Q: 'Break' time of a mechanical switch?? - UK Vintage Radio Repair and Restoration Discussion Forum

Al (astral highway) · 21st Sep 2014, 10:35 pm

I've been wondering about the typical 'break' time of a mechanical switch/ range of typical times. I'm interested in getting an approximation/ order of magnitude to compare with say a reasonably fast power electronics switching device (say an IGBT with 50nS turn-off time for the sake of argument).

Let's suppose for illustration that the mechanical switch in question is a set of relay contacts and the circuit being switched off is drawing a large DC current of 3A RMS at around 200V.

Obviously the relay contacts are supposed to spring apart quickly, but the process is far from instantaneous, right?

...because there's a measurable period when there's some arcing, maintaining the current flow, which then progressively reduces as the contacts fly apart.

So: how long before complete break including the various stages? 10mS?
50 mS?

P.S: At the back of my mind, I'm trying to work out a simple design for something that will remove power from an IGBT half-bridge *only* when neither device is conducting... don't answer that yet, please, trying to work it out for myself - but it made me think of this Q first. A mechanical switch is obviously no answer, but I wanted to evaluate just how many conduction cycles could take place if someone used a relay in error.

Cheers,

Al.

kalee20 · 22nd Sep 2014, 12:33 am

It depends on the switch, and the circuit it's switching. But generally, switches switching a lowish voltage, resistive load (so there is no arcing) can be considered to switch virtually instantaneously, certainly sub-microsecond. Of course, the repetition rate is only a few tens of operations per second.

I do recall some experiments I did years ago with reed relays and measuring the time to operate and release. Can't remember the operate time, but I remember the time for the contacts to open after removing drive (by a transistor) was about a millisecond and that's the delay between turning off the transistor and seeing something happen at the contacts. (For the record, I'd used a Zener diode to absorb back-EMF rather than the usual anti-parallel diode). When the switching edge actually happened, rise time was limited by the 'scope I was using.

The fastest switching edges around, I do believe, are still what you get if you use mercury-wetted relay contacts operating into a 50ohm coax line. MOSFETs just don't cut it, by comparison.

Al (astral highway) · 22nd Sep 2014, 1:10 am

Quote:

Originally Posted by kalee20

The fastest switching edges around, I do believe, are still what you get if you use mercury-wetted relay contacts operating into a 50ohm coax line. MOSFETs just don't cut it, by comparison.

Thanks! A very informative reply, as usual!

Ah and I see that such relays are still out there...

You've inspired me to do some experiments with various configurations and my 'scope!

Cheers,

Al

Radio Wrangler · 22nd Sep 2014, 1:37 am

There are specifications for EMC measuring receivers that require wide bandwidth and operation to GHz. The difficult thing is handling the total energy in an interference pulse, even though it is spread over great bandwidth. These test receivers have in turn to be tested and calibrated. The usual calibration source is a mercury-wetted contact embedded in a 50 Ohm transmission line operating with a high DC voltage. Schwarzbeck made them if you want to look them up. Nanosecond rise times!

David

Al (astral highway) · 22nd Sep 2014, 8:17 am

Quote:

Originally Posted by Radio Wrangler

There are specifications for EMC measuring receivers that require wide bandwidth and operation to GHz... The usual calibration source is a mercury-wetted contact embedded in a 50 Ohm transmission line...Nanosecond rise times!

Hey David, thank you. I'll take a look at Schwarzbeck and the configuration you mention...just quickly skimming a summary. The pulse-forming circuit is beautifully simple if ingenious, but there's a lot more to the theory!

Cheers,

Al

Radio Wrangler · 22nd Sep 2014, 9:38 am

Its job is to simulate an arcing commutator, The EMC measuring receiver has to handle the full spectrum from the thing hitting it at once, whlle being able to make measurements right down to very low levels in narrow bandwidths without showing signs of compression. This is why EMC measuring receivers need a huge bank of preselector filters in order to reduce the total bandwidth of the comb spectrum getting further into the receiver.

A truly evil test and very old technology!

David

G6Tanuki · 22nd Sep 2014, 11:01 am

Is the load purely resistive, or is it inductive?

If inductive you need to consider the rate at which the contacts open to a non-arc-sustaining spacing - and if necessary take precautions to extinguish the ensuing arc at the contacts.

This can be something as simple as a R-C 'snubber' across the contacts, or a suitably-positioned magnet close to the relay-contacts to force the arc to take a longer path. The magnet-trick was the approach taken on a couple of offset-litho printers I once worked on where there were some seriously-heavy-duty solenoids being fed with 500VDC. A previous 'curator' of these had replaced a couple of the Octal-based relays with generic types without 'blow-out' magnets - they lasted about a week.

woodchips · 22nd Sep 2014, 12:30 pm

If you really want to know I can flog you a Programma Timer TM-2 whose job it is is to measure relay operate/release time delays. Works on 5V DC upwards to mains voltages. A counter timer will do, until you blow it up!

One caveat with mercury switches is that they can spike enough to blow up a spectrum analyser front end or the sampling diodes in a sampling scope. There was one as a test source in some Thomson scopes I had, 4.5GHZ real time, and the rise time was indistinguishable from the scope rise time. Fast! Still got one or two around somewhere?

Al (astral highway) · 22nd Sep 2014, 1:24 pm

Quote:

Originally Posted by G6Tanuki

Is the load purely resistive, or is it inductive?

The load is inductive - a resonant RF circuit. The circuit being opened is a half-bridge whose peak current will be set around 350A. (*Very* low duty cycle/ pulse width determined by a modulator explains how this apparently high figure is achievable)...

The DC power bus is as close as possible to the IGBTs. There is the expected transient protection for both devices - suitably rated transzorbs on both sides of each, including the gates.

And there has to be a shut-down protocol, I think similar to what I've read is used in transceivers (although I have no experience of these) to avoid damage. I haven't worked this out yet.

Modulation set 0%, modulator off
Half-bridge power off
Logic power off

Opposite on power-up.

Cheers,

Al

Al (astral highway) · 22nd Sep 2014, 1:27 pm

Quote:

Originally Posted by woodchips

If you really want to know I can flog you a Programma Timer TM-2 whose job it is is to measure relay operate/release time delays.

Woah, sounds possibly overkill for me and expensive, but please PM me!

Quote:

Originally Posted by woodchips

One caveat with mercury switches is that they can spike enough to blow up a spectrum analyser front end or the sampling diodes in a sampling scope.

Amazing and really counter-intuitive...

Cheers,

Al

21st Sep 2014, 10:35 pm	#1
Al (astral highway) Dekatron Join Date: Nov 2006 Location: London, UK. Posts: 3,496	Q: 'Break' time of a mechanical switch?? I've been wondering about the typical 'break' time of a mechanical switch/ range of typical times. I'm interested in getting an approximation/ order of magnitude to compare with say a reasonably fast power electronics switching device (say an IGBT with 50nS turn-off time for the sake of argument). Let's suppose for illustration that the mechanical switch in question is a set of relay contacts and the circuit being switched off is drawing a large DC current of 3A RMS at around 200V. Obviously the relay contacts are supposed to spring apart quickly, but the process is far from instantaneous, right? ...because there's a measurable period when there's some arcing, maintaining the current flow, which then progressively reduces as the contacts fly apart. So: how long before complete break including the various stages? 10mS? 50 mS? P.S: At the back of my mind, I'm trying to work out a simple design for something that will remove power from an IGBT half-bridge only when neither device is conducting... don't answer that yet, please, trying to work it out for myself - but it made me think of this Q first. A mechanical switch is obviously no answer, but I wanted to evaluate just how many conduction cycles could take place if someone used a relay in error. Cheers, Al. __________________ Al Last edited by Al (astral highway); 21st Sep 2014 at 10:53 pm. Reason: Added a P.S

