|'); } // End -->|
when deciding the LC filter setup after the rectifier tube what are the factors determining LC or CLC is to be used? the rectifier tube?
Getting enough voltage is one, an LC will put out a lot less voltage than a CLC for the same transformer. Another is respecting the minimum requirement for Henries on the LC choke. You should also use at least a 0.01uf cap before the LC to help out the choke. In general, LCs are nice providing a very stiff volage supply.
Not all chokes are designed for LC. Hammond, if I recall, treats its chokes as being "smoothing" chokes intended for CLC and not LC. I believe plenty of inmates have used chokes indiscriminately for LC, and have not had any worse results than some buzzing and other noise coming from the choke.
Using a bit of capacitance, from .01 up to somewhere between say 1.5 to 2.0 uF, takes a load off of the L while letting the filter still function as an LC. This cap also lets you fine tune the voltage within some limits. Very handy technique.
Search old posts, as the topic has come up often. You might search on "swinging" chokes, as these were purpose built generally for choke input filters.
I never found an LC to have enough filtering capacity. A huge capacitor on the LC output produces very high currents in the first inductor. I use either a LCLC or CLC with the LCLC being the better filter.
The difference in (steady state) inductor current in an LC filter using a moderate versus extremely large capacitor is basically nil. If you mean turn-on surge, yes, this could make a difference.
Yes, but not explained well by me. A large capacitor after the inductor produces currents that exceed the tube rectifier current capacity long term as well. A second LC keeps the first cap within reasonable size or a CLC. But, I am not educating you on that.
I don't want to seem antagonistic, but the choke-input filter really does protect the rectifier from high steady-state peak currents, even when the capacitor after the choke is insanely high. Check out these two simulations, where the capacitor goes from 10uF to 1000uF. Once the turn-on transient settles down, the choke current (equal to the rectifier current) is exactly the same in both cases.
(Sorry for the inconsistency in the picture dimentions and plot colors -- I was in a hurry.)
It's the AC voltage across the choke that makes it buzz, and this is almost entirely dependent on the voltage on the upstream terminal, since the downstream terminal is effectively grounded by the capacitor on that side. For a small upstream capacitor to "take the load off" the choke, it has to make a significant change in the AC voltage going into the choke. That is to say, it has to partially smooth the rectified voltage. In the process, it dramatically increases both the amplitude and the time-rate-of-change of the diode current. It also raises the supply output voltage, and reduces the "stiffness" of the supply.
You can see this easily using PSUD 2. You'll find that up to a threshold value, the capacitor does absolutely nothing to reduce the choke voltage or current. On the other hand, the increase in diode current can be seen even for small values of capacitance. This makes me wonder whether these intermediate-sized capacitors are really such a great idea, since you're basically degrading the three virtues of a choke-input supply: minimum rectifier current, continuous rectifier current, and good regulation. IMHO, the right thing to do is to use just enough capacitance to "snub" the switching transients, but otherwise not monkey around with pseudo cap-input filters.
The flux in a choke is a function of the instantaneous current flow. Presumably you're OK if the peak current is always equal to or lower than the rated DC current. You can verify this using PSUD 2. Another factor is whether the choke is mechanically solid enough to handle the physical stresses (magnetostriction) caused by high AC voltages. I have a 10H 300mA Hammond choke on my bench right now that is dead quiet in choke-input service putting out 200mA with 500VRMS on the input. Peak current is around 250mA. Whether it will stay quiet in continuous service is another matter.
(HP) For a small upstream capacitor to "take the load off" the choke, it has to make a significant change in the AC voltage going into the choke.
(HP) IMHO, the right thing to do is to use just enough capacitance to "snub" the switching transients.
Recently I have been trying to Squeak the DCR down on some Choke input chokes and as you know Choke input and Low DCR tend to be at odds with each other. The math part of the design is pretty clear cut, and a paper design with the proper electrical characteristics isn't difficult. The transition from the paper design to physical design generally nets you a fairly noisy device that may otherwise be functioning correctly (until you get to third order effects)
Just recently I tested a pair of 18hy 27 ohm chokes designed to be used in in a Choke input filter for a 280V 20ma supply. They were pretty darn quiet except for just a touch of a buzz and an underlying hum. Measurement in circuit confirmed they were behaving as designed with the filter voltage staying constant from 15-28ma of draw. (at 28ma, the dc was saturating the choke reducing the inductance to a point below Lcrit)
Knowing the desire, (make that demand), of a quiet Choke, played a bit with the snubber cap before the choke to attempt to quiet them down even more. As expected a .5 mic cap eliminated the buzz completely and the underlying hummmm remained unchanged. The cap also increased the Vdc output of the filter and reduced the Vac across the choke so there were no surprises. What did surprise me was cutting the value of the cap in half to .25u still eliminated the buzz, but left both the AC and DC behavior of the circuit unchanged.
