Bad Caps in a Linear Power Supply

[Crystaleyes]

I've been looking online but not found anything definitive, yet I was looking for info regarding how failed or failing filter caps change the output voltage. Both with and without load?

Is it correct that with bad caps, a PSU voltage might test ok without load, then drop 10% once under load?

[stj]

without a load or maybe even with a load it could look o.k.
but if the load is not constant the output wont be stable

[Crystaleyes]

Ok.

So a heavier load wouldn't necessarily cause more of a voltdop than a fluctuating load would?

And 10% is a reasonable 'possible' ball-park figure?

12v down to 11 - and 50v down to 40 or 45?

[redwire]

The (linear) PSU sitting there with no load, if a filter cap is say 5% of it's rated value say 1,000uF is 50uF the voltages can look OK but it all falls apart when there is load applied.
With increasing load, you get more mains ripple if the filter capacitors are failing low value. This makes the output voltage sag as it becomes filled with 2xmains ripple. The output is also full of hum.
Bad or weak pass transistors can look like the same problem - the output voltage sags under load. But there will be no pulsations (hum) or ripple.

[redwire]

A voltage sag of 10% is pretty crappy, it's not regulating properly or the load current is too much.
When you say "fluctuating" load, how much current is it?

[Crystaleyes]

Haven't measured the current draw but it would be a few hundred milliamps.

400? 500?

I should probably measure the current draw eh, before asking such questions.

[eccerr0r]

IMHO you need an oscilloscope or speaker to see/hear problems with bad filter caps on linear PSUs in general. Behavior of downstream circuitry will be erratic but should "do" something, unless the peaks end up frying them...

BTW there are some confusion between "linear" psus specifically distinguishing "line frequency unregulated" vs "linear regulating" PSUs. Without a regulator, who knows what the voltage output will be as it will change under load, and linear regulated PSUs may get quite warm under load.

[Crystaleyes]

I was asking this primarily to understand more regarding which kind of symptoms failing filter caps demonstrate, specifically in linear power supplies. I'm sure that they are equally annoying on both regulated and unregulated supplies.

It's all very well replacing shorted or open condensors, but without an ESR tester, or scope to see any AC ripple, sometimes caps look and test ok, so recognising the symptoms is invaluable.

Unfortunately, such knowledge is down to experience, which I don't have too much of, but I can still ask when these questions pop into my head.

[stj]

put a multimeter on the output and set it to AC volts - it should show ripple.
or set it to frequency and see if it can detect 60 or 120Hz
but be carefull with that trick because the meter cables can act as inductive antenna and pick up mains from nearby cables or transformers

[Crystaleyes]

Originally Posted by stj

put a multimeter on the output and set it to AC volts - it should show ripple.

The question is, "how much ripple is acceptable?"

I've measured various filter caps this way and - for example - I have sometimes seen a steady 0.000v AC. Another time there might be different values all the way up to flashes of over 400v, followed by zero, then some other number, then zero, then 400+v etc. and this just keeps recurring.

Obviously no ripple is desirable, yet surely there must be some percentage value above the rated cap voltage, where the ripple level is just unacceptable?

As for doing a frequency test to see the mains 60Hz, I have never thought about that. I'll give it a go.

Thanks

[petehall347]

some meters wont detect AC properly so use a coupling cap in series with a meter probe to be sure .

[Crystaleyes]

Originally Posted by petehall347

some meters wont detect AC properly so use a coupling cap in series with a meter probe to be sure .

Ok.

So, as for the values of coupling capacitors, If for example I want to measure a 220µF condensor, would a 47µF coupler be about right?
Or if it were 1000µF, could it be a 100 or 220µF?

Just to get a rough idea...

?

[eccerr0r]

electrically ripple is impossible to remove without a regulator (and even then...), though it can be made "small enough".

It all depends on the downstream circuitry whether ripple is acceptable or not. If it's a regulator, sure it was meant to remove ripple and lots is fine. A lot of the times, dc motors do fine with ripple. But if it's logic chips, not so much.

[eccerr0r]

A fixed value is sufficient, value is due to the fact you're measuring line frequency ripple. Even 47µF is way bigger than you need, only need like 0.1 to 1 µF or something like that to block the DC component and still let the ripple to get through for most cases.

Yes, the value also depends on your meter.

[redwire]

Most digital multimeters have 10MEGΩ input impedance, so a much smaller coupling capacitor can be used.
Fluke, Brymen etc. use 10nF (film cap 1kV) which is good to around 10Hz (-0.1dB or 1.1% down).
I use junkbox 0.1uF film cap but careful to discharge them afterwards if I am measuring ripple on HV power supplies like 500VDC. They can make a big spark.

[eccerr0r]

I've seen some meters with worse AC input impedance than DC for whatever reason and cheap meters with only 1MΩ input at DC, so one might have more of a measurement error with small caps and assuming 10MΩ input impedance.

No problem with the 1GΩ meters however... but I don't think I've ever seen a meter with a 1GΩ input impedance in AC modes.