Siglent Bench DMMs SDM3065X, SDM3055 and SDM3045X

[CDN_Torsten]

These work well: https://www.ebay.ca/itm/223470935306

[Kleinstein]

The 4 wire short from Ebay is not even a good one: the connection relies on the symmetry to be really 0 ohms in the 4 wire measurement. The conntectos look like they are made from a comboination of different metals. A simple DIY one bend from some 20 cm of 1.5 mm² bare copper wire is likely better.

At the 2 V level and more 6 digit performance the wire / connectors for the short should not jet be the major concern. It can be limiting when you go lower in the voltage (200 mV or 20 mV range).
A main point is keeping temperature gradients small. So a short wire and some cover to keep draft away can really help.
Things get only more complicated with a voltage reference - the short at least needs no power and can be very compact.

Much of the amplifier and ADC noise should be white noise. It can still be interesting to see where the additional low frequency noise (e.g. thermal fluctuatoins, maybe mains hum) comes in.

To see how much the reference contributes the test would be better in the 2 V range than in the 20 V range. So a divider from 10 V to some 2 V may still be a good idea. For just a noise test, a well insulated battery cell can also work. If kept at constant temperature and not mechanically disturbed a battery can be a pretty low noise voltage source. This is still a separate test, not replacing the noise test with a short.

[CDN_Torsten]

The 4 wire short from Ebay is not even a good one: the connection relies on the symmetry to be really 0 ohms in the 4 wire measurement. The conntectos look like they are made from a comboination of different metals. A simple DIY one bend from some 20 cm of 1.5 mm² bare copper wire is likely better.

Would you mind explaining what you mean by relying on symmetry to be 0ohms?  I don't quite understand...

As for the composition of the connectors, they appear to be gold-plated copper or brass.  I can also confirm that at 6.5 digits, the connector provides a very repeatable short.

[Kleinstein]

Looking closer at the 4 wire short, it is not even working right when symmetric, it has a principle problem in that the current between the drive terminals will no only flow the shortest way, but also take the long route around, close the the sense side. The crurde estimate is that maybe 1/8 the current would flow the long way and thus one may see a resistance of about 1/8 of a square (around 500 µOhms for 35 µm copper). So expect this "short to have about 60 µohms.

It would get much better if one would cut one of the 2 links connecting the drive and sense pairs, so that much less current would take a longer way.

Attached is a picture with the main current path draw in red, the unwanted part in orange and the area to cut in yellow.

With the cut the short is no longer turnable at will. So it is not a 4 way short. There are also designs for a 4 way short, but these are more complicated. This could be 3 contacts to the corners of a regular triangle and than the 4 the contact to the center.

[mawyatt]

The 4 wire short from Ebay is not even a good one: the connection relies on the symmetry to be really 0 ohms in the 4 wire measurement. The conntectos look like they are made from a comboination of different metals. A simple DIY one bend from some 20 cm of 1.5 mm² bare copper wire is likely better.

Would you mind explaining what you mean by relying on symmetry to be 0ohms?  I don't quite understand...

As for the composition of the connectors, they appear to be gold-plated copper or brass.  I can also confirm that at 6.5 digits, the connector provides a very repeatable short.

Also see Fluke 884X 4-Wire Short.

Best

[hammy]

easiest way I found to reproduce this: just switch on the meter, set to DCV (auto), press dual, then DCI, the meter clicks and hangs.

Confirmed on my SDM3045X (newest firmware).

@tautech
Any news about this from Siglent?

[mawyatt]

The 4 wire short from Ebay is not even a good one: the connection relies on the symmetry to be really 0 ohms in the 4 wire measurement. The conntectos look like they are made from a comboination of different metals. A simple DIY one bend from some 20 cm of 1.5 mm² bare copper wire is likely better.

Would you mind explaining what you mean by relying on symmetry to be 0ohms?  I don't quite understand...

