Signal relays vs LC2MOS at low levels

[Spehro Pefhany]

Hi, all:-

I'm in the process of designing relatively high-precision signal
conditioning and data acquisition system front end.

It's desirable to have some kind of gain/range switching to maximize
the dynamic range over different operating modes.

Available options for switching are ADGxxx type analog multiplexers
(voltages are relatively high.. say +/-15V so HCxx are out) or "2A"
telecom style signal relays as are quite commonly used in bench-top
instruments.

All the relay data sheets that I've looked at have disturbingly high
recommended minimum voltages/currents. Say 10uA or 100uV-10mV (usually
the latter). I'd prefer to have just 10-20pA or so at < 1uV. With some
compromise I could arrange to have 10uA flowing through the contacts,
but that would result in a (possibly variable) ~1.5uV voltage drop. If
it varied with vibration that would be Bad. There's a reference to
some with special Silver-Palladium contacts for low-level loads, rated
at 10mA max, but I see no availability.

Any actual experience with these things? It might be possible to
"exercise" the contacts occasionally, or just before a measurement.

Also, on the ADGs, the typical leakage on the best ones is in the 10's
of pA (according to the data sheets, I have not yet made any
measurements). Any experience on the effects of long term radiation?
Usually things get worse, but how much worse?

Also a bit worried about the "sealed" relays in low pressure.

Thanks for any info,





Best regards,
Spehro Pefhany

 

[Tim Williams]

For relays, for very small signals and distortions, you might see if you
can find mercury-wetted types. They're still around, if a bit pricey.

I know normal relays are regularly used for small signals, but I don't
think I've ever seen any sort of guarantee applied to that use.

Tim

 

[[email protected]]

Hi, all:-



I'm in the process of designing relatively high-precision signal

conditioning and data acquisition system front end.



It's desirable to have some kind of gain/range switching to maximize

the dynamic range over different operating modes.



Available options for switching are ADGxxx type analog multiplexers

(voltages are relatively high.. say +/-15V so HCxx are out) or "2A"

telecom style signal relays as are quite commonly used in bench-top

instruments.



All the relay data sheets that I've looked at have disturbingly high

recommended minimum voltages/currents. Say 10uA or 100uV-10mV (usually

the latter). I'd prefer to have just 10-20pA or so at < 1uV. With some

compromise I could arrange to have 10uA flowing through the contacts,

but that would result in a (possibly variable) ~1.5uV voltage drop. If

it varied with vibration that would be Bad. There's a reference to

some with special Silver-Palladium contacts for low-level loads, rated

at 10mA max, but I see no availability.



Any actual experience with these things? It might be possible to

"exercise" the contacts occasionally, or just before a measurement.



Also, on the ADGs, the typical leakage on the best ones is in the 10's

of pA (according to the data sheets, I have not yet made any

measurements). Any experience on the effects of long term radiation?

Usually things get worse, but how much worse?



Also a bit worried about the "sealed" relays in low pressure.



Thanks for any info,











Best regards,

Spehro Pefhany

--

"it's the network..." "The Journey is the reward"

[email protected] Info for manufacturers: http://www.trexon.com

Embedded software/hardware/analog Info for designers: http://www.speff.com

The 20-bit A/D's are supposed to be good for eliminating this kind of headache.

 

[Klaus Kragelund]

We use



FUJITSU FTR-B3GB4.5Z-B10

OMRON G6KU-2F-Y-DC5

NEC ELEC UB2-5S



(all direct replacements)



about 16,000 so far. No problems. These are sealed, latching, DPDT relays. We

use latching to keep coil power hence thermal EMFs down. We use these in

low-level, thermocouple and such circuits. No semiconductor can come close to

their on resistance, off capacitance and leakage, bandwidth, or voltage

capability.

Haven't you seen fretting or low drive current oxidation of the relay contacts causing failures?

Cheers

Klaus

 

[Klaus Kragelund]

On a sunny day (Sun, 15 Sep 2013 13:19:46 -0400) it happened Spehro Pefhany


<[email protected]>:





















Generally avoid relays for small signal altogether.

I have seen those used in video / audio cross bars in TV studio,

and replacing relays was a frequent occurrence.

Indeed all relays (also gold contact) need a minimum 'wetting' current.

In the uA range that condition would not be met.



And on the high current side, I remember one of my first assignments

was in the city where we had all those film editing tables at different

locations, go there with a box of relay contacts and ask politely

if I could interrupt their work and replace those contacts as

preventive maintenance, learn to know the people and places...



