SSRs needed

[akis]

I have a device that produces from 40V p2p to 160V p2p on a 100R load at 200KHz (currents up to 700mA RMS depending on load, typical is 550mA RMS continuous). The device is attached to two such loads and switches between the two using 4 SSRs, two per load (needed because the loads are not isolated from each other).

After quite some search I have been using the AQZ205 SSR from Panasonic. These are MOSFET SSRs capable of handling AC loads, rated at 100V and 2A. However on testing I have found out that they leak as much as 25% or more when fed with 160V p2p (80V AC peak) which is well within the spec, but I suspect the 200KHz is what causes the problems.

I am now searching urgently for other SSRs to replace the AQZ205, I will order and try the AQZ207 which is rated 200V AC peak. However it is also rated at a lower current of 1A which may be a bit too close to my requirements. Maybe the MOSFETs will not like working full time at 75% of their current allowance ?

Has anyone got experience with SSRs and can maybe suggest another model?

[wraper]

IMO using SSRs is a wrong idea. Usually they are used at mains frequency, not 200 khz.

[akis]

Well can you suggest an alternative? The device is always on and switches at 1Hz if not more often. Relays (normal and reed) would make a constant clicking noise and would not last long. That's why I decided on SSRs as well as the total isolation.

[Marco]

It's weird the datasheet has nothing on output transient immunity, I assume it's because of dv/dt turn on.

[akis]

An interim solution is to use two of them in SERIES, and I am testing and absolutely nothing gets through. I am soldering them one on top of the other as if they were made as one piece.

[wraper]

Well can you suggest an alternative? The device is always on and switches at 1Hz if not more often. Relays (normal and reed) would make a constant clicking noise and would not last long. That's why I decided on SSRs as well as the total isolation.

Make your own switching device from some discrete components. But proper way of course would be switching the formation of the signal, not switching the signal itself.

[akis]

It is the equivalent of a power amp connected to two sets of speakers, with an external (to the amp) circuit which switches between the two. The amp is heavy, bulky and uses a lot of components so it is better to switch the amp's output using a small circuit than to use two amps.

[mmagin]

I suspect that's a little low in frequency for it, but I wonder if something like a pin diode switching arrangement would work with slow enough diodes?  You still might need isolated drive, or suitable input and output coupling capacitors.

[akis]

VOM1271 and 2 back-to-back MOSFETs.

It seems so.

A technical person in Omron said that the AQZ205 has an output cap of 1000pF! He actually ran it in simulation and it is bad. But not as bad as my 'scope shows. For example the AQZ205 is rated at 100V. I tested with +/-80V at 200KHz on a 100R load and the relay when "off" lets through around +/-30V on the scope, ie 40% of the input voltage !!!! This is not simply the 1000pf cap doing that...

Therefore I will try to make my own "SSR" by using the VOM1271 (thanks by the way) and maybe the following FETs - can you choose one? I picked them for 150V, very low Rdson, and important low output cap.

Will it be:

http://uk.farnell.com/fairchild-semiconductor/fdd86252/mosfet-n-ch-150v-27a-to252/dp/2083238

or

http://uk.farnell.com/international-rectifier/irfsl4615pbf/mosfet-n-ch-150v-33a-to-262/dp/1752751

[Marco]

The VOM1271 datasheet lacks specs for output transient immunity and output capacitance, there is no SPICE model either. It will probably have more robust pulldown of the gates than the SSR you used, but it's hardly guaranteed.

PS. scratch that, the resistance of the depletion mode FET which pulls down the gate can be determined from the turn off time specs ... put that in a SPICE sim with the models of the MOSFETs you picked and you can check output dv/dt transient immunity. Turn off is 24 us, so rc time is ~10 us, so the pull down is ~50 kOhm. That doesn't sound great.

[Marco]

Who cares? It is a back to back structure. Assuming the 2 FETs are ideally identical, it has infinite output immunity.

Maybe my mental model is lacking. AFAICS one of the MOSFETs will simply act as a diode and for the other the drain-gate & gate source capacitors will act as a divider which tries to pull up the gate when there is an output transient, opposed by the pull down from the VOM1271 (~50 kOhm).

PS. the ratio of Cgs/Cgd is huge though for the  FDD86252, so it seems unlikely it will turn on.

[akis]

The VOM1271 datasheet lacks specs for output transient immunity and output capacitance, there is no SPICE model either. It will probably have more robust pulldown of the gates than the SSR you used, but it's hardly guaranteed.

PS. scratch that, the resistance of the depletion mode FET which pulls down the gate can be determined from the turn off time specs ... put that in a SPICE sim with the models of the MOSFETs you picked and you can check output dv/dt transient immunity. Turn off is 24 us, so rc time is ~10 us, so the pull down is ~50 kOhm. That doesn't sound great.

The original problem was leakage with the SSR off. No current flowing in its internal LED. The reason is not explained by the MOSFETs output capacitance. It is probably a limitation of the output MOSFETs wrt the applied voltages.

