Redundant DC power supplies

[Ken Moore 23]

I have a somewhat critical application, that has a lot of analog I/O (around 300 points). This past week-end one of the DC power supplies bit the dust. It was a Sola-HeavyDuty SDN-20-24-100P, only about 8 months old. (BTW this is the first time I've had a Sola die on me.) Anyway, looking at the Sola web page, they sell a unit that allows the use of two power supplies for redundantcy, it's a SDN 2.5-20 RED, anyone ever use one of these? I've pretty much convinced myself to buy one, and I'm looking for any bad or good experience with this arrangement.

[Chris Elston 79]

Generally for redundant supplies, you'll be putting them in parallel, which will double your amperage capacity. If your going for failure, make sure your single power supply can handle the total amperage load if one fails. Example. If you have PS1 = 10A and PS2 = 10A, you could supply 20A of power to your system. If your system draws 18A, and you lose one power supply, guess how much draw that 18A is going to have on your last power supply? It will probablly kill that power supply right away also. My advice is to be sure your singles can supply the total load. Also use SOLA's "ok" contact on the power supply and monitor both of them. Or if you have an analog input, use voltage feedback in conjunction with a voltage divider to bring your 24V down to 10V. This is so you know that your "other" power supply died, and your flying solo....

[Ken Moore 23]

I'm currently using three 20 amp supplies, which is more than enough for the application. I plan to buy 3 additional supplies and the Redundantcy module. So if one fails, it shouldn't be an issue. I have to look at the specifications on the "ok" contact, I assume it is just a contact, if it is I plan to connect all the power supplies in series and run them to a panel light that says something like "DC Power OK, if anyone of the supplies fails, the light will go out and I will know to check them, the supplies are all located in the same panel, so it will only take a second to find out which one is "bad". Thanks for the feedback. Ken

[ianbuckley 0]

This power supply (as well as most commercially available anymore) has an overcurrent feature that reduces the output voltage when overcurrent is detected. This feature should prevent the second supply from frying itself if it is given the full load. That doesn't mean it will supply the load correctly as intended, but one supply going bad shouldn't kill the other as well. On a side note, depending on how sensitive the analog I/O is, you should at least consider using linear supplies. The efficiencies are much less and they take up more space, but they have much better regulation and don't introduce electrical noise the way a switcher does. If the last few bits don't make that much difference, then it probably doesn't matter, but it should at least be considered. Your "ok" contact feedback scheme sounds like it should work fine.

[jstolaruk 9]

If you're going to monitor the status of the supplies, you might as well use a contacter to choose which one is actually supplying the I/O with power. Switch in the backup when the primary dies. This would be consistant with the way I've seen it done plus you lower the risk of knocking out the 2nd supply if the circuit is what caused the first one to fail.

[Ken Moore 23]

In this application, we use good wiring practice and all shielded cables, noise has not been a factor. I have pretty much gotten away from using the linear supplies because of the reasons you mentioned.

[jstolaruk 9]

I was thinking of lightening or some other power source getting tangled up with it. But I guess you have to consider the risk of it actually happening. My customers? I wouldn't put it past their electricians, they're capable of anything!

[Ken Moore 23]

The outlet power from these units go a bank of about 20 fuses, I hope one of these small fuses blows first, plus the line voltage to these units is from an industrial 12kV UPS, which should take of spikes etc....

[DanW 41]

When I first used switchers, I was concerned about analog noise, but haven't ever had a complaint. Are small linears, 15W or 30W units, even available in DIN rail mount packages? I remember the inital appeal of switchers for small loads was the conveniece of packaging, like you mention - smaller, DIN rail screw terminal, not the open frame, solder terminal. And higher efficiency for less heat dissipation in small panels. Dan

[ianbuckley 0]

As for noise, as I mentioned before, it depends totally how sensitive the analog I/O is - especially concerning how closely you are monitoring it. The vast majority of analog applications are not that critical, so it won't matter. It will only come up if you are using the full resolution of the ADC and monitoring at a fast rate. Most process applications don't require a fast rate of monitoring as things generally don't change very quickly. The noise only becomes an issue when you start monitoring changes that occur in small slices of time (5ms or less) and every bit is critical. The other possibility for a problem is if you are amplifying a small signal to get into the ADC range, because the noise is now larger in relation to the signal. In any case, these issues won't apply to the vast majority of applications. Just thought I'd mention it. You might still be able to get a small linear as a DIN mount, but it would have to be small and you might have to dig a bit to find one in a catalog.

[catserveng 0]

We also have had some problems with SOLA units lately. Started using Phoenix Quint power supplies with buffer and DC UPS options, had good luck with those units so far.