I guess this was the first time I had noticed two distinctly different sounds to the noise a choke under high AC stress makes and the use of the so called "snubber" effectively eliminated the more objectionable buzz. Makes me wonder if different rectifiers would net different ratios of humm to buzz for a given choke.
(HP) I have a 10H 300mA Hammond choke on my bench right now that is dead quiet in choke-input service putting out 200mA with 500VRMS on the input. Peak current is around 250mA. Whether it will stay quiet in continuous service is another matter.
The DC current is actually your friend. If the Bdc is greater than the Bac, the choke will be limited to a single quadrant of operation (no reversal of the magnetic field) and will be quieter in operation than the lower flux situation that would exist with just the AC alone. Simply placing the choke across your 500Vac supply will confirm this. This is also a good test for the origins of any noise you are hearing form a choke. If a superimposed DC current quiets a choke down you need to address the mechanical aspects of the core, if the addition of DC does little to change the noise, then the structure of the coil is what needs to be addressed.
I think there's probably a lot of high-frequency crap floating around the choke and pulsing at 120 times per second. High voltage spikes should drive the chokes nuts. The small input cap probably absorbs and spreads out that energy. I haven't personally investigated with a scope what's really going on here. If I ever get the time, it's something I mean to do.
Dave, given your experience do you think it's reasonable to suppose that the small cap is eliminating (shunting) the higher even order harmonics (but not the 120Hz one) and making the chokes life a little easier by doing so?
To paraphrase Ella.
On a Choke Input filter (LCLC) what properties define a proper first choke? Is a "Swinging" choke any different than any other PS choke, or is "Swinging" just a definition of its function as the first choke in a "Choke Input" filter?
Anecdotal : On Fisher 50As I played around with some years ago, heavily modified, the stock Fisher swinging choke sounded inferior to a fixed choke I used in its palce. The music's presentation stabilized to my ears with the non-stock / fixed choke.
YMMV, but it makes ME not at all eager to employ wiggle-wagging "swingers".
Swinging chokes are only any use in circuits which change their current draw. If you changed the output to Class A the swinging choke would be pointless at best.
Non est ars quae ad effectum casu venit - Seneca.
"That which achieves its effect by chance is not art"
A swinging choke is a choke deliberately designed to change in value, from a high inductance at low DC currents to a low inductance at high DC currents. The fact that the choke "swings" isn't essential to its function as a filter. The advantage is that the very high inductance at low currents allows much smaller bleed current (needed to keep the output voltage from rising to 1.4 VRMS). To have the same low bleed current in a fixed choke would require a much larger and heavier unit. The disadvantage of the swinging choke is that it has less inductance at its maximum rated current, and therefore isn't as effective a filter.
it is also interesting to note that the whole reason the choke swings is essentially due to various degrees of saturation from the DC current.
...Which is OK as long as the choke saturates smoothly and doesn't vibrate itself to death (and turn into a dead short) as the instantaneous current varies throughout the AC cycle.
i agree, but since the DC is causing the saturation the vibration is a non issue and often becomes less as you saturate.
what i find interesting is how the saturation can be considered universally bad by some, then we have the old swinging choke sow up and it is suddenly our friend :-)
> The flux in a choke is a function of the instantaneous current flow.
The AC voltage source resistanace seen by the choke is low ( -3dB point is well below 50Hz if we want to filter anything) and hence in the first order approximation the flux in the core is determined by the AC voltage, not current, IMHO (so the flux sinusoidal, the mag. field distorted). Then the operating point of the choke's core together with the core cross-section and number of turns should be chosen such that the core is not saturated by AC flux.
