[josephkk]

On Aug 30, 4:13 pm, John Larkin

On 2012-08-29 21:42, John Larkin wrote:

On 2012-08-29 16:08, John Larkin wrote:
[...]
Well, 74-series TTL. Ironically, it's fairly recently that relays
started getting really good, small/fast/reliable/cheap enough to use
in signal circuits instead of CMOS multiplexers.

I must have been looking the other way. What relays are that?

Jeroen Belleman

We use a lot of the Fujitsu FTR-B3G type parts. DPDT, surface mount,
available as latching or not. NEC and Omron have drop-in equivalents.
As an analog switch, they blow away any CMOS part on everything but
switching speed. Capacitances are sub-1-pF, Ron a fraction of anohm,
holdoff hundreds of volts, bandwidth GHz, isolated drive.

OK, thanks. I use a fair lot of Pickering Series 103 reed switches
myself, and sometimes the parasitics get in the way.

Regards,
Jeroen Belleman

I haven't found reeds to be reliable.

Bell Labs invented them back in 1936 to get a more reliable relay for
use in the telephone system, and they worked pretty reliably for them
for many years.

Properly used, a dry reed switch is good for about 10^7 operations. A
mercury-wetted reed lasts about ten times as long.

I've seen data sheet claims of 1e9 cycles. In practise, they sometimes
stick closed, or make hi-z closures. The physics is bad: no wiping
action, possibility of residual magnetism latching them closed,
metalurgical bonding sticking contacts closed, magnetic interaction
with adjacent relays. I suspect that getting the reeds exactly
aligned, as molten glass cools and hardens, is a trickly production
challenge.

assembly starts around 4:30

-Lasse

Not the least bit like banks making money:

?-)

 

[JW]

Reeds really suck.

And yet for some reason, they are used exclusively in high end DMM
switching circuits (Keithley 2000 series, Agilent 344XX series, and even
the Agilent 3458A). Not sure why, but they must be good for some
applications.

 

[JW]

And yet for some reason, they are used exclusively in high end DMM
switching circuits (Keithley 2000 series, Agilent 344XX series, and even
the Agilent 3458A). Not sure why, but they must be good for some
applications.

Minor correction: I should not have stated exclusively, as there are a few
latching types used as well.

 

[Mr Stonebeach]

I've been perusing the Omron G6K relay series datasheet. I like
it, but I worry about RF characteristics. I worked out the open
contact capacitance to be about 160fF. It may be half that,
depending on how they got the HF isolation plot; They don't say.

I haven't found anything concerning coupling or capacitance
between the two poles in these relays. Would you have some
measured figures to share?

(I ordered a few to find out...)

Thanks,
Jeroen Belleman

Just threw a G6JU (which I use for LHe and sub-1K stuff) into
Agilent E5061 with the 16092A fixture. It showed ~0.95pF for an open
contact and ~15nH for a closed contact. The closed contact resistance
was maybe 10 mohms (less than resolvable) at low frequencies,
increasing to 2 ohms at 500MHz.

Regards,
Mikko

 

[Jeroen Belleman]

Just threw a G6JU (which I use for LHe and sub-1K stuff) into
Agilent E5061 with the 16092A fixture. It showed ~0.95pF for an open
contact and ~15nH for a closed contact. The closed contact resistance
was maybe 10 mohms (less than resolvable) at low frequencies,
increasing to 2 ohms at 500MHz.

Regards,
Mikko

Thanks Mikko,

From the specified 60dB of isolation at 10MHz I'd expect some
320fF between open contacts. The figures don't match. How did
you mount it?

Jeroen Belleman

 

[John Devereux]

Maybe for t-ohm leakage? Their long dimension helps there, assuming
you keep the coil contacts away from the signal.

Seems like the thermals would be a problem in a DVM. We use the
Fujitsu latching relays in thermocouple applications.

Here is a picture of the insides of my Keithley 6517A electrometer.

<http://ee.devereux.me.uk/img_1163-1.jpg>

The red tubes are coto reed relays, special types with ultra-high
insulation resistance and screening. There are also some conventional
NEC relays, seem to be the style you like.

Also note the sticking up thing with the meandering track. It has 1T
written on it...

 

[Charlie E.]

