EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: matthuszagh on February 25, 2024, 06:59:30 am
-
I'm designing an AC/DC power supply with an isolation transformer, bridge rectifier, filter cap, and linear regulator. I'm concerned about the possible scenario in which the filter capacitor develops a high ESR and the DC ripple at the linear regulator gets worse. I'd like to avoid exposing the regulator to that and dealing with any of the possibly errant behavior that results downstream of it. Are there any common circuits for this purpose? Ideally it would be something that tripped a fuse or similar to make the problem immediately apparent. I considered a crowbar, but the peak voltage isn't much higher when the capacitor turns into a resistor than it is before. The much more obvious symptom is the undervoltage excursions.
If there's no simple solution to this it's not a big deal as I'll use a good filter cap with a long life and I'll probably place some overvoltage/undervoltage protection between the final regulator and load (the load is what I really care about saving). So if the capacitor dies and takes the regulators with it, but leaves the load ok, that's fine, I can replace those regulators.
-
> I'd like to avoid exposing the regulator to that and dealing with any of the possibly errant behavior that results downstream of it.
There should be some power supply supervision chips that can monitor the ripple of the output voltage and provide a POWER_GOOD signal (that you can then use to gate the output through eg a relay or P-fet). I expect this is easier if you are making a fixed voltage supply; for an adjustable one you might have to roll your own protection with a microcontroller (and provide it a separate, simpler power supply).
How sensitive are your output loads? How far can they tolerate voltages exceeding spec and for how long? Human safety threat level or minor inconvenience?
-
'Under Voltage Lock Out' (UVLO) or 'power supply supervisor' chips are two terms you might want to search on, plenty of options available. It sounds like you want to have a minimum voltage for operation and disable below that.
-
I'm designing an AC/DC power supply with an isolation transformer, bridge rectifier, filter cap, and linear regulator. I'm concerned about the possible scenario in which the filter capacitor develops a high ESR and the DC ripple at the linear regulator gets worse. I'd like to avoid exposing the regulator to that and dealing with any of the possibly errant behavior that results downstream of it. Are there any common circuits for this purpose? Ideally it would be something that tripped a fuse or similar to make the problem immediately apparent. I considered a crowbar, but the peak voltage isn't much higher when the capacitor turns into a resistor than it is before. The much more obvious symptom is the undervoltage excursions.
If there's no simple solution to this it's not a big deal as I'll use a good filter cap with a long life and I'll probably place some overvoltage/undervoltage protection between the final regulator and load (the load is what I really care about saving). So if the capacitor dies and takes the regulators with it, but leaves the load ok, that's fine, I can replace those regulators.
If you design your linear regulator so that it can handle the no-load peak rectified voltage (including an allowance for higher than normal mains voltage) it should survive quite happily and not apply excessive voltage to downstream circuitry even if the filter cap dries out completely and effectively goes open circuit.
Under such conditions you will have a problem with the linear regulator dropping out on the ripple minimums but you can install a standard supply voltage supervisor IC to provide an under-voltage indication.