The flux in the core is proportional to the instantaneous current. The coil voltage is equal to the change of flux with respect to time. Together, these two relations give the definition of inductance, V = L di/dt. At a given frequency (assuming the core permeability is fixed), current is proportional to voltage since the reactance is constant. The thing is, if you want to specify the AC voltage rating of the coil, you also need to state the level of DC current and the frequency (Lundahl does this for their chokes). But if you have a way to calculate the peak current (e.g., by using PSUD or an approximate formula like the one in Morgan Jones' book), you can go straight to the answer. I don't see why a smoothing choke rated for some maximum DC current shouldn't be able to handle a lower DC current with a superimposed AC signal, as long as the total current is equal to or less than the rating. The problem, presumably, is when you take a smoothing choke with a given DC rating and and drop it into a choke-input filter supplying that amount of current without taking into account the superimposed AC current.
I believe what you are saying is correct, but I feel the relationship between current and flux is more fundamental since it doesn't depend on time.
What you say is a stationary, linear analysis. There is no such thing as fixed perm in an iron core, you have a BH loop. Look at the Ampere's Faraday's laws in a medium to see which qualities are fundamentaly related: change of flux (that is B*A) is related to the voltage, while ciculation of H-field to the current. Current *is not* fundamentaly related to the flux!
Now in the situation when driving resistance is very small you can in the first order neglect it completely and hence the voltage across the choke = the voltage of your AC source and the core excitation is voltage driven. At another extreme, high series R, the voltage can be neglected and then the current produces H field that excites the core.
In real life we are somewhere in between but handling such situation analytically is extremely difficult (a nonlinear, implicit diff. equation).
*All of the above applies to AC accross the choke only!!*
With DC the situation is reversed somehow: constant current creates constant H-field, moving core's op point more into saturation region (for realistic chokes it usually drives it deeeeep into saturation, hence we need an airgap).
I refer you to, say, Grossners book on transformers.
I do know Faraday's and Ampere's laws. I can't deny that if you bring change in core perm with DC current into the picture then you have no choice but to solve the nonlinear partial differential equations. I would argue that lacking detailed specifications (beyond rated inductance and DC current) for a choke and given the difficulty (as you say) of making the calculations, it's all but impossible for an amateur designer to come up with exact answers. I think it's an adequate approximation to treat the choke as a linear inductor of stated value for any instantaneous current up to the rated DC maximum, knowing that the incremental inductance will be higher at marginally lower operating points. In that case, designing for Idc + Vac / L < Imax should be a conservative solution.
I grant that from the perspective of the transformer engineer this is a simplification, but as a circuit designer I think it's defensible, provided one bears in mind (as you have reminded) of what has been glossed over.
I'm expecting a physicist to chime in and scold us for using Ampere and Faraday instead of Maxwell.
Things would be very simple of manufacturers would simply provide a chart stating the maximum permissible DC currents for a range of RMS secondary voltages in choke-input service.
Henry, I think you're right. Assuming a reasonably constant perm, the instantaneous flux is a function of the instantaneous current. The rate of change of flux is related to the voltage across the choke, which I suppose is important when actually using the thing, but is not required to calculate instantaneous (or max) flux.
hp> I'm expecting a physicist to chime in and scold us for using Ampere and Faraday instead of Maxwell.
A good physicist would scold you for not knowing that half of Maxwell's equations are essentially Ampere's and Faraday's laws. :-)
I'll take N-set's word for it that he knows what he's talking about, so I'm waiting for more details. In my head, I see the B-H curve. I see that H is a time-invariant function of current, and I see that for a given value of H, I can pick off the resulting value of B from the curve, even if the curve isn't straight. This ignores hysteresis, but even allowing for a loop, the maximum value of B for a given peak H doesn't increase.
I do recognize that if you apply a fixed AC voltage to the core and a variable DC bias current, the AC current will vary with the permeability as as you move to different parts of the B-H curve. If the fixed bias plus delta-H pushes you into a region where the B-H curve is flat (saturation = low perm), then the AC current will go through the roof. It follows that you cannot readily predict AC current as a function of applied AC voltage in the presence of a DC bias on a nonlinear inductor. But, it seems to me that if you know the coil can support a given value of Hmax without saturating, and if you know that Hmax is developed with DC coil current Imax, then as long as the total instantaneous current, Idc + Iac does not exceed Imax, the core shouldn't saturate. Silly me, maybe I'm missing something obvious.
The other thing is, in a choke-input filter, the coil current never cuts off or reverses, so I don't see how the choke will ever operate outside of the first quadrant.
