Measurements on emf-error of switches

[Andreas]

But what do you mean by "almost missing"

When looking through the LTZ1000 thread some weeks ago,
just thought: where is that guy who shares his systematic measurement results?

I was probably going to switch them directly with some PCA9539DWR. Simple.

I am using 5V types and put one of the coils (178 Ohms) over a capacitor directly to a 5V Port Pin (30 Ohms in my case) of a microcontroller.
You can also put the 2 coils in series to save further current with slightly slower switching.
On power up I activate the pull ups of the processor to charge the capacitors without switching.

with best regards

Andreas

[chuckb]

The Keithley Model 2000-SCAN is a 10 channel scanner card for the 20XX DVMs. It's specified at less than 1uv of offset. The Service information specifies TQ2E-L2-5V relays. The card I used actually had NEC EA2-5TNJ Orange Relays, date code 1411.

To test the card I shorted all the inputs at the terminal block with bare 22 ga copper wire from the same spool. The card was installed in a K2000 however, for better resolution, the relay offset voltage was measured with a K182 Nano Voltmeter. Each channel was selected for 1 minute and the offset voltage was recorded every 10 seconds.

I verified that the different inputs on the card do have less than 1uV of offset. The offset repeatability for each channel seems to be less than 100nV. See the attached pdf for the strange offset pattern that this card provides. The numbers in the middle of the graph note which channel was selected at the time the voltage was recorded.

[TiN]

Thanks chuckb, interesting data. How did you log the data. Perhaps I can do same test on 2001-TCSCAN card. I have 182-M as well  .

janaf, happy to see you back. Voltnut community worried about your disappearance a bit  .

[Andreas]

See the attached pdf for the strange offset pattern that this card provides. The numbers in the middle of the graph note which channel was selected at the time the voltage was recorded.

Thanks for sharing.

Looks like the Relay 1 + 6 is located near a heat source.
4 + 10 are far away from this.
Do you have a plan of the physical location?

With best regards

Andreas

[chuckb]

TiN
The K182 was logging data as part of a larger Agilent VEE data acq program. It collects voltages from 14 DVMs every 10 seconds. It runs 24/7. I did not want to interrupt the larger program so I had the K2000 automatically step through the channels with a 1 minute dwell on each channel. The data collection and the channel changes were not synchronized but it did not matter for this application. The program recorded 5 or 6 good readings for each channel setting. I did not use the slowest filter on the K182 because of the rapid (1 minute) change of input signal.

During another set of tests I also used the the K1801 Nanovolt Preamp (less than 1nV noise) to record emf of a standalone Panasonic latching relay (TXS2) with the normal and alternate contact material. The relay was on a small pcb so the thermals would be realistic. The PCB was in a heavy aluminum case that was insulated with 2" foam. The PCB was laid out so the relay would reverse a signal or pass it straight through the two contacts. The performance was excellent. The relay had less than a 10nv step when reversed. Also the balanced EMF changed less than 5nV per deg C of ambient temperature change. The relay was arraigned so the individual contact thermal EMF was balanced with the other contact. So I don't know what each contact thermal EMF is. But they balance well. Basically the relay was stable within 15nV during 7 hours of testing. I will post that data when I dig it out of the large data file.

The next project will be to test the Electroswitch C4 rotary switch later this month.
 

[janaf]

janaf, happy to see you back. Voltnut community worried about your disappearance a bit  ..

Thanks!
Frankly, I hit the wall after too much work. It took a while to recover...

[janaf]

I have a couple of NI PXI-2503 scanners. Specified as "Thermal EMF (differential) <2 ?V".
They use NAIS "TXS2SL-4.5 H11" The H11 may be some manufacturing code, not the -1 option as the scanners are rated 1A.

[janaf]

From the Keithley Model 7168 JFET scanner data sheet:
 
The Keithley Model 7168 is an 8-channel, 2-pole card with <30nV of thermal offset.
MAX. SIGNAL LEVEL: 10V, 50mA peak (resistive load only).
CONTACT RESISTANCE: <12W.
CONTACT POTENTIAL (HI to LO) BETWEEN CHANNELS: <30nV when properly zeroed with supplied leads (see manual for recommended procedure). Typically <60nV without
zeroing.
CONTACT TYPE: Solid state JFET switch.
INPUT LEAKAGE: <50pA per channel at 23°C.
MAXIMUM VOLTAGE BETWEEN ANY TWO TERMINALS: 10V.
CONTACT RESISTANCE: <1ohm.
CONTACT POTENTIAL: <200µV.
OFFSET CURRENT: <1pA (<30fA typical).

Tnd there is a teardown here at EEV
https://www.eevblog.com/forum/metrology/volt-nut-test-for-eevblog-readers/

[janaf]

i understand you want to use yours to compare references? Same here. Would like to compare a bunch of LTZ boards. Would like to measure differentialy between them. Skipping absolute values for now. Mine are within a 40mV span, so I'd use an instrumentation amp with decent gain.

My headache is how to measure difference with the low of the circuit is NOT=GND. They may differ by some (fraction of?) uV? I just may have to ignore that for now.

 

[Echo88]

Thanks for the data chuckb! Wouldnt have thought that the TX2-relays are this good. I dont intend to change the content of this thread, but would like to ask (since both TiN and chuckB have access to a K182(-M): Do any of you know, if the -M-suffix on the Keithley 182 has any meaning (couldnt found anything about it)?

[chuckb]

The Keithley Model 2000-SCAN is a 10 channel scanner card for the 20XX DVMs. It's specified at less than 1uv of offset. The Service information specifies TQ2E-L2-5V relays.

The card I used actually had NEC EA2-5TNJ Orange Relays, date code 1411.

[chuckb]

Attached is a plot of 3 days of evaluating a Panasonic TXS2-L2-12V-1 relay. This is a small 12v latching relay (L2) with an alternate contact material (-1). I have tested both contact materials in the same setup. The normal contacts seemed to have 5nV random jumps of the normal contact material. It was hours between the jumps. Either contact material is fine for Zener diode work when you have 1,000nV of noise anyway. The voltage measured is the difference in thermal EMF between the two contacts.

You use pulses to drive the coil in a latching relay, so for a few tests I kept power on one coil for 10 seconds. It may have built up 5 nV of extra offset. So it looks like a well designed relay for low signal use.

[Andreas]

Hello,

I have tested my relay multiplexer with TQ2-L2 relays.
Sorry no nanovolt meter available so I had to use the K2000 for this.
Measurement values averaged over one minute to reduce noise.

The multiplexer contains several daisy chained relay contacts.
Plus additional one relay to switch the output off (both poles).

Measurement is done at the output connector.
The active input of channel 1-7 is shorted at the input connector.

Channel 8 is a internal short to ground in the Multiplexer.
Channel 0 is without multiplexer (simply using the shorting connector to the measurement cable).

Result: no more than 0.6uV between the measured channels.

with best regards

Andreas

[chuckb]

The next project will be to test the Electroswitch C4 rotary switch later this month.

Thermal EMF evaluation of an Electroswitch C4 rotary switch.

The switch and A10 nanovolt preamp were located in an insulated enclosure. This enclosure reduced ambient temperature variations and the rate of temperature change. There was no active temperature control.
I used 22 ga solid OFHC copper wire to short the switch poles together. Then the wiper contact was directly connected to the copper input terminal of the preamp.
It looks like a single contact of the C4 switch had a 50nV / deg C thermal EMF.

When the switch was wired differentially the thermal EMF was less than 5 nV / deg C.
When the preamp input was shorted with the same OFHC copper wire the preamp drift with temperature was less than 1 nV / deg C.
The Electroswitch C4 looks like a good low themal EMF solution when it’s wired differentially.

