EEVblog Electronics Community Forum
Electronics => Repair => Topic started by: fozzyvis on February 21, 2022, 11:56:28 am
-
I got hold of a Gossen MSP 64D (https://www.sagatron-shop.de/Gossen-Metrawatt/Laboratory-power-supply/MSP-64D-KONSTANTER/GOSSEN-MSP-64D-Lab-Power-Supply.html) for very cheap.
It only has the double 16V 1.5A module, but I can wait and search for a couple of years to find others.
Even thought it seemed to work perfectly, I opened it up to clean it up a bit and check if there was anything requiring some love.
One of the common mode chokes has had it's potting compound turn viscous enough to let gravity do it's work (see bottom right)
(https://www.eevblog.com/forum/repair/potting-compound-liquified-and-oozed-out-of-common-mode-choke/?action=dlattach;attach=1421098)
I did manage to get most of the tacky goo out and the choke measures OK (both sides very similar resistance and inductance).
(https://www.eevblog.com/forum/repair/potting-compound-liquified-and-oozed-out-of-common-mode-choke/?action=dlattach;attach=1421104)
As I do not find a similar choke, I thought I'd just fill it up with new potting compound and reuse the old one.
However, I am wondering... Is this something that happens regularly?
--> Could this be a bad batch of original (around 1992) potting compound that turned soft?
Or should I worry about something like heat that might have caused it? As I said before, I see no damage other than this on the board, and the unit seemed to work very well...
-
It was probably bad from the factory, e.g. potted with resin mixed with insufficient or degraded catalyst/hardener (depending on the resin system used). However IMHO its compromised because you have no idea what long-term contact with the goo has done to the wire insulation, + the goo may cause any new potting compound to fail to cure properly.
Your best option would be to desolder the windings from the terminal pins so you can remove the core and clean off more of the goo so you can note the exact winding layout, (photos!), and finally remove the windings, clean the core and housing thoroughly then rewind it with new magnet wire of the same gage before re-terminating and re-potting it. If the goo cleans off well with any solvent once you've got the core out, and the enamel on the windings appears undamaged and resists prodding with a wooden scraper (e.g. a sharpened lolly stick), you could probably get away without rewinding it.
-
@Ian.M That does actually sounds like a good idea. I had tried getting rid of as much of the potting as possible, but I did not get the core out. I did not know what isopropanol (the solvent I was using) would do to the magnet wire insulation. Taking out the core and manually rewinding does look like the best/safest/cleanest option. I'll do that.
As I won't be sure I get the exact same diameter of magnet wire, I can slighty change the amount of windings in order to end up with the same/correct inductance, right?
-
Looking at the markings on the surrounding caps, and the general layout of that area of the PCB, it appears this choke is part of a mains filter before the bridge rectifier. The exact inductance value will be non-critical.
A small variation in wire diameter will have little effect on its inductance as that's mostly determined by the core and number of turns. It would be very different if it was a close-wound air cored coil as then the geometry of the winding is critical.
If you don't have a micrometer, wind ten touching turns of the stripped off wire round the largest convenient cylindrical object you've got enough wire for and measure the length of the winding with Vernier or digital calipers, and divide by ten. You should be able to determine the diameter and thus the wire gage with a fair degree of certainty. If in doubt choose the thicker of two possible gages, as you don't want to decrease the current rating or increase the resistance.
Note that maintaining physical separation between the two windings is absolutely critical as I believe there is full mains voltage between them. You should also try to maintain a similar average turn spacing, but that's far less important.
-
It is indeed part of the mains filter, before the bridge rectifier. I did not realize that the wire diameter would not have that much effect. I have a micrometer which I can definitely use and will replicate the windings as good as possible.
One other question... Do you have any recommendations as to primers/texts on cap selection? There's a ton of (mis)information available where generally the idea is "replace all older caps, and lower ESR is absolutely what you want." However, I don't find too much good "basic" info that actually explains why you make certain choices. A while ago I purchased the DER-EE DE-5000 LCR meter to make measurements, but I have a hard time actually interpreting the results. I can measure Q as well as ESR which seems like a somewhat more useful measurement (though I'm just following gut feeling here, with no real knowledge.)
The reason I'm asking is the same power supply PCB.
