1
\$\begingroup\$

I have a device (metal cleaning/polishing machine) which needs to switch between AC and DC outputs for its different modes.

The input is coming from a 24 Vac transformer which needs to either be routed directly to the output, or routed through a bridge rectifier -> smoothing circuit -> output.

The difficulty is that it needs to be switched digitally using a PIC. So physical switches with high current/voltage ratings are out of the question.

Current circuit

Currently, I'm using two SPDT relays but they just aren't reliable, as the output peaks at around 24 Vac 50 A and around 34 VDC 35 A. It seems like trying to find relays which can handle switching both AC and DC loads on this scale is an impossible task, especially with current stock shortages across all components. I'm currently using these but have had a few failures with them. The relays are never switching under load, they only switch when the load is disconnected.

Solid-state relays could possibly be an option, but the price for relays with this rating is way too high.

Push-pull simulation

I was thinking about some sort of MOSFET configuration, with a push-pull for the AC signal (as shown above), and just a single MOS for the DC signal. The issue there is that I'm just not 100% on how to configure the hardware so that I can switch between the AC mode and the DC mode. The simulation above isn't accurate to the system in terms of timings, the relay switching will only happen when the mode needs to be changed by the user (probably once every few minutes).

Any ideas would be a great help. Cheers.

\$\endgroup\$
7
  • \$\begingroup\$ I agree, high power relays are expensive. I found a few on a UK site but they cost between £40 and £85. Another point that I noticed was that the 50A current rating of the contacts applied to AC current. It was only 7A when DC was being switched. The contact life was 100,000 cycles. You appear to be switching 10 times a second so even if the relay could keep up with that switching rate, the contacts would only be guaranteed to last 10,00 seconds, a little less than 3 hours. Looking at the relay you are using it appears to be very marginal and only fully specified for AC switching. \$\endgroup\$ Feb 2, 2022 at 14:58
  • \$\begingroup\$ Sorry, I might be misunderstanding what you are saying but where are you getting the switching 10 times a second from? I should've probably clarified in the post but the relays are being switched very sparsely by the PIC when the mode needs to be changed (using a different relay interface circuit putting 12vdc onto the coils of the relays above), so I don't think that the contacts should be failing from the contact switching life. Again, please correct me if I've just misunderstood. \$\endgroup\$
    – M_Stenton
    Feb 2, 2022 at 15:14
  • \$\begingroup\$ Your simulation shows switching every 100 ms, that would be where the 10 times a second comes from. If you're switching less frequently than this, do make note of that in the question. \$\endgroup\$
    – Hearth
    Feb 2, 2022 at 15:38
  • \$\begingroup\$ Ah I see, my bad. Will clarify in the question. \$\endgroup\$
    – M_Stenton
    Feb 2, 2022 at 15:46
  • \$\begingroup\$ Does the AC part have to be a sinusoidal wave? Would a square wave do? If so, you might consider some form of H-bridge drive controlled by the microprocessor. \$\endgroup\$ Feb 2, 2022 at 16:16

2 Answers 2

2
\$\begingroup\$

Maybe you could route the current with the cheap relays as now, but switch the primary to the 24V transformer with a mains-rated SSR, and sequence it so that the relays only carry current, never switch it.


At around the 30A+ level relays get difficult to make because the flexure (either a wire or a spring) has to carry the full load current. Contactor type construction is an option (two sets of contacts with bridging), but they get expensive and power-hungry, and they often don't have full current N.C. contacts.

If you wanted to effectively make your own SSRs you could use 8 MOSFETs (eg. CSD18542KCS, if that fits your detailed requirements) and 4 driver circuits.

If you use photovoltaic-output optocouplers with built-in turn-off circuit (eg. Toshiba TLP3xxx) , you can minimize the component count.

They tend to be rather slow so it probably would not be practical by that approach to try the more elegant approach of replacing the bridge with 4 AC switches that are controlled synchronously with the mains.

\$\endgroup\$
7
  • \$\begingroup\$ I should've probably clarified in the post, but the relays are never switching with the output connected, so it won't be switching under load. It seems the issue is just coming from the contacts themselves failing when carrying the current. \$\endgroup\$
    – M_Stenton
    Feb 2, 2022 at 15:10
  • \$\begingroup\$ Is it the N.C. contact failing? \$\endgroup\$ Feb 2, 2022 at 15:11
  • \$\begingroup\$ I believe it is, which makes sense from the data sheet - we're clearly putting too much current/volts through it so it's pretty unreliable. I just can't find any relays that have a high enough rating, so I was hoping that there would be an alternative way of switching modes. \$\endgroup\$
    – M_Stenton
    Feb 2, 2022 at 15:27
  • \$\begingroup\$ @M_Stenton It's not hard to find relays that can switch over 35 amps, but relays made for such high current tend to be called "contactors" instead of "relays". \$\endgroup\$
    – Hearth
    Feb 2, 2022 at 15:46
  • \$\begingroup\$ @Hearth I'll have to have a better look, but from what I've briefly looked at just now it seems like they have the same issues relays have where the majority are AC or DC rated, not both. Price wise also seems really high. I'll have a better look in a bit though, thanks for the info. \$\endgroup\$
    – M_Stenton
    Feb 2, 2022 at 15:58
0
\$\begingroup\$

I disagree with your premise that there aren't any relays available with appropriate ratings. A search on digi-key turned up this one, rated for switching 1200 W (which appears to be exactly what your device uses, if your numbers are correct and assuming the load is resistive--do give more information about the nature of the load!) at a maximum of either 110 V DC or 277 V AC, with a 50 A carry and switch rating on the contacts. Since it's SPST, you'd need four of them in total, and a slightly modified control method, but that should be a surmountable problem. The relay you linked in your question is part of the same series, but the contacts are rated for no more than 40 A--since you state that your load takes 50 A at 24 V AC, that's a significant overcurrent.

I would also make sure you actually are switching under no-load conditions; relay contacts don't switch off immediately, so if you aren't waiting long enough (check the relay switching time rating, and remember that it'll switch slower if you switch the coil with just a diode as a snubber) between switching off power to the load and switching the AC-to-DC relay, you may actually be switching under load. Also be aware that an inductive load is much harsher on your relay than a resistive one, when switching under load.


If that relay doesn't suit your needs, I was also able to find this one and this one which seem potentially usable as well, though both are more expensive. Remember that you don't necessarily need a relay rated to switch both AC and DC; you could use one relay to switch the AC path and a different model for the DC path, either to cut costs or optimize performance.

\$\endgroup\$
1
  • \$\begingroup\$ You're right about it being a solvable problem. I did put together a SPST version a while back with different relays, but think I was worried about shorting the connections since the relays wont be switching at exactly the same time (thinking about it now, that would be easy enough to get around, don't know why I didn't). I've also been worried about getting stock of the relays, as I'm probably going to need a enough for at least 50-100 boards worth a year, so been looking at ones I can find a large stock of. I'll have a better look and see if I can get a redesign going, cheers for the help. \$\endgroup\$
    – M_Stenton
    Feb 2, 2022 at 20:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.