Timeline for the attached pdf.
0-12 hours, C4 switch in position 1, a single contact was measured, there seemed to be 50nV / deg C thermal EMF
12-25 hours, the leads to the preamp were manually reversed to isolate the thermal EMF of the switch from a preamp offset. The offset voltage is generated by the switch and wiring.
25-26 hours, the C4 switch was selected to position 2. An extra 10nV of offset was introduced.
26-37 hours, Preamp input shorted to determine offset voltage and temperature coefficient. The room temperature was reduced 1.5 deg C with maybe a +1nV change in preamp offset.
37-53 hours, The C4 switch was rewired to operate differentially. The offset voltage was less and the temperature coefficient was greatly improved. Note: Between hours 37 and 39 it took several hours for the switch contacts to equalize after being soldered.
53-60 hours, The C4 switch was in Position 2.
60-84 Hours, the C4 switch was in position 3.
84-100 hours, the C4 switch was back in position 1. Overall the thermal EMF of the C4 switch was within a 10nV for 60 hours, when operated differentially, even with a 1.5 deg C temperature change.

[doktor pyta]

I hope You have seen this: https://doc.xdevs.com/doc/_Metrology/qian2016.pdf

[Muxr]

Forgive me if this is a wrong topic, but I think it's related. I am interested in building a scanner/switch, and I was wondering. Has anyone thought about the feasibility of using chopper amps like the LTC2057 as buffers in parallel and leveraging their ^SD (Shutdown Mode).. to make a "poor man's" scanner?

From the datasheet:

The LTC2057/LTC2057HV features a shutdown mode for
low-power applications. In the OFF state, the amplifier
draws less than 11?A of supply current under all normal
operating conditions, and the output presents a highimpedance
to external circuitry.

Basically only one of the parallel LTC2057 would be on at the time, while the rest would have their ^SD pins pulled down (shut down mode), so channel selection would work this way. ^SD pins can probably be driven with a shift register.

I am interested in measuring drift of my various DC references ie. I am not even concerned about absolute accuracy. As long as I have linearity and resolution. Also the switching speed is not a concern since I would only measure a different channel every 5 seconds or so.

[zhtoor]

Has anyone thought about the feasibility of using chopper amps like the LTC2057 as buffers in parallel and leveraging their ^SD (Shutdown Mode).. to make a "poor man's" scanner?
[/b]

Hello,

welcome to the "Poor Man's Club" 

maybe start a thread "Poor Man's Scanner".

regards.

[Kleinstein]

Using the SD pin for a kind of scanner works, if  the amplifiers a wired for a gain of 1. The input bias on the inverting inputs would add up to load the divider - but this is not a problem with a gain of 1.  With a different gain setting here could be problems, as the input differential can be higher and this might cause higher input currents.

However the OPs input is only one side of the signals, so the scanner would be just one pole, not differential. In addition there should be input protection (and maybe some filtering) for the inputs. So it is more than just the LTC2057.

[Muxr]

Using the SD pin for a kind of scanner works, if  the amplifiers a wired for a gain of 1. The input bias on the inverting inputs would add up to load the divider - but this is not a problem with a gain of 1.  With a different gain setting here could be problems, as the input differential can be higher and this might cause higher input currents.

However the OPs input is only one side of the signals, so the scanner would be just one pole, not differential. In addition there should be input protection (and maybe some filtering) for the inputs. So it is more than just the LTC2057.

Thanks Kleinstein,

Yes I would use a buffer / gain of 1 configuration, I can definitely see the issue of using different gain settings because the inputs would then be coupled to the amplified bus across the amp even when they are shut down. Not to mention expensive resistors etc..

For the sake of simplicity I think single pole is fine.

I should study some DMM/Nanovolt Meter schematics to see how to implement a sensible protection.

Hello,

welcome to the "Poor Man's Club" 

maybe start a thread "Poor Man's Scanner".

regards.