Complete picture:
(https://www.eevblog.com/forum/repair/potting-compound-liquified-and-oozed-out-of-common-mode-choke/?action=dlattach;attach=1421377)
Right behind the bridge rectifier, it has 4 big caps (original: 47uF, 350V). I can't find a datasheet, these were from around 1992. They are in 2 series/2parallel configuration so I'd figure that ends up being 96uF, 700V (picture is with the original caps):
(https://www.eevblog.com/forum/repair/potting-compound-liquified-and-oozed-out-of-common-mode-choke/?action=dlattach;attach=1421371)
While trying to take one out, I managed to snap one of the leads. I decided to go ahead and replace them. I searched Farnell (Element 14) for similar size/capacity/voltage and found these: Ilinois Capacitor, 47 µF, 500 V, -10%, +50% (https://be.farnell.com/illinois-capacitor/476tta500arz/cap-47-f-500v/dp/2841856) as a replacement.
However, when measuring the original caps, their ESR was in the range of 0.9 Ohms at 100Hz, with a capacitance of between 49 and 52uF.
The new ones have an ESR of between 2.4 and 2.7 Ohms (at 100Hz again).
Since these are (as far as I understand) bulk storage capacitors in this circuit, I don't think that this is an issue, but once again that is just gut feeling. Should I go ahead and get lower ESR caps?
-
And, in any case, much appreciate your help!
-
IMHO, LCR metering ESR need far more than merely 100Hz
have checked it has adjuster to the higher Hz?
-
Yes, but the entire charging ripple current of those capacitors is at 100Hz. However they discharge due to the sawtooth current waveform through the SMPSU's switching transistor(s), which is at a much higher frequency, ranging from several tens of KHz to low MHz region depending on how modern the SMPSU controller chip is (as higher frequencies permit smaller, cheaper magnetic components but need more developed IC technology to be practical)
A higher ESR at 100Hz has little effect other than spreading the charging current pulses over a greater angle round the mains supply voltage waveform peaks. The ESR at the switching frequency is more important but as the voltage drop within a switching cycle due to the peak switching current is likely to be small compared to the voltage drop between mains peaks and voltage variations due to mains supply fluctuations, provided the capacitors RMS current rating is not exceeded, will only have a small impact on the overall losses.
N.B. To sum RMS currents at different frequencies, square all the currents and add them, then take the square root of the result.
-
Have seen potting run out of components before. My experience was the various inductors in the deflection system of a Trinitron monitor (HP branded, Sony chassis). Looked like it might've been some kind of resin that depolymerized, or maybe oxidized or whatever; residue was brown, gummy/waxy, seems like it moved very slowly at operating temperature. Washed off easily with solvent.
Since that toroid is only held in there by potting, I would think, best to just replace the part. Similar parts are still available; might even get lucky with same outline and pin spacing? Same or higher current rating, and same inductance say +100/-30%, no need to be picky.
As for caps, replace anything that's obviously failed. Leaking, bulged, high ESR, low C, that sort of thing. Generally speaking, capacitors are fine when they're fine; if you find a few failed parts, I'd suggest replacing all of them, as their friends likely aren't far behind. If it all tests good, fine to leave them alone.
As for what you're measuring, ESR is just that, effective series resistance. It varies with frequency, and can be modeled various ways. The most important aspects are these:
- At very low frequencies (near DC), leakage current dominates. The equivalent circuit is better expressed as a parallel R and C.
- At low to medium frequencies (typically including mains frequency), capacitance dominates, with a fairly constant Q factor. Since reactance is inversely proportional to frequency (Xc = 1 / (2 pi F C)), and ESR is a constant fraction of that (ESR = Xc / Q), ESR is also varying with frequency. (Think of it as the capacitor storing only most of the energy it's given during a cycle; like a rubber bumper, it has some "spring" to it, but it's also a bit viscous and dampens motion somewhat.)
- At high frequencies (i.e., for SMPS), ESR levels off, while capacitive reactance peters out, to be replaced by inductive reactance (ESL, mainly lead inductance) at higher frequencies. This minimum ESR point (self or series resonance) is somewhat less than the ESR at low frequency -- hence datasheets often give one ESR figure for 100Hz, then a reduction factor for the 100kHz figure, maybe 1.2x or something like that.
