer) it was a significant improvement. Cheers Allen!
Hi Allen,
Am I right in thinking that this resistor can be on ANY of the secondaries since i guess what we're worried about is what the core sees? AC-heated filaments make a nice resistor. Is this one of the reasons that AC-heated filaments typically "sound better" than DC? In other words, more to do with the transformer than the tube. Interesting. Typical values?
Hi Allen, What sort of current is needed through these resistors ahead of the diodes? Is it a few ma? Or a few tens of mA? Is it proportional to output DC? What is a good starting point? Thanx.
Hi Allen!neat trick with that resistor across the secondary. Never thought of that. Thanks for pointing it out. I will give it a try.
Will you be in Munich during the High End fair ? There will be a small meeting at my place. We will try some Goto Tweeters with my speakers. It will be on the 19th in the evening.
N/T
Dave, seems to me that they are both LC filters but probably with a sub 1H choke for green and a more conventional value for the blue. I agree with Dave that higher peak currents do not necessarily make for a better supply. Intuitively I avoid high peaks wherever possible so my vote is for the Blue ... I like blue anyway.
What I'd like to see is the green one but with full conduction angle. Possible?
If you believe the sims, insanely large inductances net you sinusodal current draw from the source. Maybe we can petition the power company to supply us with some giga-henry input chokes to play with :-)
in keeping with the theme, here are the FFT's for the simmed waveforms above.
I zoomed in a bit on the interesting bits here
Sorry I don't know how to read this. Care to explain? Am I seeing that the green trace is quieter?Oh, no need to answer now and spoil others fun.
Green. Complete guess since load conditions, etc., aren't specified but as mentioned below green will have less RF hash and appears to be capable of greater peak current. Spice sims I've run also provisionally support the notion that any generated RF can be driven below the floor by judicial choice of uH inductors and small caps.
Thermionically addicted.
Assuming that the plots represent the current in the transformers primary:Blue has the lower peak current, so it will have lower I*I*R losses than green. Green has less sharp edges then blue (the slew rate is lower), so will generate less 'hash' than blue. Life's a compromise, so what is you goal?
hey mikey,i really don't have a goal so to speak, just interested in some interpretations of some pictures if you will. The slew rate and hash comments seem in line with what i would think too.
.
At first glance, the green has a nicer shape. There is a lot to be said about a nice waveform. BUT...the green spends a significant amount at 0 A current.Blue, while a bit distorted, is ALWAYS pushing current.
not knowing the context, the place of the measurement or simulated response, the conditions of the measurement, nor the application that the power supply operates with, I'd choose green because I like the color better.
Good answer...The traces are the current waveform of a power transformer primary.
dave
... seems to have a better regulation. Doesn't it? (I assume tranny DCR is the same)Will we win any prize?
I like the green one. It's capable of a greater current excursion.
| ||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: