Replacement electrolytic capacitor selection

[BradC]

I'm replacing the electros in a Tek 2440 scope PSU, but this seems to apply across the board, so I'm after some general guidelines.

Most of the electros can't be cross referenced, but other than the standard labelled Nichicon parts, the electros are all considerably larger than their proposed replacements (between 50-100% in volume).

I came across this in the extreme with the bus voltage caps in APC SmartUPS units. The caps in-place are some 4 times larger than the ones you'd normally replace them with if you were looking spec for spec, *except* I managed to find a genuine data sheet which indicates the ripple current rating on the original caps is some 4 times the smaller replacements.

On that basis, I'm *assuming* (dangerous I know) that the larger caps in this power supply have a considerably higher ripple current rating than the much smaller replacements. I can find allusions to newer versions of the same caps which indicate a much higher ripple rating, so it seems a sound theory.

So, I need an electro with the same diameter and lead spacing, and I figure either on or some 10s of percent higher capacity. The easy solution seems to be to up the voltage rating (a lot) and capacity (a bit). That seems to up the physical size, and ripple current rating while dropping the impedance.

So, for those that know. Is there any down-size to considerably de-rating a cap? Like using a 63V or 100V cap where a 6.5V was originally used?

Replacing a cap that has a ripple rating of ~3A with one that has a rating of ~1A doesn't seem like a sane thing to do for longevity.

[perdrix]

Your logic is sound ... Once you have capacity, ripple current rating is the key factor and if you end up with a higher voltage rating - who cares?

David

[floobydust]

I found modern electrolytic caps are on the low side for value, Nichicon or Chemi-con around -10% where the old Tek parts are the opposite +10%, so I upsize most of the parts to the next value (in the power supply) i.e. 82uF I used 100uF. Some older Tek scopes were lean on the capacitance in the first place, you can compare the design of newer revisions to the model. I think the PSU is the most important place to start.

For a given physical size, larger electrolytic capacitors have a higher ripple-current rating. So going up in voltage (size) also gains ripple-current ability but after some point ESR starts to go up so you have to look at that as well. I also consider lead spacings and use higher voltage parts so they fit properly.

Spreadsheet with 2235, 2235a recap size and replacements might help.

[BradC]

I started out with the Digikey web site, specifying the diameter, lead spacing, capacity through to "capacity + a couple of values", minimum voltage through to something sane and then just sorting on ripple current. Then I could use the ESR to figure out which looked "best" for the application.

I found the RS web site easier to use to do that sort of sorting, then I picked a couple of readily available cap series (like Panasonic FS & FR for example) and just worked my way through the data sheet. Comparing a couple of data sheets made it a bit easier to narrow down some "likely suspects" to then check availability on.

Even better when the data sheet lists dimensions vs ripple current for various voltages. Can just start the selection process on physical sizes / lead spacings.

An interesting process and looks like it got me a lot closer to viable replacements for the Sprague 672Ds that came out. Can't find legit data sheets for the 90's versions in the scope, but when comparing to the modern versions by dimension it looks like I got the ESR and ripple current "close enough" while bumping the capacitance up a size or two and greatly exceeding the original voltage rating : https://www.vishay.com/docs/42054/672d.pdf

I'm sure it'll go, and given the amount of use it sees, the caps will probably be the last thing to cark it. In fact it probably would have been ok with the original set of caps I ordered, which were based on capacity & voltage, and were ~33% of the original sizes and ripple current specs. I'll put them into "service stock". The reality is the thing has a crack over 1000h on the clock, if I wasn't worried about cap leakage I'd have left them all in place. They all measured out in capacity and ESR as perfectly fine.

Your logic is sound ... Once you have capacity, ripple current rating is the key factor and if you end up with a higher voltage rating - who cares?

Well, the reason I originally asked was uncertainty around long term severe electrolytic de-rating and whether there was anything to be concerned about. A bit like unused caps needing reforming, would running an electro at less than 10% of its rated voltage have any detrimental effects? My gut and experience says no, but there are much more experienced heads here so I thought I'd avail myself of that wisdom. I know enough to know what I don't know.

[David Hess]

There is no disadvantage to increasing the voltage rating to increase ripple current rating however ESR jumps upward at about 160 volts.  I suspect this is because a different electrolyte is used.

Input capacitors are sized for hold-up time during the power line cycle so have a much higher ripple current rating than needed and smaller parts with the same capacitance can be used.  Bulk decoupling capacitors are also not limited by ripple current.  Ripple current is most important in the output capacitors, and in the case of the 2440 design, these are the capacitors at the output of the current driven inverter.

Besides raising the voltage to get a higher ripple current rating, I would also considering at least doubling the capacitance.  Like the voltage rating, the value of capacitance is not that critical above a minimum value.

[BradC]

Finally completed this job.

Replaced all the 100uF electros with the best equivalents I could find. Replaced all the 220s and one 330 with 1000uF and the remainder of the 330, 1000 & 1500 with 1500uF caps. I basically went for the highest ripple rating that would fit. Only down side is the scope takes a bit longer to power up.

Finally got the RIFA stink out. Replaced all 4 RIFAs, the 30 Ohm flameproof resistor that the smoking RIFA damaged, the line filter and both NVRAMs. Should be good for another 20 years or so given it's only got ~1100 hours on the clock.