22nd Sep 2014, 12:33 am	#2
kalee20 Dekatron Join Date: Feb 2007 Location: Lynton, N. Devon, UK. Posts: 7,080	Re: Q: 'Break' time of a mechanical switch?? It depends on the switch, and the circuit it's switching. But generally, switches switching a lowish voltage, resistive load (so there is no arcing) can be considered to switch virtually instantaneously, certainly sub-microsecond. Of course, the repetition rate is only a few tens of operations per second. I do recall some experiments I did years ago with reed relays and measuring the time to operate and release. Can't remember the operate time, but I remember the time for the contacts to open after removing drive (by a transistor) was about a millisecond and that's the delay between turning off the transistor and seeing something happen at the contacts. (For the record, I'd used a Zener diode to absorb back-EMF rather than the usual anti-parallel diode). When the switching edge actually happened, rise time was limited by the 'scope I was using. The fastest switching edges around, I do believe, are still what you get if you use mercury-wetted relay contacts operating into a 50ohm coax line. MOSFETs just don't cut it, by comparison.

22nd Sep 2014, 1:37 am	#4
Radio Wrangler Moderator Join Date: Mar 2012 Location: Fife, Scotland, UK. Posts: 22,867	Re: Q: 'Break' time of a mechanical switch?? There are specifications for EMC measuring receivers that require wide bandwidth and operation to GHz. The difficult thing is handling the total energy in an interference pulse, even though it is spread over great bandwidth. These test receivers have in turn to be tested and calibrated. The usual calibration source is a mercury-wetted contact embedded in a 50 Ohm transmission line operating with a high DC voltage. Schwarzbeck made them if you want to look them up. Nanosecond rise times! David __________________ Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done

22nd Sep 2014, 9:38 am	#6
Radio Wrangler Moderator Join Date: Mar 2012 Location: Fife, Scotland, UK. Posts: 22,867	Re: Q: 'Break' time of a mechanical switch?? Its job is to simulate an arcing commutator, The EMC measuring receiver has to handle the full spectrum from the thing hitting it at once, whlle being able to make measurements right down to very low levels in narrow bandwidths without showing signs of compression. This is why EMC measuring receivers need a huge bank of preselector filters in order to reduce the total bandwidth of the comb spectrum getting further into the receiver. A truly evil test and very old technology! David __________________ Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done

22nd Sep 2014, 11:01 am	#7
G6Tanuki Dekatron Join Date: Apr 2012 Location: Wiltshire, UK. Posts: 13,996	Re: Q: 'Break' time of a mechanical switch?? Is the load purely resistive, or is it inductive? If inductive you need to consider the rate at which the contacts open to a non-arc-sustaining spacing - and if necessary take precautions to extinguish the ensuing arc at the contacts. This can be something as simple as a R-C 'snubber' across the contacts, or a suitably-positioned magnet close to the relay-contacts to force the arc to take a longer path. The magnet-trick was the approach taken on a couple of offset-litho printers I once worked on where there were some seriously-heavy-duty solenoids being fed with 500VDC. A previous 'curator' of these had replaced a couple of the Octal-based relays with generic types without 'blow-out' magnets - they lasted about a week.

22nd Sep 2014, 12:30 pm	#8
woodchips Octode Join Date: May 2010 Location: Grantham, Lincolnshire, UK. Posts: 1,174	Re: Q: 'Break' time of a mechanical switch?? If you really want to know I can flog you a Programma Timer TM-2 whose job it is is to measure relay operate/release time delays. Works on 5V DC upwards to mains voltages. A counter timer will do, until you blow it up! One caveat with mercury switches is that they can spike enough to blow up a spectrum analyser front end or the sampling diodes in a sampling scope. There was one as a test source in some Thomson scopes I had, 4.5GHZ real time, and the rise time was indistinguishable from the scope rise time. Fast! Still got one or two around somewhere?