As there are not any MOSFET SSRs with >100V AND 2A capacity, I found the interim solution: put two AQZ205 in series. This eliminates the problem. But I am not sure about internal stresses.

For a better solution it was suggested I use two MOSFETs back to back and an opto driver. I have simulated the MOSFETs back to back (without the opto driver) and under error conditions, eg one of them shorted, the other one sees the whole output voltage across DS - that means the devices have to be much better than my max 100V peak, so I chose 150V devices.

Now I do not understand much the bit about the transients and the pulldown. You mean how fast it can switch the MOSFETs off when the LED current drops? In other words it may be that during switching from one channel to the other, both channels are ON momentarily? If yes, I can handle this in software (I can switch one channel off, wait, then switch other channel on). I already do something similar: when one channel is working at greater power than the other, then I lower the power before switching to the other.

[Marco]

The original problem was leakage with the SSR off. No current flowing in its internal LED. The reason is not explained by the MOSFETs output capacitance. It is probably a limitation of the output MOSFETs wrt the applied voltages.

My guess it's parasitic turn on. When the SSR is off there is only a depletion mode FET pulling the gate down which is relatively high resistance, so if Cgs/Cgd is small enough and dV/dT across the MOSFET is large enough it will turn the MOSFET on.

That said, I took a look at FDD86252 and the capacitance ratio is very high ... so you shouldn't be able to pull the gate up at these voltages.

[akis]

"That said, I took a look at FDD86252 and the capacitance ratio is very high ... so you shouldn't be able to pull the gate up at these voltages."

great, but I have bought the other one, the

http://uk.farnell.com/international-rectifier/irfsl4615pbf/mosfet-n-ch-150v-33a-to-262/dp/1752751

[Marco]

The margins of error are a bit slimmer for that one (Crss shoots up at lower voltages). Not really a problem though, you can just add gate source capacitance. If it still turns on add a couple of nF across the gates of the MOSFETs.

PS. normal turn on/off of your DIY SSR will be relatively slow, especially with added gate capacitance, >100 us.

[akis]

The margins of error are a bit slimmer for that one (Crss shoots up at lower voltages). Not really a problem though, you can just add gate source capacitance. If it still turns on add a couple of nF across the gates of the MOSFETs.

PS. normal turn on/off of your DIY SSR will be relatively slow, especially with added gate capacitance, >100 us.

IIUIC the test should be the same as before. Leave the LED unconnected, apply max voltage to MOSFETS (200KHz, 80V pk, load=100R), and see what passes through, on the scope.

The next test will be to test the turn on and turn off time, although this is less critical for me, switching frequency is less than 1Hz.

[TerraHertz]

Akis, do you realize SSRs always include an RC snubber network in parallel with the triac? The C value is quite small, since it's only meant to absorb very fast HV transients that could falsely trigger the triac. But it's intended for 50/60Hz, not 200KHz. At 200KHz that RC network will be passing a lot of power.

What kind of voltage drop is acceptable in your switch? How about a single MOSFET in a diode bridge?
Since your switching rate is so low, how about transformer isolated coupling of an AC drive signal, rectified for gate drive. Perhaps cleaning up the gate drive edges using one of the cheap 'power fail detect' 3 pin ICs.

[akis]

I have run some tests with the AQZ205 (100V 2A), AQZ207 (200V, 1A) and with my own back-to-back MOSFETS.

I placed the SSR in series with a 100R load and applied 82V AC pk at 200KHz. The voltage was applied for 1 sec and had 1 sec pause. The LED of  the SSR was disconnected. I then examined the amplitude leaked voltages on the load.

1) The AQZ205 (rated 100V), which is what I have been using all this time, immediately got too hot to touch, and let through around 40V pk on the load!

2) The AQZ207 (rated 200V) let around 9V pk through and was cold to the touch.

3) MY own back-to-back MOSFETS connected to the VOM1271 also let around 9V pk though, and did not get hot at all. I also shorted the Gate and Source to see whether it would make a difference and it did not.

I think therefore that the MOSFETs used inside the AQZ205 suffered a voltage breakdown and let the current through and that's why they also get hot. That is my theory.

The AQZ207 is much better but at 1A capacity is too close to my nominal load, I do not trust it either.

So the best solution is the custom made SSR comprised of the VOM1271 and two MOSFETs in this case the IRFSL4615PBF (150V / 33A)

Akis, do you realize SSRs always include an RC snubber network in parallel with the triac?

What kind of voltage drop is acceptable in your switch? How about a single MOSFET in a diode bridge?
Since your switching rate is so low, how about transformer isolated coupling of an AC drive signal, rectified for gate drive. Perhaps cleaning up the gate drive edges using one of the cheap 'power fail detect' 3 pin ICs.

No triacs on these SSRs, they are all MOSFET types. They are made to switch from very small to large AC waveforms without distortion.

The transformer idea is great I had not thought of making my own SSR this is my first attempt, but what's wrong with the VOM1271 ?