On Thu, 30 Aug 2012 08:24:51 -0700 (PDT),BillSloman

On Aug 30, 4:13 pm, John Larkin
On Thu, 30 Aug 2012 08:51:11 +0200, Jeroen Belleman

On 2012-08-29 21:42, John Larkin wrote:
On 2012-08-29 16:08, John Larkin wrote:
[...]
Well, 74-series TTL. Ironically, it's fairly recently that relays
started getting really good, small/fast/reliable/cheap enough to use
in signal circuits instead of CMOS multiplexers.

I must have been looking the other way. What relays are that?

I've used thens of thousands of reeds and ditto the little
Fujitsu-type telecom relays. The telecoms are better in every respect.
I won't use reed relays any more.

Next time you design a volume-production product, use reeds. Get back
to us on how they work.

Back in the dark ages (1980...) I worked in the office where they had
just installed an EAX2 switch. Thousands of reed relays to do the
switching of telephone calls, in boards with about 20-40 relays on
each. They had special test gear to take the bad boards, and tell
them which relays had gone bad. Automated testing told them which
boards had gone bad. It was a small switch, just two 'exchanges,' and
they typically had 10-20 bad boards to test a day...

Those guys just LOVED reed relays... NOT!

Charlie

 

[Bill Sloman]

No contradiction. The mercury eliminates contact bounce but doesn't
eliminate twang. The reeds still vibrate after closure, making
low-level electrical noise in the magnetic field, but the contact gap
remains closed by the mercury.

And the mercury film acts as an energy absorbing damper, and kills the
the twang relatively rapidly.

All em relays will twang to some extent. In my experience, reeds do it
a lot.

There's no good reason to use reeds when much smaller, cheaper,
surface-mount, more reliable, lower coil power, standardized parts are
available.

There's one good reason - reed contacts are hermetically sealed, and
regular relays aren't. After prolonged exposure to the wrong kind of
air the contacts in the relays you fancy show increased contact
resistance, as is mentioned in the data sheet.

 

[Mr Stonebeach]

Thanks Mikko,

 From the specified 60dB of isolation at 10MHz I'd expect some
320fF between open contacts. The figures don't match. How did
you mount it?

Jeroen Belleman

My relays have gull-wing style leads, I clamped the leads directly
to the 16092A spring clip fixture.I have calibrated the fixture up to
the spring clips at some past time- I forget when. My calibration file
is old (and probably done with a different clamp-clamp separation), so
the instrument did read ~minus 200fF with nothing in the clamps, ~plus
750fF with the G6JU installed. The capacitance reading was quite flat
up to 500MHz which the fixture is spec'ed for.

I agree there's a discrepancy. Maybe I did something wrong, it was
just a quick-and-dirty test as a reaction to your post.

Regards,
Mikko

 

[Bill Sloman]

How do you know that? Got data?

It follows from the fact that mercury-wetted relays don't bounce. The
process that kills the bounce moves the mercury film around more
extensively than the process that damps the twang, but in both cases
the mercury sticks the contact areas together and has to move around
if there's any displacement between the contacts.

One doesn't need any more data than that to understand what's going
on.

The ones we buy are sealed. They are amazingly reliable in real life.

But not hermetically sealed, if the data sheet is to be believed. Have
you read it yet?

 

[Bill Sloman]

On 08/30/2012 03:39 PM, John Larkin wrote:
37 pm, John Larkin<[email protected]>
wrote:
21,BillSlomanwrote:
On Aug 29, 6:30 pm, John Larkin
21:27 -0700, John Larkin
[snip]
I first invented a Manchester decoder (and encoder, and a supervisory
control system that sold megabucks) when I was an EE undergrad.
Deriving the receive clock is easy: a transition detector followed by
a one-shot of 0.75 times the bit period.
"Invented?"  Bwahahahahaha!  That's like Gore claiming he invented the
Internet.  Your "invented" method can be found in ancient textbooks.
                                         ...Jim Thompson
I didn't say I invented it first, just that I invented it. Before the
internet, it was hard to access prior art. It's not shocking that lots
of people independently discovered the same circuits.
It's a bit depressing when an undergraduate with free access to a
university library declares that he couldn't access prior art.
Oh nonsense! Some things are so simple that reinventing them costs
less than looking them up. OK, so it's been done before. So what?
And some things look that simple, and aren't. When you look them up
you find out about the less obvious gotcha's.
More often, you find a circuit in RSI or somewhere and discover that
it's fundamantally unreliable: beta dependent, bad corner cases, stuff
like that. RSI is better read for amusement than guidance.
You are telling me this? I've now published five critical comments in
RSI, pointing out where their refereeing process had failed - in 1972,
1996, 1999, 2004 and 2009 - and they rejected similar comments I
submitted in 1980 and 1995.
Nobody is saying that every academic paper is to be relied on, but
rejecting the whole of the scientific literature because some of it is
flawed rather misses the point. We make progress because some of the
papers that get published reflect dramatic steps forward.
Larsen N T 1968 Rev. Sci. Instrum. 39 1 12
http://rsi.aip.org/resource/1/rsinak/v39/i1/p1_s1?isAuthorized=no
"Authorized=no"
I'd have to buy the silly thing to see if it's worth reading.
is difficult to fault. If you'd read that more for amusement than for
guidance you'd have missed quite a bit
I used to subscribe to RSI. It wasn't worth it. I downloaded maybe 20
interesting papers from the last 30 years or so, but never actually
used them for anything.
The closest thing I've seen to a useful instruments-journal paper was
one by a Danish guy called Lars Bager, back in about 1990.  He had this
idea to minimize mode hopping in diode lasers by dithering the bias
current, detecting the medium frequency noise power from the monitor
photodiode, putting that signal into a lock-in, and using an integrating
servo in the usual way to find the minimum.

