Analog switch or SSR?

Den mandag den 7. april 2014 17.05.24 UTC+2 skrev George Herold:

voltage?

-Lasse

An SSR is a nice floating, usually bidirectional, switch, good for a lot more voltage and current than an IC. But they switch in a millisecond or so, whereas analog switches go in more like a microsecond.

We like the Claire CPC1008N... used over 12K of them so far. I have lots of data on its behavior, including destruct limits. I figured out how to derive the SOAR destruct curve without destroying an SSR for every data point.

I wish someone would make a good, affordable self-protecting SSR. I suppose it would be difficult.

Yes, it will have some voltage :^) So in the 30V range. George H.

OK thanks.. 8 ohm Ron is a bit much for me... but the CPC1002N looks nice! (or maybe the TLP222A) (I need two, one too short the current source and the other to connect to the load.) So causes the SSR to fail? Too much power dissipation? (over voltage?)

George H.

On a sunny day (Mon, 7 Apr 2014 08:29:35 -0700 (PDT)) it happened George Herold wrote in :

MOSFET

Like any other mosfet, it's a SOAR (safe operating area) thing. A mosfet has a family of curves, each for a time duration. Each curve is the voltage-drop:current envelope above which the part might fail.

An SSR doesn't have a controllable gate voltage, so it has a mostly fixed voltage-current curve. So the SOAR set can be expressed as a single voltage-time curve.

That last one is the approximate safe area curve.

Well I guess I could make that work. I can stick an SSR almost anywhere in the circuit, but I have to think about the FET.

George H.

John Fields' suggestion is good.

You haven't described how you've implemented your current source. Perhaps the switching can be incorporated in the source itself? ...Jim Thompson

Cool thanks for the pics. I guess you could try running the LED at some low current and just get the FET's to turn on... but I don't want to play with SSr's now.. for this I just want a switch.

George H.

ts of data

e the SOAR

e!

d the other to connect to the load.)

Well I was thinking about something like that. (but until Jan mentioned it I hadn't thought of just a FET.) One issue is that I really want *NO* cur rent when the switch is off. (or the current shorted in this case.) Even i f I got a low R_on FET, once I put it on a pcb I'm guessing there will be ~

George H.

Yeah... I still may use a SSR.

Hmm, I haven't thought about that part... I'll need a range of currents. M ost likely I'll just use a reference voltage into an opamp w/ current sense resistor and FET pass element. Oh, that's going to waste a lot of power a t 30V and 300 mA. Grumble.. I better rethink this. I wonder how much of a transient I'd get if I left the current source open circuited when not in use?

OK thanks for the head slap, George H.

--- Then you could simply use a single SSR like this:

.Iret>----------------+ . | . [LOAD] . +--------+ | .Iout>----|IN OUT|--+ . | | .ON/OFF>--|SW+ SW-|--+ . +--------+ | .OFF/ON>--------------+

If you're using a constant current supply, make sure the SSR can stand off the compliance voltage of the unloaded supply.

Alternatively, you could use a P channel MOSFET or a PNP transistor as the switch, like this:

. PCH .Iout>----+--E/S C/D--+ . | B/G | . | | | . +-[R]--+ | . | | . [R] [LOAD] . | | . ___ C | .ON/OFF>--[R]--B NPN | . E | . | | .GND>------------+------+

Be aware of the V(gs) limits for the MOSFET.

John Fields

FWIW, I think I've run across the specific reason why failures in linear mode occur without violating datasheet SOA curves.

Apparently the datasheet curves are not measured, but rather _calculated_, using outmoded models that are insufficiently sophisticated to account for local overheating. "the assumption that the junction temperature is essentially uniform across the power MOSFET die", according to this A/N:

You'd think they'd just leave them off the datasheet rather than publishing curves that are demonstrably inadequate to ensure their product doesn't blow up.

--sp

"Diverting", as John Fields suggested, should be fairly clean if you make your current source FB loop tolerant of voltage bumps in the drain path.

"Opening" will be chaotic, giving a recovery time established by the OpAmp et al.

...Jim Thompson

Analog switches are typically CMOS small-signal items (you cannot usefully assemble a power MOSFET version, the source-drain diode gets in the way). You don't WANT an analog switch, anyhow, because if you opened the switch, your current regulation loop would saturate its amplifiers; what you want, is to shunt across the load (short across a DC load, so for this, a MOSFET works fine). You might consider connecting (+) output to anode of a diode, cathode to the load, and shunting from the anode side of the diode to ground with your N-MOSFET switch. That way a few milliohms of MOSFET resistance still doesn't put any significant current into any load.

I also tested an IRF740 under the same conditions; it blew at twice the power, way the hell past maximum.

Guess which one had the bigger die, too...

Fairchild transistors often don't show 2nd breakdown, which makes me suspicious. I think the worst offenders I've seen are IXYS HiPerFET VDMOS, which aren't usually rated for *any* DC current at Vdss/2. Truely optimized for switching. (And let's not even mention IGBTs.)

Tim

Thanks, A FET does look like the answer. Unfortunately, Jim comment upstream has caused me to 'give up' on the current source. I'm trying to measure how much energy is dissipated in a load (~100 ohm resistor). The current source has a certain appeal, in that it's one less thing to measure. But turning it off is wasteful. (or gets more complicated) So I'm now thinking of a voltage source, SSR, and current sense element. maybe a resistor and instrument amp, or a TIA opamp George H.