super fast turn OFF Relay

[qwaarjet]

I have an application where 15ohm resistive preload need to be disconcerted very fast when a parallel load is connected, currently my system is using a mechanical relay (400Vdc 15A)attached to my current sensor for the task. The current sense circuit responds in about 100uS but the relay take a little over 4mS to respond. The switching current is 10A DC. Looking around on digikey most SSR have turn off time greater than 1mS not very much faster than my mechanical relay, and most  device seems to be optimized for fast turn on instead of turn off. An SSR would be ideal so the switching element can float separately from the current sense circuit. I've used optical FET drivers in the past for switching floating loads but they are not very fast devices either. I would like to achieve turn off time on the order of 100s uS. Any suggestions on what product / architecture to use.

General requirements for the switch :
>10A
>180Vdc

[wraper]

Just for information. I don't know if you are concerned about this but coil inductive kick suppression can significantly increase relay release time.
Some info to read if you don't know this already http://www.te.com/commerce/DocumentDelivery/DDEController?Action=srchrtrv&DocNm=13C3311_AppNote&DocType=CS&DocLang=EN
Of course that won't give you 100uS release time, but anyway.

[qwaarjet]

The current system has a  flyback diode on the coil.

[wraper]

The current system has a  flyback diode on the coil.

So the worst possible release time.

[qwaarjet]

it was there to protect the fet running it. I just do it as a matter of course I fried to many chips over the year with kickback not to. and the measured switching time is still much faster than the datasheet specifies 

[wraper]

it was there to protect the fet running it. I just do it as a matter of course I fried to many chips over the year with kickback not to. and the measured switching time is still much faster than the datasheet specifies

Of course protection is needed. If you see the table at the bottom of that pdf, zener + diode gives more than 4x faster release time compared with plain diode. I personally use bidirectional tvs to keep it in one part. Just need that TVS clamping voltage exceed relay supply voltage and switch transistor can handle supply+clamping voltage. Faster release prevents contact arching and welding too.

[T3sl4co1l]

There are "gate driver" optos available, which are little more than the PV stack inside those SSRs.  Left to their own devices (uA of output current, very large G-S resistor to discharge it), they go very slowly, but if you built your own SSR from these parts, you could put a second opto in parallel with the driver one to shunt the voltage very quickly.

Tim

[Kjelt]

I think the optiocoupler input makes it slow, if there were ssr,s with the same technology as AD,s adum isolators it would be very fast.
However looking at Farnell the Celduc SSRs claim 20us off time, see as example
Ordercode: 1305689

[Marco]

The current sense circuit responds in about 100uS

That's silly slow for a start, one of those Allegro Hall current sensors respond in 5 uS ... what's taking you so long?

Why do you want the galvanic isolation? Without it doing it fast is trivial (but if you really do then T3sl4co1l's solution seems the best).

[qwaarjet]

It  probably  the way I was measuring it with the relay attached, so coil inductance, I was really only concerned with overall performance and didn't spend much time characterizing the sense circuit. the system uses a shunt resistor  that is amplified by and op-amp and fed into some 4000 series NAND gates then fed to the FET that drive the coil.

[qwaarjet]

I think the optiocoupler input makes it slow, if there were ssr,s with the same technology as AD,s adum isolators it would be very fast.
However looking at Farnell the Celduc SSRs claim 20us off time, see as example
Ordercode: 1305689

This type of relay looks like what I'm looking for thanks

[timb]

Don't Mercury wetted relays go pretty fast?


Sent from my Smartphone

[Seekonk]

Sounds like the Indiana Jones trying to balance the loads so the big rolling stone doesn't get you.  If you are trying to place a load when the motor load disappears, wouldn't it be easier and faster to just monitor the voltage and switch in the load when it jumps up?

[Marco]

Don't Mercury wetted relays go pretty fast?

They transition fast, but getting the contact to move out of the mercury pool still takes a comparatively long time.

[qwaarjet]

Sounds like the Indiana Jones trying to balance the loads so the big rolling stone doesn't get you.  If you are trying to place a load when the motor load disappears, wouldn't it be easier and faster to just monitor the voltage and switch in the load when it jumps up?

The resistor  load and a relay is  a bypass to  a "safety feature" of a Sorensen SFA constant current power supply. The SFA has to have a load on it at all times or it faults, however the load I want to drive starts open circuit. A plasma is then formed by another system and sustained by the SFA. I need to transfer the full current of the SFA to the plasma as fast as possible currently the plasma is getting too cold in the time it take the mechanical relay to open.