Note that capacitors form resonant circuits with inductances. There might be a resonance against transistor capacitance, or local bypass caps, and the trace inductances between them. It's usually a good idea for these to be well damped -- so they aren't ringing and spiking, ultimately carrying excess noise up the line. Damping occurs in a series circuit (give or take a modest factor) when ESR >= sqrt(L/C). For higher ESR, losses are higher though, so, only so much can be afforded before efficiency sucks (ESR also tends to let through more spikiness). For low ESR, ringing can be exacerbated, perhaps worsening EMI, but also in the worst case perhaps increasing the peak voltage on the transistors, perhaps to failure. (To be clear, I don't think there would be many cases where super-low ESR caps would actually cause destruction; you'd probably have to work to make something that marginal.)
Most of the time, electrolytic ESR is already much higher than needed to meet this condition, so putting in lower ESR parts is safe to do. But perhaps this gives some context as far as what's really going on.
So, as for replacements, since low ESR isn't very important, don't worry about it. Even general purpose stuff nowadays will be a far sight better than what they had. No need to go crazy, just get something of similar capacitance (maybe -20/+50%), >= voltage, and <= size.
Note that you can always replace axials with radials, adding a fly wire to make the other terminal. Use some sealant to hold the cap down, or, those things have springs anchoring them I guess, maybe trim that to suit?
Film caps rarely go bad, but if you see cracking or crazing on them, maybe better to replace. This doesn't have any, but a lot of older equipment does -- the clear gold RIFA caps (metallized paper), which sometimes fail in spectacular fashion. Same idea: good shape, no worries; cracked and bulging, better to replace.
Oh, another wear item might be the MOV, the blue disk component there, S14K something or other, not the NTCs (resistors). Can be tested, but you're probably not going to have the necessary equipment just wherever (needs to be removed from circuit, then tested for breakdown voltage and leakage current with a megger or whatever); I'd leave it unless it's obviously been smoking, or cracked. (Incidentally, this family is still available, that's a Siemens logo -- product lines were sold to EPCOS/TDK.)
Tim
-
Some cool info here, thanks all of you. I'll need some more time and a couple of re-reads to make it "click" completely.
If I understand correctly: ESR is just a measure of how fast/slow a cap can store/supply electrons at specific frequencies, literally the resistance?
replace anything that's obviously failed. Leaking, bulged, high ESR, low C, that sort of thing.
-> Then the question is... What is "high" ESR, how low C (you often don't even know what capacitance tolerances the cap had when it came out of the factory. It might have been -20/+50% and have the same capacity as when it was new. I know that "high" ESR depends on the case, which makes it kind of difficult to judge without experience. The new caps had
@T3sl4co1l: Thanks for the tip on the MOV. We do have a megger at work (we work with equipment that goes under water), so I'll take a look. I take it the exact voltages are not really critical (as I don't have a datasheet of the original component). I'll take a look at it just for the learning experience.
At low to medium frequencies (typically including mains frequency), capacitance dominates, with a fairly constant Q factor. Since reactance is inversely proportional to frequency (Xc = 1 / (2 pi F C)), and ESR is a constant fraction of that (ESR = Xc / Q), ESR is also varying with frequency.
I don't get that right now, but will further look in how Q is defined. I've looked into that a couple of times but it never really made sense for me (on a "oh, right..." level).
Although... Q is just a unitless thing that defines the "quality" at a given frequency, so then ESR is related to Q because that is simply how we define it. That would make sense... (if my assumption is correct)
This minimum ESR point (self or series resonance) is somewhat less than the ESR at low frequency -- hence datasheets often give one ESR figure for 100Hz, then a reduction factor for the 100kHz figure, maybe 1.2x or something like that.
This is the first time this plot falls into place for me. I'll need some more time to really let it soak in, but thanks!
So, as for replacements, since low ESR isn't very important, don't worry about it.
You mean ESR isn't really important in this specific place in the circuit (being bulk charge capacitors right behind the bridge rectifier)?
For anyone else looking for info:
* I found this whitepaper (https://www.cde.com/resources/technical-papers/AEappGuide.pdf) which has some good info.
