I am working on a project that needs two or three circuits disconnected simultaneously (within 1 uSec of each other). The circuits will be 0-30 V and DC (< 1 amp) but the polarity will be unknown so half wave AC might be a better description. The circuit needs to run unmolested until it is time for disconnection so it seems most of the SSRs are out as they will trim some voltage or not even allow very low voltages to pass. I am looking at some fast release relays and matching them with a scope prior to board assembly but I feel there is a better way that I am missing. Any suggestions? Thank you.

 

You could use a N-ch mosfets on high-sides controlled with fast optocouplers, but you will need a 3 floated winding power-supply for driving the mosfets gates.

There exists a photovoltaic isolators that are able to drive the gates directly, but you need to find whether they support such speed (1us delay).

 

Why do they need to be disconnected within 1us?
I would suggest that a circuit with that stringent a requirement is not a good design in the first place.
What are you doing to match the inductances on all the supply cables, as that may adversely affect the disconnect time.

 

There was typo in my original post. I can't see how to edit it though. The time variance on the disconnect is 10 uSec, not 1 uSec.

 

You could use a N-ch mosfets on high-sides controlled with fast optocouplers, but you will need a 3 floated winding power-supply for driving the mosfets gates.

There exists a photovoltaic isolators that are able to drive the gates directly, but you need to find whether they support such speed (1us delay).

If I understand you correctly, this doesn't account for the potential that the current could be going either direction through the disconnect.

 

Why do they need to be disconnected within 1us?
I would suggest that a circuit with that stringent a requirement is not a good design in the first place.
What are you doing to match the inductances on all the supply cables, as that may adversely affect the disconnect time.

I posted a correction that it is 10 uSec.
The 10 uSec time was chosen to account for various factors like inductances on supply cables. The actual number is greater but the 10 uSec should provide overhead to keep everything under the actual number with variables accounted for. The actual number is protected by an NDA so it will not be disclosed but is largely if not entirely irrelevant.

 

In terms of what you're doing, there is no difference between 10us and 0us. Just simply disconnect power to all 3 at the same time by disconnecting their ground.

 

If that simple solution was an option, I wouldn’t be here with this question. Thanks anyway.

 

If that simple solution was an option, I wouldn’t be here with this question. Thanks anyway.

Most of the possible solutions are also under NDA from a lack of usable application information.

 

The circuits will be 0-30 V and DC (< 1 amp) but the polarity will be unknown so half wave AC might be a better description.

Then that's full-wave, not half-wave.

Does everything share a common ground?
Is isolation needed between the control and the load?

 

Does everything share a common ground?
Is isolation needed between the control and the load?

No, they are not. They do need isolation from each other. This is what makes all the semiconductor solutions I can think of not really viable relative to a mechanical relay. The relay solution has its own complexity but it is manageable. I am just looking for a product or innovative use of standard switching components I am not aware of. All of the SCR and transistor options I can think of either become too complicated to justify if they would work at all. I am pretty sure the relay option with a uC based calibration routine is going to be the answer. There may be some relays consistent enough off the shelf over time that would work without calibration or verification but I don’t want to risk it absent someone’s experience saying otherwise.

 

Trying to synchronise disconnections to that degree of accuracy is going to be a tall order, and hiding the original objective behind a NDA is going to prevent a lot of smart engineers on this forum from being able to give you a workable solution.

 

What it gets to quite rapidly is that fast is possible, but not cheap. There are very fast semiconductors (Gallium Arsenide), but isolated control for a bunch of them is not cheap and probably not fast enough. There are fast relays but those response times are in MILLISECONDS, not microseconds.The other issue with semiconductors is the voltage drop, which is never zero. What was not mentioned is how often this switching needs to happen. A higher powered laser beam could easily evaporate a number of conductors at the same instant, but switching back on would be slow. And a 20 KW laser is not a cheap item. And with fast relays, there is always the constraint of coil inductance.
What has not been presented is the timing accuracy relative to the intended opening time.
There are very fast RF transistors good for a few hundred megahertz, but they suffer with an on state voltage drop. Also they are not cheap, and they need driving circuits as well.

So I suggest going back to the circuit designer and getting the actual requirements as far as response times and differential opening times. AND consider that possibly, if the multiple circuits are related, the option of short circuiting the circuits to remove the applied voltages between them.

 

You might be able to do it with a 4 pole relay with the 4 contacts opening at nearly the same time.
10µs timing it still likely problematic, but it would be relatively simple to test.

 

If that simple solution was an option, I wouldn’t be here with this question. Thanks anyway.

Well, first, your description is lacking. Secondly, you're over complicating the issue due to lack of understanding. What exactly do you think you're trying to do? Your own description says (quote: "needs two or three circuits disconnected simultaneously") you are attempting to break the signal- Which exactly what I suggested you do.

Therefore my next question is this- disconnect from what if not from power? From each other? Learn to state your needs more clearly and you won't cause confusion.

 

You might be able to do it with a 4 pole relay with the 4 contacts opening at nearly the same time.
10µs timing it still likely problematic, but it would be relatively simple to test.

relays are slow because they are inductive. He can easily control whatever he needs with the appropriate FET(s). I've used FETs to replace relays on hydraulic farm equipment because it's faster, safer (no arc), and makes for a very neat PCB design.

IMHO

 

No, they are not. They do need isolation from each other. This is what makes all the semiconductor solutions I can think of not really viable relative to a mechanical relay. The relay solution has its own complexity but it is manageable. I am just looking for a product or innovative use of standard switching components I am not aware of. All of the SCR and transistor options I can think of either become too complicated to justify if they would work at all. I am pretty sure the relay option with a uC based calibration routine is going to be the answer. There may be some relays consistent enough off the shelf over time that would work without calibration or verification but I don’t want to risk it absent someone’s experience saying otherwise.

If they do not share a common ground, they are floating. As such, the voltage difference between every ground could be infinite. Ground is not zero in a circuit. That's why it's called 'common'. It is a common value that everything else in the circuit references against. We think of it as zero, but it could 57KV, or 12V, or -530V, or any value. It's relative. As such, it can vary wildly between independent/isolated circuits.

Therefore, if you want to disconnect grounds that are potentially different, you can use opto-isolator to control a FET at each different circuit. You an then control each circuit independently, or all at once, or just 2 of 3 etc, and can disconnect any one or all at exactly the same time. No delays, no difference between them.

If you want help, I am happy to sign your NDA, and we can discuss my rates. I do circuit, pcb, and embedded software design. Here's a smaller project we did for a customer:





This was both battery powered and could be connected to a walwart. Used in commercial applications in gas-stations and such, with RFID, etc. Full RTOS in 6K.

 

OK, perhaps we are thinking about this from the wrong point of view!! How long do the lines need to be connected??? A short on time is simpler to produce than a synchronized off switching. AND, once again, could momentarily short circuiting the three circuits be a useful alternative? Certainly mechanical switching in a synchronized manner is possible, but cost increases with accuracy, just like a lot of other things. Even high speed mechanical relays have response times of a few MILLI seconds.
So some discussion of the actual system requirements is in order.Three optical switches could be selected with close response times, but both photo-diodes and switching transistors have finite response times.

 

relays are slow because they are inductive

Of course.
I'm not referring to the slow operate or release time of a relay, but how close in time the multiple contacts would cease conducting when the contacts open (which is the time the TS is concerned about).
Whether that release time difference is close enough for his requirements, is the question.

 

Why do I get the feeling this thread was DOA?