PDA

View Full Version : OT low esr capacitor source in canada



darryl
07-22-2011, 12:39 AM
Looking for a canadian supplier for switching power supply caps. Not much luck so far. I have not done phone calling yet to my usual suppliers, but their online catalogs don't show parts like these, just regular electrolytics. Anyone know of a source in Canada?

Black_Moons
07-22-2011, 12:46 AM
www.digikey.com (or www.digikey.ca ) are great sites for electronics parts, and have VERY cheap shiping to canada, with insane fast delivery times.
Free shiping on orders over $200 too iirc.

Evan
07-22-2011, 02:07 AM
What do you need and how many?

darryl
07-22-2011, 04:15 AM
Evan, I'm going to hack a few computer power supplies to bring the -12 v output up to the same current capability as the + 12 v output. This will give me + - 12 v, or 24 v at 15 or so amps. I don't need any high current 5 volt outputs, and if I need two polarities at 5 volts for a breadboard to experiment with, I'll use a pair of 5 v regulators.

It looks like the +12v output has a single 1000 uf cap on it, which my brain says is kind of low for any real current output over a length of time. I'm thinking to boost that to 2000 uf, then I need to add 2000 uf to the - 12v circuit. Times two for two power supplies, it looks like I'll need at least 6 1000 uf caps at 16v. Size won't matter since I'll make my own pc board to mount the parts.

As you know, there's no room in there to do much modifying, so my plan is to pull most of the output-related parts off from the secondary side of the power supply. The feedback trace to the TL494 will be re-fed through a voltage divider from the +12 v output, which will keep the regulation happening, and probably better on the 12 v outputs. With all of the unneeded parts removed, I might get room inside for a larger rectifier heat sink, which should be the major heat-producing part in there. It won't bother me if it's larger and runs cooler because of that. The capacitors will mount on the secondary pc board which will also carry an output terminal block of some kind.

I'm thinking to re-wind the inductor that's in the 12v output circuit with heavier wire, but still the same number of turns. I can mount it on the output pc board with the capacitors and output terminals. I'll be getting rid of the rat nest of individual wires-

Essentially I'll be using the primary, drive, and feedback circuitry as-is, but re-building the secondary circuitry to suit my requirements.

I'm sure enough that this mod will work, so I want to also modify a higher output power supply later on. In total I'll need about a dozen or so caps.

I looked at pulling some off an old mb I have, but they are only 6.3 v rated, and anyway I'm not into being the guinea pig to see if the caps are going to live or not.

Evan
07-22-2011, 06:15 AM
I will have a look at my stock later. I have a lot of caps from all sorts of power supplies.

ikdor
07-22-2011, 07:05 AM
Darryl,

If you just need a powerful 24V or powerful +/-12V it is possible to wire two computer power supplies in series.
In general the GND and the earth are tied together, but if you remove this connection the output is floating and can be used in series.
On some power supplies this GND link is just one connection, on others it's more elaborate.
If you google around you can find some examples of people putting them in series.

This would seem simpler than rewinding the transformer on one.

Igor

Black_Moons
07-22-2011, 11:28 AM
You don't need to rewind the output transformer since its allready centertaped.

You will need to add more rectifyers and rewind the output inductor, it does the voltage matching beween outputs, And the -12v is just a parsistic winding off that inductor, so a new heavy duty winding, and new half bridge will be required to make a high current -12v.

darryl
07-22-2011, 08:37 PM
From the power supply schematic, the output transformer secondary is center-tapped, and the two remaining leads are fed to a dual rectifier, a common cathode. The output from that goes through a funny-looking inductor and on to the capacitors. This gives the + 12 v output. The - 12 v output is taken via a common anode dual rectifier, fed from the same two points on the transformer. The only difference is the rectifier for the negative output is smaller in capacity, and there's another rectifier in series before the voltage makes it to the capacitor. If that dual rectifier was as capable current-wise as the other one, the -12 v output could be rated at the same current level as the + 12 v output. Oh, and I'd have to remove the single rectifier that is in series- I think all it does is reduce the voltage on the - output, since very little current is normally drawn from there in computer use.

The primary circuit drives the transformer in push-pull, and current is delivered from both alternations. The primary won't know if the negative 12 v output has been upgraded, current wise. What it will know is how much power is being drawn in total. Basically my planned mods will be invisible to the circuitry, and of course it will be up to me to keep the total power draw within the capability stated on the case.

The torroidal inductor in there is used in a way I'm not familiar with. Directly after the rectifiers is a winding through this torroid core, then it's on to another inductor and the cap. Each rectifier, four in this case, feeds its own winding through the same torroid. In other words, there's four separate coils wound on one core. I'm not sure what purpose this serves, but possibly it's more than just giving some inductance in series from the rectifiers. At any rate, it's these windings that I would consider replacing with a heavier gauge, just to keep some of the losses down. I will in no way be opening up the power transformer to mess with windings. These things are pretty well glued up tight- all that would happen if you tried to get into it is it would break.

J Tiers
07-22-2011, 09:38 PM
The torroidal inductor in there is used in a way I'm not familiar with. Directly after the rectifiers is a winding through this torroid core, then it's on to another inductor and the cap. Each rectifier, four in this case, feeds its own winding through the same torroid. In other words, there's four separate coils wound on one core. I'm not sure what purpose this serves, but possibly it's more than just giving some inductance in series from the rectifiers. At any rate, it's these windings that I would consider replacing with a heavier gauge, just to keep some of the losses down. I will in no way be opening up the power transformer to mess with windings. These things are pretty well glued up tight- all that would happen if you tried to get into it is it would break.

That inductor is wound with a proportional winding from each output to improve "cross-regulation", meaning that a load on one is reflected more equally in others.

Otherwise, a load on one that has the feedback from it may cause others to go drastically upwards in voltage, as the PWM duty cycle varies upwards to hold the load.

It is usually the main inductor, and the others are more like "hash chokes".

If you mess with it, and want the other outputs, keep the windings at same proportional turns ratio. (otherwise delete the other rectifiers). However, more current may end up saturating it, which will not be what you want. PC power supplies CAN be hacked, but they usually are pretty optimized designs, so any changes need to be compensated for. Removing the other unused windings gives more winding window, and also more ampere-turns left for the one(s) you want. But the issue of inductance is still there, the inductance needs to be "designed" for the supply to work well.

Some PC supplies are more advanced types, not simple "forward converters", but high efficiency quasi-resonant types, etc, which are only hackable with expert knowledge.

darryl
07-23-2011, 02:09 AM
A couple of points- first, I don't want to wire two power supplies together for 24 volts when one will produce it, though I know it could work. Secondly, if I re-wire the output inductor I wouldn't be altering the number of turns on it, just increasing the size of the wire (or winding it bi or tri-filar). I think in this case I will be eliminating the 5 v output and the -5 v output, so I would only need two windings through the inductor. In addition to making the gauge larger, I would wind the same number of turns through it for the -12 v output that already exist for the +12 v output.

Referring to my schematics, some have the feedback coming from the +5 output, others have it coming from the +12 output. I will have to physically follow the feedback trace in the power supplies I'm using to make sure I don't leave it open circuit when I pull the parts out that won't be needed.

Here's something I didn't notice before- one of the schematics shows some very low capacitance values on the outputs, like 1 uf. They must be relying on the motherboard to provide sufficient filtering. The ATX 200 is like this.

J Tiers
07-23-2011, 09:32 AM
Eliminating the 5V will save lots of ampere-turns on the inductor, since the 5V or a lower voltage is usually the highest current output.

But the 5V MAY be a primary regulating output, as you say...... "somebody has to be the boss". It will be hard to move that to a different winding unless you know some things, and do some "design". Or you may get lucky.

And, some supplies COMBINE two feedbacks, making it difficult to separate them

Black_Moons
07-23-2011, 10:49 AM
Iv never seen a supply with low output capacitance on anything but the -12v and -5v lines, it just would'nt work without localized capacitance on the +5v/3.3v and 12v lines, the wires have too much resistance to the motherboard to use the motherboard. (Not to mention hard drives and such have next to no onboard filtering, and motherboards only have enough filtering for there own needs and would likey pop if they had to filter psu ripple too)

That said, Iv never seen -12v generated from the transformer itself, its allways a parasitic winding on the common inductor (ie, ground->inductor coil->diode->cap, it uses voltage induced from the buck cycle of the other supplys to actualy produce -12v) And then -5v is generated from a 7905 off -12v (making -5v by far the best regulated, lowest noise supply of a ATX psu!)

Often ATX supplys will have a compairitor or two that will act as an overvoltage cutout, and sometimes overcurrent cutout (they usally have some method of overcurrent cutout, however its more for protecting from shorts/sudden huge loads then overheating from overcurrent), Not sure if any of them have undervoltage cutouts, However a GOOD psu should'nt put powergood high untill all voltages are up to spec (Many bad psus tie power good to +5v)

And yes, psus do vary a little from power supply to power supply, however most fall into a few catagorys of design for the 150~400 watt range. Bigger psus often start using a diffrent design that iv not really looked into because I quit PC repair and stoped taking psus apart before over 400W psus became commonish.