In short, relays are not reliable, at any level.

Just as an idea, is it not possible to amplify your signals to a more

usable level before doing the switching?

Or are you also a TeraOhm collector?



Relays

trimpots

batteries

electrolytics

I wouldn't agree that electrolytics are unpredictable. If you know the working conditions, it will be determistic

Cheers

Klaus

 

[Spehro Pefhany]

In short, relays are not reliable, at any level.
Just as an idea, is it not possible to amplify your signals to a more
usable level before doing the switching?
Or are you also a TeraOhm collector?

The level is pretty high (volts), but I'm looking for uV. Similarly,
the resistances are not very high (1K ~ <100K), but 10pA into 100K is
1uV. That's 1.5T ohm with a 15V supply, so maybe you could say Tohm
(not quite correct, but gives an indication).

And even the datasheet minimum wetting current causes a possibly
undesirable variable voltage drop in one application where the input
is effectively chopped.

On reflection, I think unmodified relays are out for that application
because of potential outgassing, but for a second application (more
benign enviroment) might be okay.

*If* the datasheet minimum wetting current is, for some reason, not
applicable.


Best regards,
Spehro Pefhany

 

[Spehro Pefhany]

FUJITSU FTR-B3GB4.5Z-B10
OMRON G6KU-2F-Y-DC5
NEC ELEC UB2-5S

(all direct replacements)

about 16,000 so far. No problems. These are sealed, latching, DPDT relays. We
use latching to keep coil power hence thermal EMFs down. We use these in
low-level, thermocouple and such circuits. No semiconductor can come close to
their on resistance, off capacitance and leakage, bandwidth, or voltage
capability.

Thanks for the part numbers, John.

Just to be clear, do you use them with essentially zero contact
current?

The Omron datasheet, for example:
http://components.omron.com/compone...2CFA49FCB185257201007DD585/$file/G6K_0313.pdf

(front page)

---
Min. permissible load (See note) 10 uA at 10 mVDC

Note: This value was measured at a switching frequency of 120
operations/min and the criterion of contact resistance is 50 Ohms.
This value may vary depending on the switching frequency and operating
environment. Always double-check relay suitability under actual
operating conditions
---

Which is pretty disturbing if you want to operate at 6 orders of
magnitude lower current, and (maybe) 3-4 orders of magnitude lower
voltage. And that's with the relay being whacked 2x/second; not much
time for oxidation to build up.

Not sure what they mean by load voltage- presuably the open-circuit
contact-to-contact voltage. That may be >> 10mV, but closed circuit
current will be <<1nA.



Best regards,
Spehro Pefhany

 

[JW]

Generally avoid relays for small signal altogether.
I have seen those used in video / audio cross bars in TV studio,
and replacing relays was a frequent occurrence.
Indeed all relays (also gold contact) need a minimum 'wetting' current.
In the uA range that condition would not be met.

I wonder - is this why I see so many failures of relays in scope
attenuators? They usually fail with high contact resistance anywhere from
..5 Ohms up to 20 ohms or so.

 

[Klaus Kragelund]

On a sunny day (Mon, 16 Sep 2013 02:09:49 -0700 (PDT)) it happened Klaus


<[email protected]>:






I have had a radio-tv repair shop for many years.

Electrolytics will just stop working whenever they feel like it,

in the same equipment and circuit where other ones will last for 30 years.



Electrolytics, tubes, high value resistors, solder connections too.

Even some foil capacitors.

HV transformers, HV rectifiers, CRTs, all common stuff to replace.



As to elcos there also was this series of those in PC mobos that went all bad after a short while ...

But even the good ones, sometimes go bad.



Elcos are most certainly a life time limiting factor in electronic equipment, more than anything else.

Come to think of it I replaced the caps in the power supply of this Samsung monitor too some years back,

reported here, else it would have been end of life already.

Yes, but that was a case of planned obsolescence (they lost the law-suit over this) and the other cases are caps used in designs without doing the math.

If you contact the major manufactors of electrolytic capacitors, you get quite good and reliable lifetime data (mostly IRMS and ambient temperature related). I have even gotten special capacitors with burried thermocouples for testing.