Sounds crazy. When we had trouble with mode hopping at Fisons Applied
Sensor Technology - because it changed the direction of the laser
beam, and thus the angle of incidence of the beam on the surface being
interrogated, which we were monitoring to about one second of arc - we
just went over to a single mode laser diode and stabilised it's
temperature well enough that it always stayed in the same mode.

That was switching between transverse modes, which is a horse of an
entirely different colour.  Diodes don't do that anymore, by and large,
except for VCSELs.  *Mode hopping* is a fairly complicated
thermal-electrical-optical instability that makes diode lasers jump back
and forth between longitudinal modes at around 100 kHz, especially in
the presence of optical feedback.  A backreflection of ~ 10**6 is
generally enough to set it off.

So where do they get published? The British Journal of Scientific
Instruments - now Measurement Science and Technology - got some of
R.V. Jones papers, and Blackett's as well. Today the journal isn't
quite so mind-numbingly academic as RSI, and the refereeing is
marginally better, but they still publish the usual proportion of
trivial rubbish.

They turn up all over the place, but of course not very often, because
there aren't many that are that good.

That's just Sturgeons Law - there are a lot more mediocre papers
published in every field of science than there are good ones.

It's not precisely Darwinian evolution - the changes from paper to
paper as you track back through the citations aren't random - but
better papers do get read - and cited - more often than mediocre
papers.

And different readers are looking for different things - a rubbish
paper drawing lunatic conclusions can still provide useful data.

 

[Bill Sloman]

If the entire reed structure is ringing in multiple modes (which they
do... it's obvious from the induced waveforms) and the contacts are
closed and pushing against one another, and there's a bit of mercury
in the gap, I don't see how that bit of mercury is going to suck up
all the mechanical energy that's rattling around in the structure.

I'm sure that that bit of mercury is not going to suck up all the
mechanical energy rattling around in the structure - the vibrations
eventually die away die away even in conventional dry reeds, so
internal friction in the rest of structure is going to soak up at
least some of that energy - but viscous damping is always a handy
extra, if you can arrange it.

You are making all this up. I doubt that you have ever observed the
electrical noise from reed relay twang.

I haven't. I've always known that reed coils are noisy because you've
got ferro-magnetic metal moving in a magnetic field, and changing the
magnetic field as as the contacts open and close, but it's less of a
problem than the primary source of noise, which is changing the
current in the - appreciable - inductance of the relay coil to open or
close the reed in the first place.

All that twang does is a funny-shaped tail to the noise.

Getting excited about the twang is straining at a gnat after
swallowing a camel.

You probably didn't know it existed until I told you about it.

Dream on.

You have time on your hands: do some experiments.

Why? It's trivial.

 

[Bill Sloman]

Unless you actually want to use reed relays for low-level,
millisecond-range signal switching. I've done it, and it can be ugly.

Low level millisecond-range switching isn't going to be trivial.
Switches always inject noise into the circuit being switched at switch-
over.

Reeds suck.

And so do all the alternatives, but for different reasons. What you've
posted here suggests that you could have learned a bit more about reed
switches before you tried to use them, and could have thought a bit
harder about what was going wrong when they weren't doing what you
wanted.