And yes, +5v will be the master regulated output, With 3.3v regulated by its own TO-92 transistor+another inductor+Black Magic (Honest! 30A at 3.3v is regulated by a TO-92!) and some extra circuity. the +12v is regulated just by the inductor since nothing that uses 12v is gonna care much if its 11v or 13v. (the inductor acts as a transformer, Except sideways... Basicly, Every coil to ground will produce the same 'ratio' voltage as its turn ratio during the 'buck' stage (Input transistors off, Current flowing from ground (through off transformer) through diodes (not important) to inductor to output.)

Note very little current flows to +12v during the 'on' stage of the input transistors, Because the transformer is actualy 20:10:0:10:20 and the 20v taps feed the 12v, so the ratio doesnt work out to allow much current flow during the on stage. Also results in a typical +10v to the 12v output when there is no 5v load to cause higher duty cycle and proper matching via the inductor.

darryl
07-23-2011, 05:39 PM
Well now from all you guys have said and from what I've read, I feel confident that this mod will work out fine. The TL494 ic has been common in car amplifiers for years, and I did spend some time studying it both in aid of servicing equipment and for use on my own. I have an application guide on it, and I don't think I'm going to have a problem maintaining the output voltage properly through the use of the feedback terminal.

I did find some old vcr power supplies that have switching rated capacitors in them. Have not checked the values, but there may be some usable caps I can rob. They sure are small physically- it's hard to imagine them handling much ripple current, but then depending on the waveform feeding the output circuitry, they may only need to deliver short spikes of current between pulses from the rectifiers. The average power dissipation in the caps is likely to be quite low- maybe I don't need to worry about having enough capacitance.

It's hot here today, much nicer to be in the shop. Maybe I'll start to strip parts and get this phase of the project in gear.

darryl
07-23-2011, 09:00 PM
You all have tolerated this OT thread very well. Some of you are into electronics in some ways and probably don't object- others, well, it doesn't hurt to learn some things about electronics- some of it is useful in the metalworking field.

Just wanted to pose a question to a few in the biz- I've got one of my power supplys unpopulated of parts that won't be needed in the finished version. One of those parts is a dual 30amp, 45 volt rectifier. It was in the 5 volt circuit- I want to use it in the 12 volt circuit. Am I going to get in trouble with the voltage rating? I'm thinking no, because the reverse voltage it will see is double the output voltage plus perhaps 20% for spikes, so about 30 volts. Because the transformer is never unloaded for either polarity of voltage swing, there shouldn't be a major spike being generated. I'm concerned though that the addition of the output inductor would allow higher spikes to occur on the rectifier side. Does this make sense?

I can always 'scope the secondary side of things to see what the waveforms look like, using an unmodified power supply.

J Tiers
07-23-2011, 10:33 PM
Switching supplies are unlike others, the voltage requirements may be lots higher than you think.

The power supply transformer turns ratio is almost certainly a lot different than it seems like it should be... because the effective ratio is an average, and the thing regulates over a wide range by varying the pulse width.

Translated, that means that the supply applies higher voltage pulses to the inductor, which (with the capacitor) averages them. The 12V winding may get 20V or more applied to the inductor depending on the design, and mains voltage, etc.. But the pulses are short, and the average is 12V.

Reverse voltage pulses of similar voltage will probably occur, and it is wise to allow at least 30% of the total extra for spikes.

When switch is "off", with significant current flow (continuous conduction mode), the input side of the inductor will be just below ground. but at low current it may not be, it may have "run dry" and be "ringing" at some intermediate voltage.

So you need to figure that you may need 20V, plus 20V, plus 30%. Or more. Depends on the design.

So..... 45V might work, but it looks too close for me.

darryl
07-24-2011, 01:33 AM
Thanks, JT. I think I'm going to 'scope an unmodified psu to see what the waveforms look like. I'll do some load testing as well- some power resistors, some motors, etc for an inductive load. Chances are I'm going to use some higher voltage devices. There has to be something in my collection of dead car amps with some dual rectifiers, common anode and common cathode. Hmm, I had an amp in with about 10 output devices per side- that should have some usable parts, maybe some caps too. Just need to find it-

J Tiers
07-24-2011, 10:53 AM
The rectifiers will need to be fast recovery types.... how fast is needed depends on the switching frequency

Black_Moons
07-24-2011, 01:07 PM
45 is not enough no.