Cheers

Klaus

 

[Spehro Pefhany]

Yes, but the best feedback is always from the service departments...
Comebacks are expensive, very expensive under guarantee for manufacturers of consumer stuff.
Manufacturers will always present their data so you really want to buy, that goes for datasheets too

I have very little doubt about the reliability of most aluminum
electrolytics. Even off-brand parts are pretty good if they're not
pushed too hard ripple-current wise. They are like power relays-
lifetime is known fairly well and will be less than semiconductors,
but good designs can be created using them.


Best regards,
Spehro Pefhany

 

[George Herold]

Hi, all:-
I'm in the process of designing relatively high-precision signal
conditioning and data acquisition system front end.
It's desirable to have some kind of gain/range switching to maximize
the dynamic range over different operating modes.
Available options for switching are ADGxxx type analog multiplexers
(voltages are relatively high.. say +/-15V so HCxx are out) or "2A"
telecom style signal relays as are quite commonly used in bench-top
instruments.

All the relay data sheets that I've looked at have disturbingly high
recommended minimum voltages/currents. Say 10uA or 100uV-10mV (usually
the latter). I'd prefer to have just 10-20pA or so at < 1uV. With some
compromise I could arrange to have 10uA flowing through the contacts,
but that would result in a (possibly variable) ~1.5uV voltage drop. If
it varied with vibration that would be Bad. There's a reference to

some with special Silver-Palladium contacts for low-level loads, rated
at 10mA max, but I see no availability.

Any actual experience with these things? It might be possible to
"exercise" the contacts occasionally, or just before a measurement.

Hi Spehro, I wish I could help.
For my own edification, do you have any idea what the problem is at low currents? (and how that might show up in the circuit.) Do mechanical switches have the same 'issues'. Or does the sliding /wiping of the contacts help keep them clean?

George H.

 

[Spehro Pefhany]

Some time ago somebody posted a picture here of a Chinese capacitor with a high uF marking,
opened with in it a small elco

Might have been me.

So much for 'off-brand', or even 'brand', you would have to order directly from the manufacturer.

Or from an authorized distributor. Just no ebay and be careful with
"market" items. I did get some caps directly from the manufacturer
(not China, this time) that were completely missing the rubber bung.
Something like 47uF/16V 0.002CV. Naturally, their capacitance
measured a bit on the low side by the time we got them.

One of the major distributors was hit by fake stuff too.
I had some fake MAX232s.

Were they good enough? Or just re-marked garbage?

 

[Spehro Pefhany]

Yes, but the best feedback is always from the service departments...
Comebacks are expensive, very expensive under guarantee for manufacturers of consumer stuff.
Manufacturers will always present their data so you really want to buy, that goes for datasheets too

There are electrolytic caps in just about every piece of electronic
mains-powered consumer goods out there.. light dimmers being a rare
exception. CF bulbs, AC adapters etc. etc. are all dependent on the
cap working well enough, and many applications are really sub-optimal
being a hot environment. A lot of stuff the service department will
never see if it lasts much past the warranty period. I've seen repairs
come back on stuff that was 15+ years old in the industrial world.
Re-capping the PS caps only (not the signal ones) was a frequent
repair. That's why I use better rated caps in those locations.

 

[Spehro Pefhany]

Hi Spehro, I wish I could help.
For my own edification, do you have any idea what the problem is at low currents? (and how that might show up in the circuit.) Do mechanical switches have the same 'issues'. Or does the sliding /wiping of the contacts help keep them clean?

George H.

I think it's a surface film effect where you can get some kind of ugly
metal-oxide semiconductor action until enough current passes to
microweld the metal parts together and form a resistive path.

Looking at an SRS 830 Lock-in amplifier front end schematic, and they
use similar Hasco relays to what John was mentioning, in an
application where the current is set by some leakage and a 10M to
ground, so it could be fairly low (~100pA) with a 1 mV level input.
There are lots of other examples around, but that's from a mature
product.

That particular Chinglish-esque relay datasheet shows no minimum
current, and also refers to optional gold-clad silver-palladium
contacts, but with no part number variant to cover that option (newer
versions of the datasheet show only gold-flash silver-nickel alloy.

My _guess_ is that there is enough wiping action in the contacts
closing in a sealed environment that it won't be a problem, provided
they _only_ ever switch extremely low currents, but the datasheets
from the Japanese suppliers (the only datasheets I take seriously)
look to be covering their butts fairly extensively.

My automated ratio transformer has relays, and it goes through a noisy
clackety-clack session whenever it's powered, exercising all the
relays.

 

[Spehro Pefhany]

I'll check with testing, but I think we've had zero failures. We are working at
microvolts and picoamps, high impedance thermocouple signals and similar stuff.
We're switching ranges and switching channels to a test bus, so we don't switch
often. I'd expect some applications to switch once a month or so.