* TRX LAB has a seemingly promising video on Youtube "#83 LCR's meter D, Q, Phi, ESR what is it all about demonstrated on DER EE DE-5000 (https://www.youtube.com/watch?v=ivVSq0IiZGo). Seems like that could be some good reference to learn from. However... That guy is just so incredibly slow, repeating himself over and over, demonstrating and proving things that turn out not to be of any relevance. I've watched half of it at 1.25 speed, which makes the rambling somewhat more bearable. I'm at about 45minutes now, where the information really seems to start coming. (Same goes for MrCarlssonsLab, seems like he has some great content, too bad it's spread in between an hour of rambling and side stories.) Can't complain of course as you're not (directly) paying for it :).
* W2AEW has some things on his youtube channel (https://www.youtube.com/user/w2aew) that are much more to the point. He really has that amazing gift of being able to explain things in a matter that just *click* (for me)
* And I just realized I have the ARRL manual, which sometimes manages to convey information in a way that works for me (They've had a few revisions to get it right :) ).
-
If I understand correctly: ESR is just a measure of how fast/slow a cap can store/supply electrons at specific frequencies, literally the resistance?
Yes, that's pretty close. Most often, it's not being dis/charged very far, and just some small ripple voltage is present; in that case, we'd like to know how much voltage for a given ripple current, so need to know the impedance of the capacitor. The resistance part of which is, well, effectively in series, hence ESR. :) At most frequencies, the capacitive reactance is minuscule, leaving ESR as the dominant part; obviously, reactance gets more important at low frequencies, which is why we're concerned about a bit of both at mains frequencies. But at switching frequencies, basically ESR.
Handy trick, well, not really a trick, more an observation -- if ESR * C is the time constant at which the part, by itself, can be discharged, then that's also the rate at which you extract zero energy from it. Because, it's dissipating it all internally, and if you connect a load, you necessarily increase the resistance of the circuit, making it go slower. And the slower you go, the smaller the fraction ESR is, out of the total resistance. So the ratio of resistances also gives the efficiency.
That's most important if you're doing deep discharges. Like power supply hold-up, or supercap energy storage, or batteries in general (which can be modeled as nonlinear capacitors: capacitance goes up with charge, so it gets harder and harder to raise the voltage, the more charge you put into it -- eventually flattening out to give the nominal terminal voltage at full charge).
Note that, even at a fairly slow rate, significant heating can still occur in the component -- it only takes say a few watts to overheat a small capacitor, and if you're drawing 100W on average (say a 10J pulse, 10 times a second), and the Q factor is about 10, that's 10 watts dissipated in the component.
So we tend to use electrolytics at relatively small ripple voltages, so that the amount of energy being used is quite small in relation to the total energy stored.
In the extreme -- for applications such as power supply filtering, bypass, etc., we're just using them for the fact that they give a usefully low impedance between the terminals (at AC), while drawing hardly any current (at DC). Indeed in such applications, we hardly even care that the part stores more energy than that tiny increment we're using; we just don't have any other solution to avoid also charging it up to that nominal voltage. (If we had something with effectively high capacitance under bias, and low near zero, we'd be set -- which, sounds suspiciously like a battery, if you've been paying attention! -- but alas, batteries have impedances much too high to be generally useful in circuits this way. There are, however, some specialty ceramic capacitors (specifically: poled ferroelectrics) that have this effect!)
Or put yet another way: whereas we might only need so-and-so capacitance to store the energy used during a cycle (in a bypass application), or such-and-such [additional] capacitance to also keep the voltage ripple low; we might still not have low enough ESR in an electrolytic of that value, to keep total voltage low enough; so we brute-force it by simply using excessively large values. Because electrolytics are so much cheaper and denser per uF than most types, this is often worthwhile!
I don't get that right now, but will further look in how Q is defined. I've looked into that a couple of times but it never really made sense for me (on a "oh, right..." level).
Although... Q is just a unitless thing that defines the "quality" at a given frequency, so then ESR is related to Q because that is simply how we define it. That would make sense... (if my assumption is correct)
Right, Q == Xc / ESR. Or for electrolytics, you often see "tan δ", which is essentially the reciprocal. (δ being the phase angle of the impedance, so tan gives the ratio of resistance to reactance; this can be a more accurate representation for high-loss components like these, which I guess is why they traditionally use this rating.)