For any given gap size the Bdc can be formatted in "gauss per madc"
For Bac you need to add time into the equasion, so it becomes "gauss per volt AC @120hz"
armed with this info and some very simple math, the end user can be in control of how they operate the device.
Somehow I didn't expect you to agree with me from the beginning...
1)The equations are not partial. If I have time I will put them on my site
2)Design for Idc + Iac is a nonsens, as AC flux matters more and can drive the core into saturation and/or out of first quadrant
3)Treating iron as linear, even by a circut designer is shooting his own foot
Does your argument depend on hysteresis? If not, isn't it true that if you know H, you know B, and if you know I, you know H? If not, please explain.
Yes it *strongly* does: no hysteresis loop means
=> no saturation, no worry with AC flux
=> B & H easily related to each other (say linearly) and what you say is true
Presence of histeresis makes it hard to calculate B having H or H having B, because:
1) hard to build a good mathematical model of the hysteresis loop; classical model is Rayleigh loop, built from two parabolas, more advanced treatments analyze more complicated loops, say given by Taylor series around Rayleigh loop, with fitting parameters to match the real life.
2) even when we have it the relation between H and B can be found only locally (loop is a plot of a function only locally) and is non-linear
3) we land with nonlinear, implicit diff. eq.
Hence one tries to find a small parameter (order parameter) in the equations and build approximate series - standard technique with "weak" nonlinearities. All the mainstream particle physics with their Feyman diagrams etc, is based on that.
For the transformers sin driven the order parameter is roughly -3dB frequency of the LR circuit/driving frequency.
All this just realizes that i should finally wrote those things down once and forever. If I dont do it till the end of the week - call me a moron!
OK, this makes a lot more sense. I was afraid that perhaps I was just totally losing my mind. I don't have any data on core losses in typical power supply magnetics. I've operating under the assumption that basic approximations such as the one Morgan Jones presents in his book are adequate. The alternative is to get degrees in Physics, Math, and Computational Methods...
As I said core losses play a secondary role here and you don't have to have any degrees to calculate amplitude of the flux and see if the total flux is below or above 1.2T.
I'm not familiar with MJ book.
I have a choke sitting here. It's a Triad C-16A, rated 10H @ 200ma, 150 Ohms coil resistance. I have no other data on the choke. I propose to use it in a choke-input supply with a 400VRMS transformer and 170mA DC current. Given this information, how would I compute the core flux density in the intended application?
you can't. the three numbers you suggest are like talking about a blue vehicle with 4 tires and 280 horespower.
your best bet is to look at the size and shape of the choke and figure out what design might fit.
luckily, those numbers scream cookbook design. The 10 hy 200ma 150 ohm choke is the old standard you seem to see from everyone.
it generally represents a layer wound coil with around 3250 turns of #30 on a square stack of EI-100 lams with a gap of .022 inches.
The Bdc at 170ma will be just about 10KG, and at 120hz the gauss per volt AC will be about 10
chances are it is either M7 or M19 so 14KG is a safe bet for Bsat.
While the whole reverse-engineering is fun, why not just hook the damn thing up to your 400V trannie with a pair of diodes and a cap and compare the results to what a simulator says? the only unknown is the behavior of the choke so its electrical characteristics should be pretty easy to spot.
Thank you for making my point. You can't characterize the coil without knowing the number of turns, the core dimensions, and the lamination material. You also have to have the magnetics knowledge to work out the answer. It isn't rocket science, but it's outside the mainstream of hobbyist knowledge, including my own.
If I understand you correctly, the AC flux is only 40 Gauss, compared to 10,000 Gauss for the DC flux. I'm surprised it's that low. The Jones formula and PSUD 2 predict a peak AC current of about 30mA in this application. I picked the 170mA operating point based on this number plus the 200mA choke rating. Also, that's about the lowest current (including the bleeder) that I can get my amplifier design to run at. I have other chokes rated at higher current, but they're a lot larger and I'd rather save them for other projects.
So perhaps its a good point for taking your hobby out of the mainstream and moving it to the next level? Again: treating magnetics as linear is shooting ones own foot! Spend a night with google: look for some sample BH loops to get some feeling, write down a formula for Bac, or better yet derive it if you can, look at what the gap does, etc. And at the end be smart and improvise, just like Dave did!