-
You could use a TL431+passives+PNP for a power good signal (https://www.ti.com/lit/an/slva987a/slva987a.pdf). See attached and simulation (http://www.falstad.com/circuit/circuitjs.html?ctz=CQAgjCAMB0l3BWc0EDYAsAmVBmMAOfdA-MAdkxASRsioFMBaMMAKADdxURNJ9wwlXvzoRUopHSkpWAEy48+IDOEz9Ks+gDMAhgFcANgBc5AoUpWD1ITbsMn5Ktfxbnrt-cdNPCC4TxttTxMAc2V0Hl8VHEgIqVYAd3DVFzJuKyhWAGNwNJTc7n9pOEowHGgwPkx0agBOSEwERuooWEg2MPJCpS6eaqhMpN6MxvS1TIAnHjR80cUREHRoTBxWMLn-Df74pIRat2UZzdZiOnQ4PoicJsiXEAA-ABUAGXQ8KAAdAGdznm+wJiCb4wMjofBoTBpEh7cjAv5fTCAzDA6CoWppbCQHC1UjY1D4VBwyD-FFojHibG42r4wlfEFgiFQwgwsj-c7fe5GAxvAGyRgARwAdgB9ejyYlfHBqKTfaoXCUUeXfHDlFVq9Uawr-ZZIlEUHCENFoc5lXAIOHIhHQdC6r6MGDwWKQMj7Wq1dBpHA1Ikku0O+DnF2YN0e3DeunQfWG2rG9o4M1siWc7l4MUCkVpgCCYGWny+BAqpCLBBLRfhmGWjqrVZwtpgqBdKtiTXQ4mq6HQFu+S3O1erAOYlvtbWuDVqYHQ1I7HcIYB9+e+w7go+DE6n09npMbMRq1Tb08THK5PLTQtFfMzFcoEvwSp+dEtA6BEbIBsqqAQoIw+Dd4K7CLrSNKnOVAwAQfBeAQMo8HnOcX2AyBQPAyDoLgkE30Qz9WzBX9zXzdkvmTE8+TPMU817GV73hJ9LRBSEcDSGkVnjchaPLQCVmdL1TlQcRyHOHBYJRTjXw7d8+NBLE9XoxiCWY0CKEPQjj1TEiM3Fb5b3hBBAUEl9mPwMhwPaHEWAg-8aMXGACUwhpd3HAlQSEv1oBsj87KaByG07fSVUM4ywFMghHwIojVPTc9GEzGBJE0h9-ltYcyEQsFsXIN0GPBWl4slQDHXRU1Av2CD0V4-B52YFF8rIQq3TUYNZKsyMUvwNKXVqTK0G+cQjxTXkBQAB3PPMtMfRKQUw8dIVDChqlqf9a0HJriEkrACDIQzX3HPDKMqlyVvONbDM2tK8KSybBFBNJZsnbqkxU-r+SGsibzvW9KN+MalpfeMoLSZ1UEwMAP1bf9EW+4cvVfNEamB5L4wbZzIY9eN3SghssV41kfo-LoAaBkHaQne6+tPdS8zAyj1G+HTvuslhWyK6kEA9YhyrpajATg5gKkC47bxWbBXycjmJW5nMWBxDaBalBsGJ8+mJ1At1mdZgglLCx71Ki2L80dcicssu0c1iGqsWpGqNoNeaOctcHn3waA8HRObIHqRAyEBCVKLgx3nYoSc3fgT8vcXE2PUqPFLdan8NYesnhstTW01kYaxfgeFk75VOyK+lgUShmIGowcdJ24sGueEmINpYLFSDRyBqGBVgpnjdQZjb8B9gGJs4ESRYLjmTu5kgFuQGHjv8XH3gBkqeB+87qUzilaf4jCN5KCXxYV5Vbh4imMFRG7w-8lEeekg3rvKBPwLSgXqfb-Hh-xlHi+tIyTuRnvlxxjmDJX6oN3DIN8X6sCMFQDsV9AGlG7nQZgIBhwfnAsQYMH5DJ8RwAg+sRlfpvClEXQgmC56BDsF4A+79u4swiI-M+fddhAPGHsUoeRR5hCYTwCg0Dx6xAGKPKY7DISUAEUoWhfcAD2ygQALDeC0GeghXIDEoNwVYEjBBSNno0Vs4ByjVE-MDDq2I8CQlqK0eAEAHyqHHiAAACgAeQAOoAFEABK3wADitjbEABFWAGnAHQAAYuY8AigEEQGeGInQshvjOPoF8AAll8IwOhBRZHoKwIAA).
Not sure how to measure ESR in-circuit other than indirectly through temperature differential of the filter caps and ambient (slow reaction time). It's impractical to measure current differential before/after the caps. I think you're stuck with waiting until excessive ripple to appear and let the power good signal pick it up.
-
I might measure the minimum voltage across the regulator and if it falls below perhaps double the dropout voltage, then trigger the shut-off or alarm or whatever. Systems which monitor the supply voltage to trigger safe shutdown sometimes work this way.