As for the composition of the connectors, they appear to be gold-plated copper or brass.  I can also confirm that at 6.5 digits, the connector provides a very repeatable short.

Remember an old Fluke App Note about shorts, both voltage and resistance. Recall the voltage short is best served by a simple short thin copper wire with low thermal mass which quickly achieves thermal equilibrium with the DMM terminals and a resistance short is a more involved heavier/wider copper fixture like shown with the Fluke 884X.

The voltage short is simple in concept but we've found it somewhat difficult to get a coiled spring type thin copper wire to make good repeatable contact in the DMM input terminals, finally yielding to the technically less favorable but easier to use #12 copper wire shunted across a dual banana plug.

Best,

[oz2cpu]

here is my short, and the best type of banana i could find,
and the result is not zero volts :-)
but close, this is 10V range, 100 mains periodes, each measurement takes 3 sec,
and then 100 of them added, 153nV noise
 and ups.. a KS DMM, ya you got to have a little bit of other brand toys too

[CDN_Torsten]

Kleinstein & mawyatt - thank you both for your responses.

I do agree that for a 4W resistance short the layout is not quite ideal.  I would have removed the connection between the two sense terminals and I probably would have used thinner traces between Sense High and Force High as well as Sense Low and Force Low.  From a current path perspective this makes the most sense to me.

mawyatt - I am starting to warm up to your suggestion that the design of a resistance short and a voltage short could/should be different to help with thermal equalization.  The short I recommended takes quite some time to reach equilibrium...but it does (see image)


I still think that for the SDM series of DMMs - my recommended short certainly appears adequate and may be a quick way for someone to obtain a usable short.

[mawyatt]

here is my short, and the best type of banana i could find,
and the result is not zero volts :-)
but close, this is 10V range, 100 mains periodes, each measurement takes 3 sec,
and then 100 of them added, 153nV noise
 and ups.. a KS DMM, ya you got to have a little bit of other brand toys too

Have those bananas also, like better than the ones that allow the "cage" to spin. Also use the dual plugs, and use for both 2 and 4 wire shorts, they are convenient but none of these are up to "Metrology" standards.

Best be careful as the "Voltnuts" are now primed to rip me to shreds

Best,

[Johnny B Good]

Looking closer at the 4 wire short, it is not even working right when symmetric, it has a principle problem in that the current between the drive terminals will no only flow the shortest way, but also take the long route around, close the the sense side. The crurde estimate is that maybe 1/8 the current would flow the long way and thus one may see a resistance of about 1/8 of a square (around 500 µOhms for 35 µm copper). So expect this "short to have about 60 µohms.

It would get much better if one would cut one of the 2 links connecting the drive and sense pairs, so that much less current would take a longer way.

Attached is a picture with the main current path draw in red, the unwanted part in orange and the area to cut in yellow.

With the cut the short is no longer turnable at will. So it is not a 4 way short. There are also designs for a 4 way short, but these are more complicated. This could be 3 contacts to the corners of a regular triangle and than the 4 the contact to the center.

 Whilst that's all technically (academically) correct, with a test current of only 1mA, that 60 µohm short is going to be buried deep into the noise of even a good 6 1/2 digit bench meter.

 Also, I'd like to point out that this seller (kkrasimir.k), unlike his Chinese competitors, does provide a full and comprehensive description of his test plug assemblies :-

https://www.ebay.ca/itm/223470935306

 I've added this https://www.ebay.co.uk/itm/223610672208?hash=item34103c1450:g:fv0AAOSwMi5dx-VL to my shopping basket as well as this not so fully described cheap Chinese offering here:- https://tinyurl.com/ycknf6b3

 I haven't "Pulled the Trigger" just yet. What I'd really like to do is just  buy a few sets of those gold plated banana plugs that seller kkrasimir.k is using to mount directly onto 3mm thick solid copper heatsink shims (I'm not entirely convinced that the ones used by the Chinese sellers are as good, quality-wise - but I may well find out in due course ).

 I rather think that with a test current of only 1mA, it would be (technically at least) better to keep the sense shorting link isolated from the energising terminal shorting link when testing 4W resistance accuracy/calibration (or at least only tie the commons together if needs be).

 If anyone happens to know of a reasonably priced set of quality gold plated 4mm banana plugs, similar to the ones used by kkrasimir.k, I (and I suspect others) would be only too grateful for such information.

 In the meantime, I'll go in search of these 'fabled gold plated banana plugs' in what I suspect will be yet another hunt for "Unicorn Droppings" / "Hen's Teeth". If I manage to track down any reasonably priced contenders, I shall of course, report my findings back here.

[mawyatt]

Kleinstein & mawyatt - thank you both for your responses.

I do agree that for a 4W resistance short the layout is not quite ideal.  I would have removed the connection between the two sense terminals and I probably would have used thinner traces between Sense High and Force High as well as Sense Low and Force Low.  From a current path perspective this makes the most sense to me.

mawyatt - I am starting to warm up to your suggestion that the design of a resistance short and a voltage short could/should be different to help with thermal equalization.  The short I recommended takes quite some time to reach equilibrium...but it does (see image)


I still think that for the SDM series of DMMs - my recommended short certainly appears adequate and may be a quick way for someone to obtain a usable short.

Gotta love those old HP/AG34401As! Turn em ON, No Noise, No Fan, wait a couple minutes and you know the reading is right 

Best,

[mawyatt]

 Whilst that's all technically (academically) correct, with a test current of only 1mA, that 60 µohm short is going to be buried deep into the noise of even a good 6 1/2 digit bench meter.

Think the point was it doesn't cost any more to do this PCB correct in the first place!!

 I haven't "Pulled the Trigger" just yet. What I'd really like to do is just  buy a few sets of those gold plated banana plugs that seller kkrasimir.k is using to mount directly onto 3mm thick solid copper heatsink shims (I'm not entirely convinced that the ones used by the Chinese sellers are as good, quality-wise - but I may well find out in due course ).

 I rather think that with a test current of only 1mA, it would be (technically at least) better to keep the sense shorting link isolated from the energising terminal shorting link when testing 4W resistance accuracy/calibration (or at least only tie the commons together if needs be).

Please track down the mentioned Fluke app note regarding "shorts", recall they even showed how the massive shorting "Plate" doesn't perform as well as a simple yet technically correct short, example is the mentioned Fluke 884X. If one is to roll their own, this would be the concept to clone.

 If anyone happens to know of a reasonably priced set of quality gold plated 4mm banana plugs, similar to the ones used by kkrasimir.k, I (and I suspect others) would be only too grateful for such information.

 In the meantime, I'll go in search of these 'fabled gold plated banana plugs' in what I suspect will be yet another hunt for "Unicorn Droppings" / "Hen's Teeth". If I manage to track down any reasonably priced contenders, I shall of course, report my findings back here.

Though we had found such some time ago, until I held up a magnet

Good luck with your search , and please do inform if you discover some of these Droppings

Best,

[Kleinstein]

60 µOhms and 1 mA would result 60 nV. this is not much, but given some patience this can be seen, even with a 6 digit meter.
It is not that difficult to improve: cut at one side beween the sense and drive side (e.g. as indicated with yellow) and the residual resistance would go down quite a bit (maybe a factor of 100).

The shorts shown by Mawyatt look reasonable. The thick copper helps to get the connectors close in temperature. Otherwise I did not have good experience with this shape of banana plugs, but this was with normal cables and the same shape may come with different materials.

here is my short, and the best type of banana i could find,
and the result is not zero volts :-)
but close, this is 10V range, 100 mains periodes, each measurement takes 3 sec,
and then 100 of them added, 153nV noise
 and ups.. a KS DMM, ya you got to have a little bit of other brand toys too

A 100 PLC auto zero conversion should take 4 seconds in a 50 Hz mains country. Non AZ should be a bit faster, but is obviously not a good choice unless you have a very high impedance source.
153 nV of RMS noise would be quite good for 100 PLC, but not for the average over 100 such readings. Using an average over 100 reading that take 3 or 4 seconds would anyway be very slow to than still measure noise from.
So I don't hink the short is at fault.

For the noise analysis it may be more sensible to record data at 10 PLC (or comparable speed) and than calculate the Allan deviation from this.
The Sigilent meters use a SD ADC with a SINC3 or SINC4 fitler function - at least that is what the ADC supports directly. This filter implies readings that are not mutually independent, but slightly overlapping and this makes it a bit tricky to calculate the noise level. The Allan deviation would still be reasonable comparable for slightly longer times (e.g. > 5 samples time).  A poor quality fault would mainly contribute very low frequency variations and this a high Allan deviation for longer times, though such noise can also be intrinsic to the meter.

[mawyatt]

Here's a very simple model that shows kinda what Kleinstein is discussing regarding a 4 wire shunt.

Someone with lots of patience can analyze this, but a little intuition goes a long way here.

If the Meter function has a very high resistance and Ry approaches zero which is what a single point would yield. If Rz approaches a very high resistance which is by simple isolation, then all the other resistances don't matter much. This is the concept behind the 4 wire shunt with the 2 Dual Banana Plugs shown above, the only reason for the heavy cross wire is for mechanical support.

Best,

[2N3055]

Mike is right..

This is how original Fluke one looks.

If crosslink was right in the middle it would be ideal, but I guess it doesn't make much difference in practice.

Most important was to ensure that force currents never go into sense branch. If there is some minor (nV level) voltage drop imbalance from force current, it is seen as a common voltage on sense line and therefore suppressed.

That is for 4W resistance. 2W resistance sees one side copper resistance, that will be low enough compared with cables anyways.

[oz2cpu]

i still have not figured how it select the LO drive to be either 2.5V or 0.5V
and where it is originated from,
I think the top left stuff AGW633 is the 2.5V / 0.5V creator, but have not found the connection needed for this yet,
also i have not found the parts connected to the MAX REF 2.5V other than the two pins on the ADC as shown,
i dont plan to play with 2W or 4W ohm modes.

But a few questions : R4 is 100k !! WHY this will create a demand for AGW1408 and 633 to have impossible low leak,
those parts do affect the voltage signal in 20V mode, when heated


THIS SCHEMATIC IS FAR FROM DONE YET

[Johnny B Good]

@mawyatt

 Regarding getting the PCB 'right', I'm going to try bypassing this issue (maybe replacing it with a different issue) by using this bunch of gold plated banana plugs I just ordered from here:-

https://www.ebay.co.uk/itm/294120874481?hash=item447af861f1:g:zDYAAOSwF8Zgdrrp

which look ideal for mounting onto a piece of that 3mm thick copper heatsink spreader plate I mentioned.

 I removed the other two items from my shopping basket for now (and hopefully for the foreseeable future if my experiment works out), reducing my expenditure by almost 35 quid. At just £4.89 from a UK seller promising delivery by next Monday, it seemed well worth taking a punt.

 They can hardly prove worse than my home made 4mm ex-GPO banana plug shorting link's ~3μV performance. If they do prove to be worse, then we will have all learnt a valuable lesson about these particular fine looking gold plated banana plugs, won't we! 

 Having just retested with my home brewed shorting link, the problem seems to be a small temperature difference between the Hi and LO voltage sockets generating different thermocouple voltages between the shorting link's banana plug tips.

 Given enough time to regain thermal equilibrium, the voltage drops down to ~ -900nV. Auto-zero just confuses the issue by turning a -2.5mV into a 2.5mV reading or -8mV  into a 3mV reading. Obviously a minor re-calibration at the zero end of the scale is being called for. I get less confusing results if I just use 100PLC without auto-zero.

 If I reverse the shorting plug, I see a reading of ~-8mV which starts slowly ... er, no, actually this time it very slowly increased to -9.5mV! Time to swap it back round again... Gah! my banana plug shorting link now seems to be acting like a 6mV galvanic cell with just a soupcon of Seebeck effect thrown in for seasoning

 One thing to keep in mind when testing the SDM3065X is, despite its very low tempco at more sane voltage levels, it (and presumably the 3055 to a lesser extent) doesn't take kindly to sudden room temperature transients.

 I've been running thermal stress testing of my temperature stabilised Rubidium oscillator project all day. Starting at 19 deg C for the first couple of hours before turning the radiator on (and a 2.5KW room heater for the first half hour) to zoom the temperature up by 5 or 6 degrees before finally deciding, at 27 deg, that enough was enough by shutting the radiator off and flinging my hobby room door wide open to cool down to a less stressful temperature.

 Within just ten minutes, that 9.4μV  reading you can see in the attached photos I took shortly before, was dancing within less than a μV either side of the zero mark. Talk about becoming familiar with your test instrument's limitations and quirks!

EDIT: BTW, the object dangling by top left corner of the bench meter is that 3mm thick copper heatsink spreader I'm going to use as a shorting bar (it was actually recovered from the cpu fan I'm using to regulate the LPRO's base plate temperature to within 100mK of 36.1 deg C that can be seen in the next three photos).

 For added interest, I've included photos of my Rubidium oscillator project. The ambient temperature readout is about 1 degree low on the actual room temperature. I've had to compensate for about a 2 deg self heating effect which will vanish once the BMP280 module is thermally bonded to a heatsink exposed to the room air in the final build (the thermally isolated over-sized aluminium enclosure I fabricated for the project). At present it doesn't follow room temperature very well but it's enough to compensate for increased cooling rate demands  at higher ambient temperatures.

 Final note: I reversed the shorting plug and the reading went back up to 9.5μV or so after a few minutes after which I reversed it again and observed it stabilise close to the zero volt mark once again. It seems my shorting plug is not quite conforming to the rules of thermodynamics as I'd naively expected. Never mind, perhaps that cheap set of goldplated banana plugs with a copper shorting bar will do better.

[Kleinstein]

The 100 K R4 looks like it is there to get about the same source resistance for all voltage channels. This would allow to use less zero levels to compensate for the effect of the OPs bias current.  I still don't like the 100 K very much, as it would add noise to the 1000 V range.
The 100 K resistor has slightly more noise than the AD8629 amplifier, though less than the ADC without gain. 

With the high impedance leakage at the ADG1408 would cause some error. This could be an issue when really hot, but should not be so bad at a more normal temperature.

[mawyatt]

@mawyatt

 Regarding getting the PCB 'right', I'm going to try bypassing this issue (maybe replacing it with a different issue) by using this bunch of gold plated banana plugs I just ordered from here:-

https://www.ebay.co.uk/itm/294120874481?hash=item447af861f1:g:zDYAAOSwF8Zgdrrp

which look ideal for mounting onto a piece of that 3mm thick copper heatsink spreader plate I mentioned.

Sorry to rain on your parade, but those look exactly like the ones we have from a few years ago. They were not copper but some magnetic metal alloy

They aren't that bad though, but don't count on them being a low thermal EMF plug.

Maybe try a simple thin copper wire for a zero volt source, it's a little tricky to get reliable and repeatable contact with the DMM terminals though, recall Fluke even recommended this.

Good luck hunting for those "Unicorn Droppings", they are rare indeed 

Best,

Edit: BTW our SDM3065X is ~2.3uV offset with ~0.23uV SD, with Auto-Zero OFF the offset doubles.

[Johnny B Good]

@mawyatt

 Thanks for offering that information. I keep a small magnet stuck to the side of my desktop PC case ready to test K type plugs and adaptors so that's going to be the first check I'll be doing when those banana plugs arrive.