[Alex Eisenhut]

I have an application where 15ohm resistive preload need to be disconcerted very fast when a parallel load is connected, currently my system is using a mechanical relay (400Vdc 15A)attached to my current sensor for the task. The current sense circuit responds in about 100uS but the relay take a little over 4mS to respond. The switching current is 10A DC. Looking around on digikey most SSR have turn off time greater than 1mS not very much faster than my mechanical relay, and most  device seems to be optimized for fast turn on instead of turn off. An SSR would be ideal so the switching element can float separately from the current sense circuit. I've used optical FET drivers in the past for switching floating loads but they are not very fast devices either. I would like to achieve turn off time on the order of 100s uS. Any suggestions on what product / architecture to use.

General requirements for the switch :
>10A
>180Vdc

Sounds like the Indiana Jones trying to balance the loads so the big rolling stone doesn't get you.  If you are trying to place a load when the motor load disappears, wouldn't it be easier and faster to just monitor the voltage and switch in the load when it jumps up?

The resistor  load and a relay is  a bypass to  a "safety feature" of a Sorensen SFA constant current power supply. The SFA has to have a load on it at all times or it faults, however the load I want to drive starts open circuit. A plasma is then formed by another system and sustained by the SFA. I need to transfer the full current of the SFA to the plasma as fast as possible currently the plasma is getting too cold in the time it take the mechanical relay to open.

Use a capacitor as a load instead? It pretty much disconnects itself after ~5 time constants...

[C]

you are currently using

Turn on relay to open contacts:
normally open contacts using active on with passive off.

You could change your logic and use

Turn off relay to open contacts.
Normally Closed contacts using passive off with active on.

Not saying this is a option for you, just to get you thinking.

The passive above is the reverse diode.
 
As you have stated,
  you see a lot of fast on times but slow off, WHY?


Relays are not really voltage based devices. Think more of a relay as a current based device.
It is an inductor and resists current change. It takes time for the current to increase or decrease.
One common thing to do if current increase is to slow is to use more voltage to make it increase faster. 
An example may help
The holding Voltage/current to keep a relay on is normally lower then the voltage needed to turn it on. Keeping the higher voltage/current across the relay just heats the relay.
I think your using
Voltage -> Switch -> Relay -> Voltage
an example of above for DC would be
(More?)Voltage -> Resistor -> Cap -> Switch -> Relay -> Voltage)
Note: in steady on state the relay voltage would be between the dropout voltage and the turn on voltage, the remaining voltage is across the resistor.
 The cap must not be so large that the active voltage across the relay goes above it's max voltage rating when changing.

two more options.
passive on, passive off
 it's a switch not a relay
  and
Active On, Active Off
you reverse the power to the relay for a bit of time to turn off.

Be sure to think of other things that happens. Normal turn on for a relay may be like closing a door gently. When you make it faster you start slamming the door closed. To much, you break the door or door frame and it will not last as long.
To much active open and you have broke the door or broke the door stop.

I actually saw a design once where they had no other options. They used over voltage to start relay turning on. A set of on contacts to sense when the contacts closed and when reversed the voltage on the coil for a time to prevent slamming in to the relays stops.  Still would not last long, contacts were hammered and the faster close made the contacts bounce longer.

Guess I am missing something here,
a very quick look a SFA says very low capacitance on the Output so that the output can respond fast. Why idle at high power level, why not switch from a low power, low current state to a high power, high current state when needed?

C

[qwaarjet]

Guess I am missing something here,
a very quick look a SFA says very low capacitance on the Output so that the output can respond fast. Why idle at high power level, why not switch from a low power, low current state to a high power, high current state when needed?

The SFA can not even be turned on without a load attached or it will fault, not that it would indicate that or anything (I hate Sorensen products). I also want the DC bias from the SFA on the load prior the the plasma forming as it assists in plasma formation. I can't command the the SFA to go from a low power state to a high power state fast enough through GPIB to heat the plasma before it cools too much. My current system responds in 4mS, that is much faster than my GPIB +DAQ + PC +LabVIEW system used to control the SFA, and still to slow.

Use a capacitor as a load instead? It pretty much disconnects itself after ~5 time constants...

The purpose of using the SFA is to lower the system capacitance.

[wiss]

I have an application where 15ohm resistive preload need to be disconcerted very fast when a parallel load is connected, currently my system is using a mechanical relay (400Vdc 15A)attached to my current sensor for the task. The current sense circuit responds in about 100uS but the relay take a little over 4mS to respond. The switching current is 10A DC. Looking around on digikey most SSR have turn off time greater than 1mS not very much faster than my mechanical relay, and most  device seems to be optimized for fast turn on instead of turn off. An SSR would be ideal so the switching element can float separately from the current sense circuit. I've used optical FET drivers in the past for switching floating loads but they are not very fast devices either. I would like to achieve turn off time on the order of 100s uS. Any suggestions on what product / architecture to use.

General requirements for the switch :
>10A
>180Vdc

This is in the realm of MOSFETs.