-
I'I'm concerned about the possible scenario in which the filter capacitor develops a high ESR and the DC ripple at the linear regulator gets worse. I'd like to avoid exposing the regulator to that
If the ripple depth gets low enough for the regulator to drop out that should be detected by an under voltage lockout circuit as several other people have described. But beyond that, don't worry about it, this is what regulators are made for. The ripple rejection of even crappy linear regulators is practically infinite at these low frequencies.
-
I'm designing an AC/DC power supply with an isolation transformer, bridge rectifier, filter cap, and linear regulator. I'm concerned about the possible scenario in which the filter capacitor develops a high ESR and the DC ripple at the linear regulator gets worse. I'd like to avoid exposing the regulator to that and dealing with any of the possibly errant behavior that results downstream of it. Are there any common circuits for this purpose? Ideally it would be something that tripped a fuse or similar to make the problem immediately apparent. I considered a crowbar, but the peak voltage isn't much higher when the capacitor turns into a resistor than it is before. The much more obvious symptom is the undervoltage excursions.
If there's no simple solution to this it's not a big deal as I'll use a good filter cap with a long life and I'll probably place some overvoltage/undervoltage protection between the final regulator and load (the load is what I really care about saving). So if the capacitor dies and takes the regulators with it, but leaves the load ok, that's fine, I can replace those regulators.
Is the load especially sensitive ?
You can ensure the regulator is not killed by simply using one with enough headroom for peak charge, which you likely already have.
Voltage dips should not usually destroy anything, but certainly will have it cycling rapidly.
As you already plan, use a good cap, (or even two) and if you want to be paranoid, add a OVLO/UVLO as added insurance.
Addit: for voltage checking there are expanding families getting higher voltage operation.
see
https://www.nisshinbo-microdevices.co.jp/en/products/reset-voltage-detector/?inputMore-Vin_max=10.0 (https://www.nisshinbo-microdevices.co.jp/en/products/reset-voltage-detector/?inputMore-Vin_max=10.0)
-
What about using the ripple to drive a charge pump, and Zener clamping and low pass filtering the result so it doesn't trip on the power-up transient? You'll need a bleed resistor to set the Vout vs ripple transfer function, then into an overvoltage trip input of your PSU supervisor.
-
50Hz mains filter caps tend to have very long lifetimes. Sure they see some pulsing current, but they are physically big to take it and they hold a lot of electrolyte volume to make drying out less of a problem. So a good quality cap should last many decades.
The electrolytic capacitors that typically die are the ones on the output of switching PSUs, those have a much harder life.
But if you do want to detect it. The best way is to add a power supervisor chip (sometimes called a POR or reset controller...etc). Then you just create a divider from the pre regulated voltage down to what the threshold on the supervisor is, then take the output of the supervisor to drive the enable pin on the regulator. This way if the input voltage ever drops bellow below what is required the supervisor will disable the regulator for like 100 to 1000ms. Tho this might cause it to start cycling since when the load disappears the input voltage might rise up again, but the user will definitely know something is wrong. Tho mains power brownouts will trigger it just the same.
-
Big capacitors with screw terminals or snaps seem to usually fail not by drying out, but when the seal leaks allowing water and oxygen inside which then attacks the aluminum strip connecting the terminal until it breaks.
-
Thanks all for the very helpful comments. Based on that, my current plan is to use something like the attached image (from https://www.analog.com/en/resources/analog-dialogue/articles/driving-a-high-side-mosfet.html (https://www.analog.com/en/resources/analog-dialogue/articles/driving-a-high-side-mosfet.html)). I was looking at a different supervisor, like the TPS3762 but I still need to do more research there, and I still need to test things out with a sim. Not that it's particularly relevant to the original question I asked, but I'll probably add another supervisor between the final regulator and load to shut off the load in case of any issues. I'll probably use something very similar to what I'll use for the "ripple shutoff".