I dispute that you shoot yourself in the foot by ignoring nonlinearity in magnetic components. As circuit designers, we are pretty much forced to rely upon off-the-shelf chokes and transformers in our work. We know that these components are nonlinear, but we take care to design with sufficient margins that the impact of nonlinearity is minimized. I can't deny that more knowledge is always a good thing, but I will assert that it's possible to get preoccupied with specific details at the expense of the overall picture. My goal designing a power supply is to deliver clean, steady power to my amplifier circuit. It is not to understand in intimate detail the operation of the filter choke. Magnetic design is a fascinating and worthwhile subject, but it's not the aspect of the hobby that most interests me most. Lacking sufficient time, energy, and intelligence to be an expert in every subject, I am forced to pick and choose.
I am satisfied that I can get reasonable performance and economy using rougher estimates than the ones you would prefer. I understand and respect your point of view, though, and appreciate your explaining it. It's motivated me to revisit topics I haven't thought about in such detail for a number of years.
the AC flux will ve 10 gauss per volt so if you have 400VAC that translates to 4000 gauss.
when you mention peak current do these calculations just assume 120hz operation (or 100hz) in any event it seems with either method you seem to be operating the choke in a place where it will be happy. Now all that is needed is to build the circuit and see if it complains :-)
Sorry, that was a brain fart on my part.
The Jones formula is based on a Fourier analysis of the rectified AC waveform. It assumes the choke current is dominated by the second harmonic of the powerline frequency, both because the higher-order components are lower in amplitude and because they are more strongly rejected by the choke. It also assumes the choke acts like a fixed inductor of the rated value and that the flux adds linearly in proportion to the sum of the DC and peak AC currents.
It's interesting that your numbers and mine seem to agree almost exactly, but I think that is mostly accidental. Still, to anyone that says my approach is "nonsense", it would help to have a quantitative example giving an estimate of the magnitude of the error.
(HP)It's interesting that your numbers and mine seem to agree almost exactly, but I think that is mostly accidental.
I don't. if you look at it aren't you just using a different method to essentially solve maxwells equasion in a relative manner.
if you double the frequency, the impedance of the choke will double and for a given AC voltage your current will be half. This is the same relationship that happens with flux so i would expect the results to still be accurate as frequency is changed In my method i am simply using a few assumptions to convert DC current (H) to an equivalent value of B. MJ just appears to do the opposite buy converting the flux to current. The end result is getting both the AC and DC characteristics in the same scale for ease of comparison.
Actually, MJ's formula doesn't know anything about flux. The only tricky part about it is using the Fourier transform to find the amplitude of second-harmonic component of the ripple voltage. Beyond that, it's just a simple matter of computing the inductive reactance at 120Hz and dividing that into the voltage. It looks like our numbers agree exactly because 10K + 4K Gauss is a good number for core saturation flux density, and 170 + 30 ma adds up to the rated Idc of the choke. The reasons I say this coincidence is an accident are:
1) The actual 120Hz AC voltage component is significantly less than 400V. It's also an RMS voltage, not a peak voltage (but these two factors may roughly cancel each other out).
2) The inductor isn't perfectly linear so the value of incremental perm at 170mA is only an approximation over the entire AC flux range.
3) It's unlikely that the rated current puts the inductor right at the edge of saturation.
On the other hand, I don't think it's entirely accidental because, clearly, the calculations we've been discussing take you will into the ballpark of the actual numbers -- good enough for our purposes, I would think.
Me too, thanks Henry!
However, I have heard that using a small cap (about 0.68 uF) before the choke can be helpful in curing buzz, IF it turns out to be a problem, without compromising the LC filter characteristics significantly.
Also, I suspect that when people get away with using smoothing chokes instead of swinging chokes in LC power supplies it's because there is not a great difference between quiescent current and maximum current of their amps. If there were a big difference, such as with a class B output stage, then I think a swinging choke would serve them much better.
There's no major disadvantage to using fixed chokes provided 1) you take care not to exceed the peak current (or DC current plus superimposed AC voltage, depending on your point of view) of the choke, 2) you can afford the higher bleed current, and, 3) you can tolerate the larger size and weight. If these constraints are met, then you're pretty much good to go. The only other advantage I can think of with the swinging choke is that the lower inductance at rated current should give you less filter "bounce", but this is easily made up by using more capacitance.
The big advantage of fixed chokes is that you can actually buy them, whereas nobody seems to make swinging chokes anymore.
This post is made possible by the generous support of people like you and our sponsors: