DIY Low Thermal EMF Switch/Scanner for Comparisons of Voltage and Res. Standard

[zlymex]

1. Why Switch/Scanner
Comparison of standard(voltage/resistor) is precise, repeatable, low Thermal EMF.
Comparison can be easily made automatic.
Multi-reading results make statistics/averaging at ease and achieve high confident level.
Standards to be compared may be relatively permanently connected thus inconsistency due to contact variation is greatly reduced.
May compare multi-standard(ex. up to 16 or greater) for a long time.
Derived use i.e. measurement of temp-co.


2. Why DIY powerless Switch/Scanner
Very simple to use. No programming, no GPIB.
Powerless, and therefore no direct interference from the mains.
Powerless, and therefore very little heat generated and less Thermal EMF.
Compact, low cost.
No power switch, consume virtually no power when idle.

[zlymex]

3. Features and Principles of the DIY switch
Provide 2*2(DPDT) or 4*2(4PDT) switch, or 2*4, 4*4, 2*8, 4*8, 2*16, 4*16 switch.
MC control(Multi-meter Measurement Complete Signal).
MC power supply, consume about only 0.3mA current.
Use 74HC CMOS and related low power IC.
Use Latching relay to achieve low power consumption.
Manual switch and free running mode provided.
Operates in loop manner only.

MC or VMC signal is usually provided by precision multimeters as a standard feature.
It usually stay logic high when idle or during measurement, and goes to logic low briefly when a measurement is complete.

When measurement by the multimeter is complete, it's the perfect time for switch/scanner to operate.

Newer multimeters use CMOS IC for VM output thus the high level is nearly 5V. Old ones such as my Agilent 3458A (made in 2006) uses TTL IC which gives only 3.65V when drawing 0.5mA current. However, this is enough for the switch to operate properly.

Although is mechanical, the relay will complete it's switch over in less than 5ms, plus settle down of the measured V or R, which normally within the default delay time of a typical multimeter. For 3458A 10V range NPLC=50, default delay is 10ms. This delay may be increased if needed.

[zlymex]

4. Schematic of My first 4*2 switch for two standards(V or R) comparisons


The Latching relay, Omron G5AK-237P





Omron G5AK-237P, DPDT low signal double-winding latching relay.
Data sheet is here http://datasheet.octopart.com/G5A-237P-DC5-Omron-datasheet-109999.pdf


The Drive
Normally the set or reset coil operate at 5V individually.


However, the set and reset coils are identical and can be connect in series or in parallel. When in series, it also operate at 5V but consume half the current.


When paralleled. it will operate at 2.5V but the current will be doubled. This is the way I use.



Power supply



I use HT7136 low dropout CMOS low power regulator. Connect a 200 Ohm resistor and a Schottky diode to VM, and output is 3.5V minimum through the use of a 470uF filter capacitor.
DS of HT7136 is here: http://www.holtek.com/pdf/consumer/71xx_1v170.pdf

The Principle is straight forward, every time a measurement is complete, VM goes low briefly toggle the J-K flip-flop and hence the relay.


Optional trigger is also provided for manual momentarily switch K or by activating the free running astable vibrator.

[zlymex]

5. Implementation and testing
Breadboarding first


Then assemble the device on universal board




Cat. 5 twisted pair, pure copper and screened, good for this purpose.





Test is performed by making measurements of two inputs alternatively


Sampling is done by another DIY interface device:




One end of the device is connect to multimeter GPIB, another end is connect to PC's USB port by a RS232-USB converter(early version) or direct USB connection(later version). This device is designed by llycom and is widely used among Chinese volt-nuts.

By running the dedicated program, one can send any GPIB command to the multimeter.




It also provides a precision temperature sensor(DS18B20).

The final graph is the test result when two inputs shorted. Ideally the dark blue line and purple line should be equally flat to zero but it's not owing to multimeter's noise and offset. However, the average of the two lines are quite the same with less than 5nV difference, well below the noise level of 3458A.

[zlymex]

6. Next version of the 2*4 switch

First 8 photos show the implementation of PCB version by Mytek, last 4 photos show the miniature version.

[zlymex]

7. 17*2 scanner for up to 17 voltage standards comparisons

[zlymex]

8. 16*4 scanner for 4 wire standard resistors comparisons

believe or not, 32 latching relays and it's related circuitry power by multimeter VM and consume as little as 0.2mA (When 3458A choose NPLC=50 and AZ on)

[zlymex]

9. Application examples

By applying very stable 100mA current thr two one-ohm resistors, 3458A reads two 100mV alternatively thus will compare to sub-ppm level.


Compare two 10V.
Blue line: 3458A measurement of a 10V standard (Warming up curve)
Red line: difference of two 10V standard plus 10V.


Standard resistors comparison chart and photo.



For those who can view 38hot.com properly, I have other scanner related topics:
Mechanical switch http://bbs.38hot.net/thread-1006-1-1.html
Overview of metrology scanners http://bbs.38hot.net/thread-1016-1-1.html
Data Proof 160A Scanner, Principle and modification http://bbs.38hot.net/thread-1082-1-1.html
DIY 16ch Powerless Scanner for Voltage http://bbs.38hot.net/thread-3658-1-1.html
DIY 16ch*4wire Powerless Scanner for Standard Resistor Comparisons  http://bbs.38hot.net/thread-4041-1-1.html
Precise measurement of alpha/beta temp-co of SR104 standard resistor http://bbs.38hot.net/thread-1352-1-1.html
DIY 100mA constant current source for comparison of 1 ohm standard resistor http://bbs.38hot.net/thread-17593-1-1.html
Tests on 7V and 10V voltage standards http://bbs.38hot.net/thread-645-1-1.html


<The End>

[quarks]

Great work, thanks a lot for sharing

Here is one of my scanner cards for Keithley 7001 with very nice COTO relays (see att. pic.)

on Reply #2 unfortunately I cannot see your pictures
edit: now it is the same on other replys also

on Reply #3 Where did you get the spade lugs?

on Reply #7 SR104-1, 742A + VHP101 look flat as expected, but what is the jump 11y/2W?

[zlymex]

Thanks quarks, that's is a very good switch board. My approach is to generate less heat to compensate not-so-good relays.

Reply #2 I see the photos fine, both reply #2 and others.

Reply #3 those spade lugs was a custom made group-buy in China, pure copper, later I changed to lugs with gold flashed.

Great work, thanks a lot for sharing

Here is one of my scanner cards for Keithley 7001 with very nice COTO relays (see att. pic.)

on Reply #2 unfortunately I cannot see your pictures
edit: now it is the same on other replys also

on Reply #3 Where did you get the spade lugs?

on Reply #7 SR104-1, 742A + VHP101 look flat as expected, but what is the jump 11y/2W?

[quarks]

Hello zlymex,

here is what I see (see att. pic.)


Those spade lugs look very good. Is it possible to buy some?
They would be a great addition to my
https://www.eevblog.com/forum/buysellwanted/diy-low-emf-cable-and-connectors/msg190302/#msg190302

[zlymex]

Hello zlymex,

here is what I see (see att. pic.)

Those spade lugs look very good. Is it possible to buy some?
They would be a great addition to my
https://www.eevblog.com/forum/buysellwanted/diy-low-emf-cable-and-connectors/msg190302/#msg190302

I see your reply pictures fine, but I don't know why my pics won't show up on your side.

Those spade lugs was bought 5 years ago and I was a group buy participant. Usually the seller will be long gone.

lly/2w is the DIY two-wire  standard resistor by llycom sent to me for test. I cannot recall what's inside as I did not open it for photo. I guess it's a single VHP101-10k.
Because it's two wire, there is a value jump when I disattach and reattach the lugs from the binding post during test.

btw, My VHP101 DIY consist of 4 matched resistor elements.

[quarks]

now I know it is IE, because with Operea I can see your pictures (still wonder what the cause is)

btw, My VHP101 DIY consist of 4 matched resistor elements.

Very nice!!!
Have you used 4x 10k values (2 in series and than same in parallel) or did you use 4x 40k in parallel?

[zlymex]

now I know it is IE, because with Oprea I can see your pictures (still wonder what the cause is)

btw, My VHP101 DIY consist of 4 matched resistor elements.

Very nice!!!
Have you used 10k values (2 in series and than 2 parallel) or did you use 4x 40k in parallel?

Here is the large photo with resistor values on it, 9k999.
Mine is 4 wire(actually it's 6 wire), but llycom's 2 wire.

[zlymex]

Talking about matched resistors, I selected the 4 from a batch of 40.

2 resistors may match alpha temp-co, it usually takes 4 to match beta as well.

[quarks]

Here is the large photo with resistor values on it, 9k999.

thanks again

Your board looks like ceramic. Can you share some details?
When I did some home work for a "as good as possible" DIY resistor reference, I bought some ceramic solder stands (see att. pic.).
I hope with this stands, it will be possible to also build high resistance values (I plan 1GOhm). Have you tried high resistance values also?

[zlymex]

thanks again

Your board looks like ceramic. Can you share some details?
When I did some home work for a "as good as possible" DIY resistor reference, I bought some ceramic solder stands (see att. pic.).
I hope with this stands, it will be possible to also build high resistance values (I plan 1GOhm). Have you tried high resistance values also?

Ceramic solder stands is a very good idea as I bought some as well.
I use Teflon board in hope it's better insulation. However IMO, ceramic is good enough for 10k even for sub-ppm level requirement.
For higher resistor value such as 1GOhm, still good if precision not exceed 0.01% and keep the surface dry and clean.
I tried even higher values like 10GOhm and 100GOhm, guarding must be applied even the insulation is Teflon.
What resistor element you plan to use for 1GOhm?
I once use 100 Dale RN55D 10MOhm for 1G.

[quarks]

I use Teflon board in hope it's better insulation.

thanks, that makes it much easier to drill the holes

However IMO, ceramic is good enough for 10k even for sub-ppm level requirement.
For higher resistor value such as 1GOhm, still good if precision not exceed 0.01% and keep the surface dry and clean.

good to know

I tried even higher values like 10GOhm and 100GOhm, guarding must be applied even the insulation is Teflon.
What resistor element you plan to use for 1GOhm?
I once use 100 Dale RN55D 10MOhm for 1G.

Yes, guarding is a must for high resistance values (att. are some details from Keithley Low Level Meas. Handbook)

Until now I have bought OHMITE MOX 500MOhm (but only 1%), several Caddock USF 0.01% (but only 10MOhm, which will be extremly expensiv when I need 100 pieces) and I have a sufficient large batch of promissing Caddock MG745 40MOhm 0.1% on order.

[zlymex]

until now I have bought OHMITE MOX 500MOhm (but only 1%), several Caddock USF 0.01% (but only 10MOhm, which will be extremly expensiv when I need 100 pieces) and I have a sufficient large batch of promissing Caddock MG745 40MOhm 0.1% on order.

That's great
While I cannot find Caddock USF(or cannot afford), I bought 100+ used Caddock TF020R-3 20MOhm instead.

Another expensive option for 1G or 10G maybe Supper MOX(MOX910)

Your blue Omite is Mini-Mox, I also bought 20+ similar Slim-Mox 104 500MOhm for 1G and 10G.

[quarks]

Super MOX 910 looks also good

do you know by chance what is used in the Fluke 8508A-7000K calibration kit (see att. pic.)?

[Vgkid]

Well done.
This will be a rather interesting thread to follow.

[quarks]

OT: I indeed normally do not use IE with my private PCs.
But with my business PCs I am forced to use it, so sometimes I open eevblog with IE

[Mickle T.]

This idea has been widely used in the Prema DMMs. I used a similar scheme to control the sealed relay (RPS-25) in the selfcal module of the 8.5-dig. DVM.

[Andreas]

Hello Lymex,

thanks for sharing also from my side.

There are many good ideas in the cirquits.
Especially using the DMM trigger as power supply was a fact that
I could not see from the link on volt nuts to 38hot.com.

8. 16*4 scanner for 4 wire standard resistors comparisons

believe or not, 32 latching relays and it's related circuitry power by multimeter VM and consume as little as 0.2mA (When 3458A choose NPLC=50 and AZ on)

I guess it is only for resistors and not for reference voltages.
Or do you have a possibility to get into some "reset" state for the relays after "power on".

The bad news is that Mouser is saying that the OMROM G5AK-237P is "discontinued"    ...

But the good news is that I found an alternative relay, the Panasonic TX-S series  ::

I am using Panasonic/Matsushita TQ2 series (the TN2 series would work also) running for about 1-2 weeks from a 9V block including microprocessor.

With best regards

Andreas

[splin]

The bad news is that Mouser is saying that the OMROM G5AK-237P is "discontinued"    ...

But the good news is that I found an alternative relay, the Panasonic TX-S series  ::

I am using Panasonic/Matsushita TQ2 series (the TN2 series would work also) running for about 1-2 weeks from a 9V block including microprocessor.

With best regards

Andreas

Panasonic specify a minimum switching load of 10uA, 10mV which might be a problem for comparing references - but probably not for a few thousand or 10s of thousands cycles! They recommend the SX series for low signal levels but even those specify a minimum of 10uA, 1mV.

I know that they ran into this problem at a company I used to work for - minimum loads aren't always quoted for relays but some contact materials rely on the current disruption to clean the contacts.

I wonder if the contacts get noisy over time or just suffer increasing contact resistance?

Are reed relays better in these regards being hermetically sealed?

The thermal EMF spec is for the TXQ is 5uV and .3uV for the TX-S but those are for the non-latching parts. Are you using latching versions and if so have you measured the thermal EMFs (or tried - I guess it's quite tricky to do)?

Thanks,  Splin

[Vgkid]

What about switching a resistor to ground previous to making the measurement. Or as part of a self test sending a current through all of the relays.

[Andreas]

The thermal EMF spec is for the TXQ is 5uV and .3uV for the TX-S but those are for the non-latching parts. Are you using latching versions and if so have you measured the thermal EMFs (or tried - I guess it's quite tricky to do)?

Hello,

I use the TQ2-L2-5V version (the latching type). Otherwise a 1-2 week operation would not be possible from a 9V block.
picture is here: https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/msg567203/#msg567203

I am not able to measure thermal EMFs reliable since my 24-Bit ADC is quite noisy. (1uVpp with a integration time of 1 minute).
A partly compensation of the thermal EMFs is done in my setup since I am doing a offset measurement through the relays.
I can actually switch the +/- signal of the output to every + pin of the references or to the common ground. So its possible to do difference measurements between 2 references also.

Some cleaning of the contacts is also done by daily switching of the relays.

With best regards

Andreas

[quarks]

finally received the Caddock MG745 40MOhm 0.1% resistors.
In my first checks they are well within spec and spread positive and negative in value.
That should help in combining them to a my specific values.

[timb]

Wow, thanks so much for this thread! I'm working on a scanner myself and it never occurred to me I could do it self powered!


Sent from my Tablet

[timb]

By the way, these might be a nice stand-in for those Omron relays: http://www.mouser.com/ProductDetail/TE-Connectivity-Axicom/V23079B1201B301/

$3/ea, 5V, DPDT 2-coil bistable in through hole. According to the datasheet thermal emf is <10uV, which is as good as the baseline version of the Coto Low Thermal relays (much better I imagine though as they are latching so no internal heat generation).


Sent from my Tablet

[Gyro]

It would be prudent to stock up on Mercury wetted bistable reed relays for this kind of use while they are still available on ebay / surplus market.

They're the only contact technology capable of truly dry switching, and with stable contact resistance and a cycle life long enough to never have to worry about.

Edit: ...and of course zero bounce.

[branadic]

The scanner of the Prema 5017 SC uses Omron G6AK-234P, just for the info. The scanner is specified with thermo forces typ. ±1µV, max. ±2 µV after 1,5 h warm-up. The Prema 2080 scanner, that I had, is specified the same way using the same relays.

http://ohh.de/5017.htm
http://ohh.de/2080.htm

[Mickle T.]

Prema has a nice and bulky series 2000 of precision scanners. The top MUX 2010, that I had, have a max 50 nV of thermo forces.

[branadic]

Prema has a nice and bulky series 2000 of precision scanners. The top MUX 2010, that I had, have a max 50 nV of thermo forces.

Never heard of that model, the only info I found is the manuel:

http://www.prema.com/index.php/de/service/messgeraetewartung/item/download/116_124e71e4309f5edffbaa2bfc370ef7b5

Smells like a teardown and a lot more photos of your unit, cause less then 50nV (measured on the plugs) is very interesting for the volt nuts here

[dr.diesel]

Smells like a teardown and a lot more photos of your unit, cause less then 50nV (measured on the plugs) is very interesting for the volt nuts here

And another saved eBay search!

[doktor pyta]

Inspired by Zlymex, I started to design my own scanner based on the same idea.
The goal is to make some improvements:

1. indicate active channel's number (1 of 16 LEDs). The measurements are made very slowly.
       If there is some bad connection etc. it will be easy to identify.
       I tested green LEDs that work satisfactory biased with 50uA of current.
2. provide Manual control of the channel
3. provide power-on reset (quite tricky part...)
4. use RJ45 connectors, so the user just cuts off-the-shelf CAT6 cable in two pieces and strips wires only on one side.
      This also makes testing of relays easy - just disconnect RJ45 cables and plug in RJ45 plugs with pairs shorted by loop of wire.
      The RJ45 connectors will be hidden inside metal enclosure to ensure thermal equilibrium.
5. use SMD components (prepared to reflow soldering, better EMF performance expected)
6. the board should fit standard TUFxx Fischer Elektronik enclosure
7. It's nice to have a spare relay on board (tribute to Data Proof scaners)
8. the design should be expandable to 2 wire 32 channels; 4 wire 16 channels; 4 wire 32 channels by using the same boards (with some PCB jumpers, desoldering of some components but basically same design)
9. last but not least: the scanner shall provide break-before-make action not to disturb expensive gear and save the relays.


The quiz is: Open the schematic pdf file and find as many bugs as possible.
If anyone is interested in factory made PCB for this project let me know ->PM.
Zlymex has a free PCB if he wants

[beanflying]

Interested! Weighing up another 6 1/2 - 7 1/2 Meter or a 34970A and boards. 

[doktor pyta]

First bug found: net NE0 added to J5 connector.
Without it the slave board would not break-before-make.
Schematic upgraded to rev 0.2

Edit:
SW1 pin 6 connected to wrong node.
Schematic upgraded to rev 0.3

Edit: Switched from 74HC4538 to 74HC123 for two reasons:
4538 has built-in power-on reset circuit which would disable generation of a pulse after turning scanner on. This is not wanted in this circuit.
4538 draws much more idle current.
Schematic upgraded to rev 0.4

[beanflying]

Now Interested for interests sake. Santa is bringing me an Agilent Box 

[cellularmitosis]

Heyo,

I've been thinking about scanner switching arrangements.

What I'd like:

• Vrefs are switched to (floating) battery supply while being measured
• DMM can measure a single Vref or the difference between two Vrefs

Here's what I've come up with (attached).

You can't compare any Vref against any other Vref: they are separated into two groups, and any Vref from group "A" can be compared to one from group "B".

With this restriction, you can get away with only two DPDT relays per Vref, which is pretty cheap.

Here, four Vrefs are shown, but the idea is that any number of Vrefs is supported.

The DMM also uses two SPDT relays and can do the following:

• Measure CH A to CH GND (i.e. measure a single Vref from group "A")
• Measure CH B to CH GND (i.e. measure a single Vref from group "B")
• Measure CH A to CH B (i.e. measure the difference between a Vref from group "A" and group "B")

Just to walk through some examples:

To measure Vref 1:

• Vref 1 input is switched from SUPPLY to BATTERY A
• Vref 1 output is enabled, placing it across CH A - CH GND
• DMM inputs are switched to CH A and CH GND

To measure Vref 2:

• Vref 2 input is switched from SUPPLY to BATTERY B
• Vref 2 output is enabled, placing it across CH B - CH GND
• DMM inputs are switched to CH B and CH GND

To measure Vref 1 against Vref 2:

• Vref 1 input is switched from SUPPLY to BATTERY A
• Vref 2 input is switched from SUPPLY to BATTERY B
• Vref 1 output is enabled, placing it across CH A - CH GND
• Vref 2 output is enabled, placing it across CH B - CH GND
• DMM inputs are switched to CH A and CH B

Edit: the caps on each Vref are large enough that they can be switched from SUPPLY to BATTERY A/B without losing power.  This is a somewhat simplified schematic, as there would probably a regulator after each capacitor, keeping a stable voltage supplied to each Vref.

Edit 2: to elaborate, Vrefs 1, 3, ... are in group "A" and Vrefs 2, 4, ... are in group "B".  So you can measure Vref 1 against 2 or 4, but not against 3.  To get the difference between 1 and 3, you'd have to do it indirectly by measuring 1-2 and 3-2, then calculate 2-3.  With a large number of Vrefs, many cross-checks are possible, so errors in the system should become apparent.

Edit 3: the batteries can be charged, but all Vrefs must be on SUPPLY, so no (floating) measurements can take place while the batteries are being charged.

Edit 4: I wonder if this thread should be moved to metrology?

[cellularmitosis]

doktor pyta, I no longer see the pdf of your schematic, was it removed from your post?

[doktor pyta]

Hi cellularmitosis,

it was weird situation when only I was attending my own quiz of searching for bugs 
Besides the thread was filled with schematics which all contained bugs.

You will find rev1.0 schematic in the attachment.
Take a look on the list of operation modes.
*I don't know if anyone ever will use 8 wire mode but it is also available.

EDIT: new file loaded: operation mode table improved for clarity.

[Echo88]

Dont tell me that scanner-mainframe you just bought with the load of 20-relay-cards dont suffice cellularmitosis 

[cellularmitosis]

Dont tell me that scanner-mainframe you just bought with the load of 20-relay-cards dont suffice cellularmitosis 

The scanner gets me started immediately with something that works, but does no good for the community!  Ideally I can produce some smaller scanner kits, benchmark them against the Keithley cards with the COTO relays, and publish an open source design to help out fellow volt nuts

Buying old gear is nice, but reproducing equivalent results in a DIY project is a lot more satisfying

[Echo88]

Hmm, since my old Keithley 7064 card produces unstable readings in some configurations i would be interested in some new cards. The easiest solution for me would be designing keithley 705-compatible cards (defined lengths and a 15pin-card-edge-connector), but then i have the problem that they were build to drive non-latching relays and i would have to add drive logic on the cards to use latching-relays. Meh.

[cellularmitosis]

Just whipped up a tiny board for a single resistor which uses machine-pin headers for a pluggable 4-wire connection.  It has multiple footprints, so it will accept a vishay foil, a wirewound, etc.

Now, to design a scanning motherboard to plug them into...

https://github.com/pepaslabs/resistor-plug

https://oshpark.com/shared_projects/hp93BAlQ

[splin]

Just whipped up a tiny board for a single resistor which uses machine-pin headers for a pluggable 4-wire connection.  It has multiple footprints, so it will accept a vishay foil, a wirewound, etc.

It would be better to have the sense connections next to each other to minimise any temperature difference and thermal EMFs.

[dr.diesel]

Dont tell me that scanner-mainframe you just bought with the load of 20-relay-cards dont suffice cellularmitosis 

The scanner gets me started immediately with something that works, but does no good for the community!  Ideally I can produce some smaller scanner kits, benchmark them against the Keithley cards with the COTO relays, and publish an open source design to help out fellow volt nuts

Buying old gear is nice, but reproducing equivalent results in a DIY project is a lot more satisfying

An open low thermal scanner would be a great community asset.   

I'll offer my Data Proof 160 (still unproven, but works) for benchmark as well.

[doktor pyta]

Update. Try to find a typo (engraving company fault  )

[2N3055]

Update. Try to find a typo (engraving company fault  )

C'mon, what's a two CH16 between friends...  ?     

[doktor pyta]

Some photos showing details of RFS135B.

[doktor pyta]

++

[doktor pyta]

Just before putting into enclosure.

[branadic]

Time to measure the performance like thermal emf Nice build.

-branadic-

[2N3055]

I'm making doktor pyta scanner too.. Two boards for a 16ch/4W scanner config..
All parts are here and ready. Time to get busy.

[2N3055]

After few evenings of soldering i assembled scanner..
It has two boards, so I can do 4 wire x 16 channels..

All parts were easily ordered from TME and Farnell.
Boards are good quality, switching part with relays is isolation routed form rest of board..
Power consumption is very low, so no problem for DMM logic output to drive it...

Now I need to make cables. 
Will be testing it with my 7562 Yokogawa (I'm waiting for a connector for Yokogawa, it doesn't use BNC) and Rigol DM3068.
I will also do some timing testing to see what are parameter boundaries..

Regards,

Sinisa

[cellularmitosis]

Very cool!  Where did you order the kit?

[plesa]

RJ45 is OK for low EMF scanner? What type you have?

[2N3055]

It's not really a kit. It's Doktor pyta scanner from above, I just got PCBS , and front and back panels for a standard Hammond box from him. His schematic is above in this topic..
Regards,
Sinisa

[2N3055]

Hi Plesa, long time no hear...

Yeah, it looks too easy, but...

That connector is very compact, and isothermal by virtue of that, it is shielded (also acting metal thermal spreader on the outside). Contacts are very close, made of gold plated beryllium copper, very low thermal mass, attached directly to board, being isothermal there... Also SFTP cable is twisted pairs, pure copper, individually shielded with aluminium, that acts as shield and thermal spreader...

Generally, doktor pyta went with making everything as isothermal and low mass as it can be, instead of relying on exotic components. We shall see how well it does, but in his measurements so far, he got double digits nV TEMF..
I don't have equipment at the moment to measure whatever TEMF it makes now.. My crystal ball sees more equipment in my future... Nanovoltmeter would be nice....:-)

What these connectors are not is that they are not low leakage and are SELV only (max 60V)...
I would think it cannot be used for precision measurement of 100 Gohm with 500V..

My use for this scanner is measuring voltage refs (up to 20V), 4W resistor comparisons (up to 10Mohms), and 4W PT100/1000 and precision thermistor measurements. It should do this with less influence than 6.5 digit DMM can see.
That is more than OK for me now..

Of course, I will keep you guys posted.

Regards,

Sinisa

P.S. This is the connector type..

[2N3055]

I did a little test, scanner works with voltages of more than 4,5 V and low going pulses from longer than cca 950 nsecs.

So 4.5V and up and more than 1 usec pulse going low works reliably.

When I make cables I will also measure brake-before-make interval too...

Regards,

Sinisa

[doktor pyta]

Hi Sinisa,

few observations after some time of operation:

1. if You change D1  (BAS86) to eg. LL41 You may reduce minimum operating voltage by approx. 200mV.
2. as I warned before custom cable must be prepared. This is due to wire allocation in standard Ethernet cable (see RJ45 ethernet cable pair drawings in google)
3. Power-On-Reset is weak point of the design. It allows to charge capacitors and turns on channel 1. The problem is that You never know which channel remained active at the end of previous operation cycle. The best ad hoc solution after powering is to press NEXT button for 16 or 32 times to make sure all relays are in known position.
4. current revision of  the schematics below.

[2N3055]

Hi Sinisa,

few observations after some time of operation:

1. if You change D1  (BAS86) to eg. LL41 You may reduce minimum operating voltage by approx. 200mV.
2. as I warned before custom cable must be prepared. This is due to wire allocation in standard Ethernet cable (see RJ45 ethernet cable pair drawings in google)
3. Power-On-Reset is weak point of the design. It allows to charge capacitors and turns on channel 1. The problem is that You never know which channel remained active at previous operation cycle. The best ad hoc solution after powering is to press NEXT button for 16 or 32 times to make sure all relays are in known position.

1. Yes, of course..
2. Yes, I didn't mention this. Cables are not made as standard ethernet cables, were pairs are intermixed. Here pairs are in PAIRS, so cables have to be crimped 1-2  3-4  5-6  7-8 .. That's why you have to make them otherwise I would just cut few ethernet cables in half.
3. I did notice that from schematic. I also went for a scan through channels once on startup and then reset.... Not a problem for now. I have some ideas how to work around that . I'll let you know...

Thanks for your help..

Regards,

Sinisa

[MiDi]

When thinking about low thermal emf scanner and building my own, I came across with thermocouple scanner cards from Keithley (7402, 7057A) rated at <1µV thermal EMF.

They use Coto 3400-0066 dual reed relays which could be predecessor of low thermal EMF 3600-Series.
Connection to a µC should be quite easy as the coils are routed to edge card connector directly and are working with 5V supply.
The signal path is fully guarded - not shure if this is good or rather bad for this purpose (capacitance between guard and signal).

At first look they seem to be suitable for the job, what do you think?

Keithley 7057A:


Keithley 7402 top


Keithley 7402 bottom

[niner_007]

this is an interesting idea and an interesting project, I'm working on something similar but my goal is a little bit different, I just need a compact (think 5cm x 8cm) standalone scanner card, in my case I need it for monitoring cell voltages and general power electronics projects, likely an handheld DMM (289, U1270 etc) will be used as well; somehow I independently end up choosing the exact same relays, G6K, but I'm using an MAX4821 relay driver, all controlled by an STM32 with an USB interface; power is supplied over USB; the general idea is that the scanner, and the DMM (over GPIB, or USB/serial etc) are both connected to a raspberry pi, which does the logging and uses the scanner to cycle the channels, then read the DMM samples

[dkozel]

Thanks to everyone who's shared information here! I'm interested in switching four references between two DMMs so i suspect I'll have to make something custom based on all the parts and circuits described here.

doktor pyta, thank you for sharing your design. Have you or 2N3055 measured the thermal EMF in use? Any chance you have more PCBs available, I could do the end plates of the enclosure myself.