To add some clarity about the load, given the questions, I'm not actually sure how sensitive it is to damage from a regulator cycling on and off. It's an OCXO and has a wide voltage range (+18 to 30 V). But, it's expensive and hard to find nowadays so I'm willing to be overly cautious. The other supply sensitivity it has (which I'm more concerned with in general operation than in a fault condition), is that the output frequency does have some supply sensitivity. It's not a lot (<1e-11 for a 1% change in supply), but the whole purpose of this OCXO is to be very stable so I'll design the power supply to provide a very stable and temperature-insensitive rail.
-
Any decent linear regulator should be able to keep a output under control while the input is swinging wildly. The output just dips down when the input becomes too low to maintain regulation. Especially when you are talking about slow things like mains hum, even the ancient slow as molasses LDOs can follow that. Tho LDOs do have a tendency to just give max output if they blow up.
If you are worried about overvoltage just put a SCR crowbar circuit on it.
-
Any decent linear regulator should be able to keep a output under control while the input is swinging wildly. The output just dips down when the input becomes too low to maintain regulation. Especially when you are talking about slow things like mains hum, even the ancient slow as molasses LDOs can follow that. Tho LDOs do have a tendency to just give max output if they blow up.
I'm not worried about the regulator failing to regulate with excessive ripple above the minimum dropout voltage. I'm worried about ripple that gets bad enough that puts the input below the target output, causing the regulator to cycle on and off, or at least drop it's output alongside the ripple. I don't want that getting passed onto the load, hence the supervisor and high-side load switch to disconnect everything downstream of the filter cap in this case.
If you are worried about overvoltage just put a SCR crowbar circuit on it.
I'm not worried about overvoltage, I'm worried about undervoltage. I probably will put overvoltage protection in too, but just because the supervisor IC will probably provide that along with undervoltage, but I don't expect it to be necessary.
-
Usually over voltage is not part of the power supervisor chip.
The purpose is these chips is primarily to hold digital circuitry in reset during a period where the power supply rails are still ramping up. That way they don't do anything funny on power up and ensures they always get a clean reset signal on startup. They also save them from doing something funny during a brownout so that they are pulled into a predictable reset before they could do something unpredictable once the supply gets out of spec.
There is nothing magical about these supervisor chips. They are just a comparator, voltage reference and turn off delay circuit packed into one chip. You can build your own from those components, but using the single chip solution is just simpler and cheaper.
However when on marginal conditions such a supervisor chip might still cycle on and off. It might trigger on a dip, but then the ripple gets high enough for the chip to be happy again and after 100ms give a power OK signal, then the ripple swings down low enough and again makes it trip for a few seconds and then recover. You might also want to add a RS flipflop and a manual restart button if you absolutely must not power cycle your load.
-
Usually over voltage is not part of the power supervisor chip.
The purpose is these chips is primarily to hold digital circuitry in reset during a period where the power supply rails are still ramping up. That way they don't do anything funny on power up and ensures they always get a clean reset signal on startup. They also save them from doing something funny during a brownout so that they are pulled into a predictable reset before they could do something unpredictable once the supply gets out of spec.
There is nothing magical about these supervisor chips. They are just a comparator, voltage reference and turn off delay circuit packed into one chip. You can build your own from those components, but using the single chip solution is just simpler and cheaper.
However when on marginal conditions such a supervisor chip might still cycle on and off. It might trigger on a dip, but then the ripple gets high enough for the chip to be happy again and after 100ms give a power OK signal, then the ripple swings down low enough and again makes it trip for a few seconds and then recover. You might also want to add a RS flipflop and a manual restart button if you absolutely must not power cycle your load.
Well many of the ones I've looked at (eg TPS3762) have window detection, so I can use them for under and overvoltage. The input voltage range of that chip is also very wide (threshold range 2.7 to 60 V) so I think the ripple would have to get really bad for the supervisor IC iteslf to start cycling. The ripple should progress slowly, so I have trouble imagining this might happen. I'm getting one with a manual reset, which I'll leave unconnected, so it will just power everything off until I notice the problem (there will be an LED indicator on the front displaying power directly across the load). And yeah I suppose I could just build one, but these are cheap and they fit my voltage range nicely.