Consider the '12v' output of the transformers output 20v when given a nominal 120v AC input.

Thats +20v on one side, and -20v on the other side. 40V allready.

Next we have some flyback leakage that will easly cause spikes (And general ringing of.. everything in a high speed high power circuit)
Next you have to surive voltages as high as 140v input, or another 20%+

So you end up with at LEAST 40V+20% or 48V during just normal operation, before counting ringing. You'll probley find the diodes on the 12v side where rated at least 60v. Maybe even 80V. Id go with 80V to be on the safe side personaly. 60v would be absolute lowest id use.

And they definately need to be faster recovery, or schottky (most smps use schottky), schottky are ALLWAYS fast, but leak a fare bit in reverse bias mode. 'Normal' speed diodes will go up in smoke fast in a SMPS.

J Tiers
07-24-2011, 01:21 PM
45 is not enough no.

Consider the '12v' output of the transformers output 20v when given a nominal 120v AC input.

Thats +20v on one side, and -20v on the other side. 40V allready.

Next we have some flyback leakage that will easly cause spikes (And general ringing of.. everything in a high speed high power circuit)
Next you have to surive voltages as high as 140v input, or another 20%+

S


not sure where the 140V comes from........

And with a "forward" converter, there is no guaranteed "-V" of the same magnitude..... it isn't like a regular line freq application.

There is a power pulse, and there is a "volt-second recovery" and "leakage" spike, the magnitudes of which is a design factor in both forward and flyback supplies. IIRC most computer PSUs are "forward" types, although with higher power there may be some half bridge types, quasi-resonant, etc.

The bulk of Schottky diodes are relatively low voltage, 100V max, although specialty ones can be had in considerably higher voltages

darryl
07-24-2011, 04:50 PM
The existing dual cathode rectifier in the +12v side is a 60 volt rated part. I'll be using four single rectifiers rated at 100 volts, so there will be a bit of heat spreading, and they are fast devices.

Aside from my schematics, I've physically traced the feedback lines, and it's coming from both the 12v and the 5v outputs. With a bit of math I'll be able to change the resistor in the 12v line and remove the one in the 5v line. The voltage at the feedback pin will end up being the same as it would with both voltages feeding it through the existing resistors. On this same note, I can bring out a wire from each side of this resistor and put a pot on it. As I dial the pot down and increase the feedback, the output voltage will get lower. I will have a minimum load on it at all times as well, since I know at some point it could go wild without that. It would be nice to be able to vary the voltage without having to go with a series regulator.

This will be instructional for me as well, and I'll see what the circuit is doing by 'scoping it. This could end up being more useful than I originally thought.

I have stripped the output inductor and counted the turns. Now I will figure out if I can wind on 4 bifilar windings of the same gauge wire, which currently is #18. If that won't quite fit, I'll wind on two #16s. According to the AWG, there's a bit more meat on two 18s than there is on one 16, so if I can do it I'll go with four #18s- each will be a bit easier to wrap tightly on the core.

J Tiers
07-24-2011, 11:07 PM
There are some details that affect how adjustable the voltage is, the feedback on that side may, or may not, include components which "compensate" or adjust the frequency response of the feedback for proper operation.

Those components MAY be on the DC side, OR they can be in the local feedback circuit of the control chip.

As for the wire, with 4 #18 you may find you have more options. The 4 wires can be wound flat and parallel, which may be the smallest practical volume, or "cross-section" per turn. While round things pack the same at any size, smaller ones make more different net cross-sectional shapes, and may be easier to fit in.

They also are easier to wind, being less stiff. And if you think you will be packing the windings tight, their flexibility is advantageous in reduced winding forces and less chance of chipping off the wire varnish as you manipulate the last few turns.

darryl
07-24-2011, 11:56 PM
Well, I rewound the inductor awhile ago. I used two runs of #16 instead of 4 #18s. It was a real bear getting all the turns on there, especially with both windings going on together. I cut two 5 ft lengths, then passed them both halfway through the core. I put a couple turns on from one side, then the other, etc, getting the inside of the turns as close to the core as I could. Once I got around once, the turns started overlapping and it was easier. At the end, I barely got the last wire through the core to complete the turns.

It's done now, but I would not recommend that anyone try this. It was very hard on the fingertips. To do it again- I would just go with two #18s and be done with it, still winding them both side by side for closest coupling. The four original windings were done this way, which I found out as I tried to unwind them one by one. I had to unwind them all together for the most part.

So, that part is done- on to the rectifiers and heat sink.