That would be fine for the intended application. Thanks for the info.

I think we mostly buy the Fujitsu parts. We use non-latching versions, too.

That's a good brand. Without further knowlege I would pick
Fujitsu-Panasonic/NEC-Omron-Hasco in that order.

The part number you gave has gold-silver-nickel contacts- the
low-level variant has 'P' instead of 'Z', but they look to be
special-order.

I have some timing data around here somewhere. They will go on/off in about a
millisecond, something like that.

So about 1/3 the data sheet limits. The Fujitsu data sheet has a
distribution of operate times.. 1ms for most and 1.5ms for about all,
and they might bounce for some hundreds of usec after that.

 

[Spehro Pefhany]

They were useless, would go into latchup if you plugged in any cable,
did throw those away, ordered locally from conrad, those worked OK.

I got some 3V NS RS-232 drivers that were tetchy as all get out..
latchup if you look at them crossways.

 

[Klaus Kragelund]

Might have been me.






Or from an authorized distributor. Just no ebay and be careful with

"market" items. I did get some caps directly from the manufacturer

(not China, this time) that were completely missing the rubber bung.

Something like 47uF/16V 0.002CV. Naturally, their capacitance

measured a bit on the low side by the time we got them.

They actually all do, the ones I have measured anyway, goes for ceramics too. If the spec says +/-20%, you would be pretty stupid, financial wise, to add to much material and sell +20% parts. Better to optimize the production and earnings

Cheers

Klaus

 

[George Herold]

I think it's a surface film effect where you can get some kind of ugly
metal-oxide semiconductor action until enough current passes to
microweld the metal parts together and form a resistive path.

Looking at an SRS 830 Lock-in amplifier front end schematic, and they
use similar Hasco relays to what John was mentioning, in a
application where the current is set by some leakage and a 10M to
ground, so it could be fairly low (~100pA) with a 1 mV level input.
There are lots of other examples around, but that's from a mature
product.

That particular Chinglish-esque relay datasheet shows no minimum
current, and also refers to optional gold-clad silver-palladium
contacts, but with no part number variant to cover that option (newer
versions of the datasheet show only gold-flash silver-nickel alloy.

Thanks Spehro, I know next to nothing about metallurgy (Well except that gold is magical.) But it's weird that the underlying materiel would make a difference. Gold over some silver alloy... why would the exact silver alloymake a difference? (Yeah, I read some relay app notes that talks about silver-palladium as 'the best' for low signals.) But one wonders if that's only better because it's a better alloy once the gold has been zapped off bya few arcs. I guess I'd look for a contact with a heavy gold layer.. and not some wimpy flash.

Say, I don't suppose there's enough room to put two relays on the critical signal path. (Or would that be worse? Now the dribble of a current is split two ways.) Could you arrange it so that you pushed some 'big' (1mA) current through the relay once it was closed? (Clear it's throat, so to speak..)

George H.

 

[Spehro Pefhany]

Thanks Spehro, I know next to nothing about metallurgy (Well except that gold is magical.) But it's weird that the underlying materiel would make a difference. Gold over some silver alloy... why would the exact silver alloy make a difference? (Yeah, I read some relay app notes that talks about silver-palladium as 'the best' for low signals.) But one wonders if that's only better because it's a better alloy once the gold has been zapped off by a few arcs. I guess I'd look for a contact with a heavy gold layer.. and not some wimpy flash.

Say, I don't suppose there's enough room to put two relays on the critical signal path. (Or would that be worse? Now the dribble of a current is split two ways.) Could you arrange it so that you pushed some 'big' (1mA) current through the relay once it was closed? (Clear it's throat, so to speak.)

George H.

The contacts are bifurcated, so you get 2 for the price of 1.

An R + low leakage C through an analog switch could give a > minimum
(say 20uA) current for a short time. . Have to think about relative
risks with that one.

 

[George Herold]

The contacts are bifurcated, so you get 2 for the price of 1.

An R + low leakage C through an analog switch could give a > minimum
(say 20uA) current for a short time. . Have to think about relative
risks with that one.

At some point do you think about measureing a bunch? (sounds time consuming)
So what's wrong with analog switches? too much leakage? from where?
can you bootstrap it somehow?
(I know about as much about analog swtiches as I do about relays..
not much, every time I've used either they've worked fine.)

George H.