 If they do prove to be a magnetic alloy, my hope is that the 3mm thick copper shorting bar will help flatten the temperature disparity in the sockets a bit more than my present banana plug shorting link can manage. In other words, I'm hoping to see at least some improvement.

 In any case, they're likely to be a nicer fit - those ex-GPO banana plugs I used are a very tight fit making it a bit of a struggle to extract them even though I don't fully insert them as far as they could be pushed into the sockets if I were to be foolish enough to "Give it the old college try".

 With regard to pushing wires directly into sockets, that's what I did with a 5 metre long K type thermocouple wire I'd converted into an experimental thermocouple. I'd simply folded the conveniently springy wire ends back to form a contact spring and this worked very well - little thermal mass and no inappropriate alloy material of a cheap banana plug connector to displace the "cold end" away from the meter's isothermal reference connection at the Hi and LO socket's internal connections running some 3K above room ambient.

 I've got a bunch of miniature K type plugs and adaptors where the use of Chromel and Alumel prongs has been specifically stated, ambling their way to me from China. Needless to say, when they do finally arrive, they'll be subjected to the magnet test and, for good measure, a soldering iron test.

 Assuming they prove to be as described, I'll be able to upgrade my cheap thermocouples' miniature K type plugs for use with the SDM3065X via a proper chromel/alumel K to banana plug adaptor. In the meantime, I've got the use of a zero cost alternative to the expense of 'made for purpose' K type plugs and adaptors to tide me over.

[oz2cpu]

to Kleinstein : YES you are right, good seen,
all 3 selected voltage inputs, do look the same, about 100k when see from U406 pin3,
BUT : all ranges uses a separate calibration file, where both offset and gain is handled,

also i dont see any over/under voltage protection clamp.
I think the high series resistance into the first ADG1408, will clamp to 15V
and then again R6 into ADG633 and U406 serve as the last stage of voltage protection,
in that case R4, could be 5k1 since the 10Meg caddock serve fine as input protection,
and since the 20V and 200V ranges are the most important, i like to try improving the temperature drift caused by the mux parts

i have not yet investigated if the analog switch / mux ic's temperature sensitivity,
is due to leak, or changes in on-resistance

[Kleinstein]

Still having seprate calibration files make sense, but the idea of getting the same offset may still have been an idea in the HW design phase. Sw could have changed later. A lower values for R4 may improve the noise in the 1000 V range a little, but not very much as there are also other noise sources:
R4 (100K) ~ 41 nV/sqrt(Hz)
AD7190  with gain 8 ~ 42 nV/sqrt(Hz)    (SINC3 with 1.5 SPS should be around 0.5 Hz noise bandwidth)
AD8629    ~ 22 nV/sqrt(Hz)     (2 x)
 + some current noise (the 5 fA/sqrt(Hz) number in the DS looks too good to be true, I would guess more like 100 fA/sqrt(Hz)) * 110 K
10 K from the divider output impedance  ~ 12 nV/sqrt(Hz)

For the 20 V range the divider output impedance is around 100 K and not much to change here.

[oz2cpu]

i am not even thinking on that noise from the resistance,
but the variable leak / or on resistance, in ADG parts,
if R4 is much lower, the leak change will affect much less, at least that is my teory, now i need to prove it by real science experiments :-)

[Kleinstein]

The leakage current with the CMOS mux chips is usually quite low, often in the 10 pA or below range, though getiing higher with temperature.
So R4 would contribute a little to temperature drift / offset in the 1000 V range. 100 pA*100 K would be 10 µV and after a gain of 1000 this would be 10 mV (for the 1000 V range). So on the display one would hardly see this.  To get 100 pA of leakage it may already need a rather high temperature and the rate of change would be more like 10 pA/K if really high - corresponding to 1 mV/K for the 1000 V range. So I would not consider the leakage current so be such a problem.

In the 20 / 200 V range most of the resistance is from the divider and notmuch to change there.

A variable on resistance is no an issue in the shown circuit, as the on_resistance is still low.