MiDi, that looks like a really interesting option. Did you end up getting one and testing it?

niner_007, How many inputs are you designing for? I'm also using a RPi as the brains of the operation and a USB based scanner would be ideal. I'm considering getting the Keithley Thermocouple card that MiDi mentioned and designing a little STM board to interface with it, but it would be much better to have a complete design.

cellularmitosis, did you continue on with your design?

[MiDi]

MiDi, that looks like a really interesting option. Did you end up getting one and testing it?

Got two 7402 a while back, but since then I had other things going on, so no test so far.
Want to build custom case with custom drive incl. an Arduino nano.
Have to make design and order parts, so this will hopefully go for christmas holiday...

[niner_007]

I'm working on a 8 independent channels one

[niner_007]

Here is some progress on my scanner card, this one is not designed for low EMF, I'm working on a similar design, with same features, larger size and with COTO 3500 relays with < 0.5uV EMF noise

Some features:

- tiny size
- USB powered
- communication/control over USB
- daisy channable, one scanner via 1 USB port can control up to 10 other scanner (power included), thus expandable to 80 or more channels
- outputs can be connected

[dietert1]

Another DIY low thermal scanner including EMF measurement is here: https://www.eevblog.com/forum/metrology/scannermultiplexers-for-voltage-references/msg3569019/#msg3569019.

Regards, Dieter

[branadic]

I this week received a low thermal scanner from doktor pyta. I wonder if someone with such a device has already implemented a remote control for a fully automated setup?

-branadic-

[branadic]

This is my take on the "DIY Low Thermal EMF Switch/Scanner for Comparisons of Voltage and Res. Standard".
In Dezember of last year I received a unit from doktor pyta, again, thanks for that. It took me a while to carefully prepare the test setup and to perform the measurement.
The scanner was configured as 4x16 (jumper D and F on Master PCB, jumper F on Slave PCB).

I equipped the output cable with a LEMO FFA.1S.304.CLAC62 connector that I had at hand, with the top board of the scanner connected to channel A and the bottom board connected to channel B of a 34420A Nanovoltmeter.
To control the scanner I've added a level shifter based on 2x 10k resistor and 1x 2N7000 to my Raspberry Pi and used one of the GPIO pins to control the scanner via Python:

   import RPi.GPIO as GPIO
   GPIO.setmode(GPIO.BCM)
   pin = 17
   GPIO.setup(pin,GPIO.OUT)
   GPIO.output(pin, GPIO.HIGH) #set Pin HIGH (ON)
   time.sleep(0.2)
   GPIO.output(pin, GPIO.LOW) #set Pin LOW (OFF)

while capturing the readings of the 34420A via GPIB (20 NPLC).
The 32 input channels of the scanner were shorted using clean bare wires. I twisted together HI/LO of each channel and crimped them with some 10 mm long copper tubes.
I then ran a program capturing 20 samples per channel before switching to the next and repeated the round 20 times.

Attached are very first results, blue channel A of the 34420A and red channel B of the 34420A, with each additional round plotted in lighter color.
We can see some offset and an offset difference between channel A and B of the 34420A, while it is yet unclear where this offset and the difference between both is coming from, could also be an offset from the Nanovoltmeter itself, I have to check on that.
Apart from that yet unkown offset, the differences between channel 1-16 of top and channel 1-16 of the bottom scanner board are fairly small. Note, these are just very first results and not a comprehensive study or analysis, but the results look pretty amazing to me.

-branadic-

[iMo]

Double check your cables (those twisted pairs) are made of copper (a lot of cheap copper clad aluminum used in those patch cables).. 

[branadic]

No worries here, the twisted cables are bare copper I need to create a short to rule out the initial offset, connectors are on their way.

-branadic-

[branadic]

I today created a short myself using a FFA.1S.304.CLAC62 connector. I then performed a short measurement, with the results below (100 samples @ 20 NPLC and 100 samples @ 100 NPLC, blue again Ch1 of 34420A and red Ch2 of 34420A).
I corrected the measurements taken before with the scanner at the 34420A and subtracted the measured offsets, result below too.
That is as far as I can go for this week.

-branadic-