And, it would be nice to think of a capacitor as some simple lumped equivalent, like a fixed resistor in series with a fixed capacitor; but that would grossly overestimate the Q at low frequencies, because ESR would be constant, while Xc is inversely proportional to frequency. So Q would rise arbitrarily, as frequency goes down. In reality, it levels off to about 1/tan δ, then falls again as leakage current takes over (in the mHz, that's milli with a small 'm'!).
I'm not sure what a good mechanical analogy is, because I'm not sure that there's any good and intuitive experience that's also frequency domain, and that's unambiguously not just viscous damping (~fixed resistance). For sure, there are fluids with some weird properties, like, consider a nice bowl of tomato soup: you stir it around, it seems... soupy enough? But let it settle, gently twist the bowl, see how it moves; it's viscoelastic, it acts as a gel, a rigid body, at low shear rates. This... seems like a much stronger effect than what you get in a capacitor (except maybe some ceramic capacitors, when you look at them up close enough that hysteresis becomes apparent?), so I'm not saying it's an example of the effect, more just to say that, definitely, resistance can vary with frequency.
The most direct analogy is probably some types of rubber; often used for shock absorbers, rubber consists of an extended molecular structure with fluid-like domains interwoven with cross-linked molecular bonds. The bonds give it solid strength, but the circuitous paths between crosslinks give it lots of room to stretch and deform, while the dangly bits of the structure act almost like a liquid sloshing around inbetween. As a result, the Q factor tends to be modest -- certainly still enough that you get a good bouncy reaction from the stuff, but also not like a bouncy superball that returns like 90% of initial--- well, actually that's fine, 10% loss is the same as saying a Q of 10, right? Anyway, particularly the stuff used for damping, tends to have a modest Q and over a wide frequency range I suspect, so it should be a good representation of this. I'm just not sure how intuitive it is, that that particular characteristic might apply (~constant Q vs. frequency).
You mean ESR isn't really important in this specific place in the circuit (being bulk charge capacitors right behind the bridge rectifier)?
Right. And the other ones, pretty much anything you find will outperform the originals too (but also, likely not so thoroughly outperformed that resonances might be uncovered), so, nothing to worry about.
Cheers!
Tim
-
Or put yet another way: whereas we might only need so-and-so capacitance to store the energy used during a cycle (in a bypass application), or such-and-such [additional] capacitance to also keep the voltage ripple low; we might still not have low enough ESR in an electrolytic of that value, to keep total voltage low enough; so we brute-force it by simply using excessively large values. Because electrolytics are so much cheaper and denser per uF than most types, this is often worthwhile!
In that case, would it -in some cases- not be a better solution to put multiple (possibly cheaper) caps in parallel? As I suppose the total ESR as seen by the input/output follows the 1/ESR(tot)=1/ESR(1) + 1/ESR(2) +... Or is that too much simplification to still hold up to reality?
-
You've really managed to get some essential point across. Time to revisit some whitepapers/video's I've attempted to get through in the past :)
-
Yes, ESR drops when connected in parallel. If the capacitances are also different, you really ought to just throw it at a simulator and see how the currents share, and what the total impedance is; but absolutely, paralleling is a common strategy. And, when they're all the same type, there's nothing weird about the frequency response between them, they're all the same and the total impedance is just 1/N times an individual. Or in series, N times; etc.
I mean, first of all, a smaller electrolytic is nothing more than the guts of a bigger one, chopped shorter and rolled tighter. They're already in parallel, in the sense of being a continuous array, if you will. ;D
When using different types of capacitors, obviously their responses aren't any kind of similar so you can't just say it's divided up; applications range from, reducing the impedance rise due to ESL, or reducing ESR a bit, say with smaller ceramic bypass caps; to materially reducing the ripple current seen by the electrolytics (using enough ceramics to dominate over electrolytic ESR), thus increasing overall (power) capacity without having to spend for say 1000s of uF of ceramics.
Or likewise with film caps, at higher voltages (say, industrial inverters with 320V, etc. supplies), which have similar characteristics to ceramics but which are much cheaper in higher voltages and values.
Or using better types of electrolytic in the first place, i.e. aluminum (or other) polymer types. These are polarized, like electrolytics, but the electrolyte is solid (a conductive polymer) and can have much lower ESR. They're more expensive, so not usually used for bulk (1000s+ of uF). They're a common sight in computer parts, where the low impedance is necessary for the low voltage, high current regulators for CPUs, GPUs, RAM, etc.
Tim