Did you got to the trouble getting reeds with rhodium-plated contacts
for your low-level switches? Or did you just buy a packaged reed
switch from DigiKey and hope for the best?

 

[Mr Stonebeach]

Do you run them cold?!!!

I do.

Regards,
Mikko

 

[Jamie]

Wow. Presumably whatever's inside, dry air or nitrogen or whatever,
puddles and freezes in a corner somewhere. I'm surprised it doesn't
occasionally freeze up the mechanics.

Hmm
I can't picture devices with plastics that would work very well at 77k
or below in the first place? Unless I am miss understanding something here ?


Jamie

 

[Mr Stonebeach]

Wow. Presumably whatever's inside, dry air or nitrogen or whatever,
puddles and freezes in a corner somewhere. I'm surprised it doesn't
occasionally freeze up the mechanics.

In refrigerators the cold plate is located in vacuum. If the relay
is not hermetically sealed, I'd expect its interior to get pretty much
evacuated while the main vacuum can is pumped. If it doesn't, the
interior probably cools via the metallic parts, so that the residual
gas condenses on the metal surfaces where the surface tension makes
the liquid to spread rather uniformly before it freezes. The frost
layer cannot be thick, the solid volume is maybe 700 times less than
the gas volume. Still, you made a good point.

I find it amazing that the latching mechanism still works there,
it's based on what, a permanent magnet?

Regards,
Mikko

 

[Jamie]

Nothing wrong with plastic when it gets cold. It just gets a little
harder.

yeah, that's true but just don't perform any tensile motion on them.
Also I question if the seal would maintain at those temps?

Jamie

 

[Jamie]

In refrigerators the cold plate is located in vacuum. If the relay
is not hermetically sealed, I'd expect its interior to get pretty much
evacuated while the main vacuum can is pumped. If it doesn't, the
interior probably cools via the metallic parts, so that the residual
gas condenses on the metal surfaces where the surface tension makes
the liquid to spread rather uniformly before it freezes. The frost
layer cannot be thick, the solid volume is maybe 700 times less than
the gas volume. Still, you made a good point.

I find it amazing that the latching mechanism still works there,
it's based on what, a permanent magnet?

Regards,
Mikko

Ah, but a PM works better when they get colder..

Jamie

 

[Bill Sloman]

I read the data sheets and believed them. What I learned later is that
they have some nasty problems, including failure rates maybe 1e4 worse
than claimed. What I mainly wanted them to do is open and close
reliably.

Those failure rates strongly suggests that you were abusing the
contacts, whether you knew it or not.

I read the data sheets on the little Fujitsu telecom relays, and
learned later that they are true. Nobody would last in the telecom
business if their products weren't very, very reliable.

But reed relays were invented by Bell Labs, for the telecom business,
and were very. very reliable when used carefully.

I mostly bought reeds from the manufacturers, supposedly suitable for
my applications.

But you can't tell us which reeds, or why they were supposed to be
suitable for your articular job.

Mostly signal switching, but some cold (ie, zero
current at open/closure time) switching of liquid helium level probes.

It's actually "dry" switching. The threshold is something like 100mA.

Strange, but several of the reed makers are now out of business.

Not strange. Farnell's catalogue now has a lot more solid state
switches in the relay section than it did thirty years ago - the
market for reed relays has shrunk and that does drive manufacturers
out if business.

You seem to have deployed low-hundreds of reeds, and I don't know if
you would have gotten feedback on their long term performance. I've
used tens of thousands of reed and conventional relays (about 15k of
the Fujitsu telecom types so far) and I *do* get long term feedback
about failures. Yet you keep insisting that you know more about this
issue than I do.

Experience is a hard school but fools will learn in no other. I'm not
that kind of fool. I try to anticipate that kind of experience by
reading about what has happened to other people and how they coped
with it - it's what academic engineering is all about. You don't seem
to do that kind of reading, nor understand why other people would want
to, and insist that your own experience, limited and specialised as it
is, is worth more than the - admittedly imperfectly - documented
experience of the rest of the world.

 

[Mr Stonebeach]

Ah, but a PM works better when they get colder..

They indeed seem to.

Do you happen to know any high-frequency ferrites which would
survive cryo temperatures, BTW? Every one I have tried seem to
effectively turn into an air core in LHe. They must exist, as
evidenced by Pamtech circulators. There are some powdered iron and
nanocrystalline cores for lower frequencies, but I know nothing off-
the-shelf which would work for RF.

Regards,
Mikko