I think I will design and build a simple 8 channel scanner based on this idea and share my findings in a new thread. Thanks!

[zhtoor]

Using the SD pin for a kind of scanner works, if  the amplifiers a wired for a gain of 1. The input bias on the inverting inputs would add up to load the divider - but this is not a problem with a gain of 1.  With a different gain setting here could be problems, as the input differential can be higher and this might cause higher input currents.

However the OPs input is only one side of the signals, so the scanner would be just one pole, not differential. In addition there should be input protection (and maybe some filtering) for the inputs. So it is more than just the LTC2057.

Thanks Kleinstein,

Yes I would use a buffer / gain of 1 configuration, I can definitely see the issue of using different gain settings because the inputs would then be coupled to the amplified bus across the amp even when they are shut down. Not to mention expensive resistors etc..

For the sake of simplicity I think single pole is fine.

I should study some DMM/Nanovolt Meter schematics to see how to implement a sensible protection.

Hello,

welcome to the "Poor Man's Club" 

maybe start a thread "Poor Man's Scanner".

regards.

I think I will design and build a simple 8 channel scanner based on this idea and share my findings in a new thread. Thanks!

how about a "flying capacitor" scanner based on some 8 channel differential multiplexer like MPC507A (or equivalent).

regards.

[Muxr]

how about a "flying capacitor" scanner based on some 8 channel differential multiplexer like MPC507A (or equivalent).

regards.

I did consider analog multiplexers but I am worried about large voltage and current offsets.

MCP507A says this in the datasheet:


Although my source impedance will be much lower, I thought going with a tried and true chopper op amp was a known quantity. Worth of consideration though, thanks!

[Kleinstein]

Unless the Bias currents are really high, the offset voltages due to the R_On are not that high: Typical R_on is more in the 100 Ohms range. it is only the protected types that have extra R_on.
The leakage current is typically lower, 10 nA is over full temperature range.
When measuring low voltages one usually does not have the full +-15 V across the MUX - this also helps to reduce leakage.
A mux with only 8 Inputs would also have less leakage (e.g. half).

If you are into very low voltages, one might not need the highest gain precision. So something like a MUX / preamplifier might be still very useful, even if the gain is not super stable. The LTC2057 noise level is lower than some of the DMMs - of causes this also due to not having much protection.
So an protected MUX like the MPC 508 (8:1)  and an LTC2057  with some optional gain might not be such a bad solution. It has it's limits though: gain drift and accuracy, bias currents, and limited maximum voltage.

The advantage of relays is usually the higher voltage capability and less leakage - though thermal EMF can be a problem. Especially it is hard to get good data / estimate on the thermal EMF. Of cause the thermal EMF is not just depending on the relay, it also depends on the thermal setup. Another problem with many relays is the not so well defined performance at very low currents - no such problem with CMOS switches. Getting at least some data on switches / relays it the purpose if this thread.

[Andreas]

Using the SD pin for a kind of scanner works, if  the amplifiers a wired for a gain of 1. The input bias on the inverting inputs would add up to load the divider - but this is not a problem with a gain of 1. 

Hello,

be carefully!!!

Like many precision OP-Amps the LTC2057 has input protection diodes between the +/- inputs
So you can get a very low impedant path between the output and the postive input even when the output is shut down in a unity gain configuration.

Life is not easy, but you can easily kill your LTZ1000.
I had only luck that I usually use a capacitive load protection cirquit on all my buffers.

with best regards

Andreas

The next thing is: the output leakage is not specified.

[Muxr]

Thanks for the input and advice guys. When all is considered, a mux seems to be the right way to go in order to keep everything simple. Will see how it goes.

[Andreas]

Hello,

on my LM399 ageing box (with floating supply voltage) I use MAX4051A multiplexers in a differential configuration.
They are low ohmic and have relative low leakage. (but are limited to 16V total supply voltage)

with best regards

Andreas