Siglent Bench DMMs SDM3065X, SDM3055 and SDM3045X

[eplpwr]

Have Siglent released a calibration manual for the SDM3065X yet?

My unit have been collecting dust for some years now, I would like to adjust the meter ...

[Neper]

I like this meter except for the relay switching at around 2V when the leads are picking up noise..

Here, it's still worse when the meter is left in resistance mode. Sometimes it rattles like a machine gun.

[Johnny B Good]

 I bought my SDM3065X from Telonic a fortnight ago. It was delivered the very next day (for once I'd not placed my order on a bank holiday Friday evening as is my usual purchasing habit of high value items ). I've been playing with it ever since, mainly to monitor voltages in a temperature stabilised Rb reference oscillator (Efratom LPR101) project I've been working on over the past 13 months since acquiring the LPR101.

 I've tried running a few "sanity checks" with a well rested AA alkaline cell and a CR2032 with a 10M and 610K potential divider to test the 200mV range down into uV territory. Although neither of these "references" are temperature stable, they did at least permit me to check the behaviour of the least significant digit's 'dance routine'.

 The AA cell test with its very slow variation of voltage allowed me to see a well behaved transition from one 10uV readout to the next, typical of any DMM that's faced with slow analogue variations and the task of assigning a digital representation of a messy analogue quantity. So far, so normal.

 I didn't bother adding a 10Mohm in series to create a 2:1 divider with the 10Mohm impedance of the meter to get into the 1.5v range to examine the meter's behaviour on the 2v scale, electing instead to 'go for broke' and look at the 200mV scale with the CR2032 and voltage divider described above. Here, the result wasn't quite as smooth, hinting that the meter was right at its stability limits, very likely aggravated by microvolt scale thermo-electric effects (which appear to be the weakness of trying to measure sub 1 ohm resistors using only a 1mA test current regardless of whether it's a 2 or 4 wire measurement setup).

 I was so intrigued by the difficulty in getting stable uV readings in the 200mV range, I decided to check how accurately and stable a reading I could get out of a 'Zero Volt Reference" (something we can all make up for a fraction of a penny's worth of solid copper wire). The result was a little disturbing compared to my previous experience with pretty well every hand held DMM I've ever owned, including the two 9999 counts Mestek DM91As I've been using of late, where such a zero volt reference would halt the random digital dance at a reading of exactly 0.00V or 000.0mV.

 I was seeing a readout hovering close to either +2uV or - 2uV and anywhere in between and was beginning to think I had a faulty unit until I took a closer look at its specifications after remembering that I was looking at a resolution three orders of magnitude better than those Mesteks (0.1mV versus the 0.1uV of the SDM3065X). Interpreting the full scale accuracy figure given for the 200mV range, it would appear that the meter is actually (but only just!) within specification.

 Despite this 'reassurance' by the specifications, I still find this digital dance with the final two digits a little disturbing, even if using auto zero can get this down to a single count either side of a zero voltage reading, so I am impelled to enquire whether anyone else has used a zero volt reference to quantify these meters' accuracy at the zero volt end of their 200mV scale. IOW. am I the only one to be plagued by this less than perfect representation of a zero volt reference? I suspect not but I'd appreciate some feedback from other SDM3065X owners as a sanity check of my own experience with my one and only 6.5 digit bench DMM.

[2N3055]

I must say I'm confused a bit because you mention Ohms measurement and voltage measurement....

If I understood correctly,  you are saying that you connected copper short on the input and measured accuracy and stability zero on 200mV range.
And measurements were flickering + - 2 uV.

For these measurements I use proper Fluke Zero Ohm plug. It has low thermal mass design.

If you don't have one, try wrapping a cotton towel or something around input sockets to block air circulation. Cotton is good because no static..

Also look at using histogram function. It should plot distribution of results.

You didn't mention how fast was acquisition.  Try 10NPLC and 100NPLC.  1NPLC and less won't suppress 50Hz noise and also faster sampling has separate noise and resolution specification.

[Kleinstein]

The test a zero volts is pretty common to see the meters own noise. With the high resolution meters it is quite normal to see a little movement in the last digit. I normal target is to have an RMS noise of  1/2 the last digit and thus some 3 steps of the last digit as peak to peak noise. Some meters even have twice as much (RMS noise = last digit).
For the higherst accuracy one normally has to use the auto zero mode - the non AZ mode with many meters has quite a lot more noise low frequency noise. The accuracy specs usually assume the AZ mode on.

0.1 µV is really not much - with normal cables it usually takes something like 1 minute for the thermal EMF from touching the plug to decay down to this range.

Coming from the time nut corner, one may use a similar way to represent the stability. Record the data for some time (e.g. 20 min) and than plot the Allan variance.

A divider with 10 M and 10 M is rather high impedance and thus adds extra noise. Chances are one would see the resistor noise of the 5 M divider impedance and maybe some current noise from the input amplifier.

An interesting test is to look at the input bias current:  so does a 10 M resistor really read 0 V or are there a few tenth of a mV.
The other way to look at the input is using a 10 nF (or maybe 1 nF) low loss (e.g. PS or PP type) capacitor and watch how fast it charge or discharges (with the meter in high impedance mode and some shielding).

[mawyatt]

Just did a very quick and crude comparison between the new (10/14/21) DMM6500, over year old KS34465A and couple months old SDM3065X. All DMMs were on over 30 minutes and short used was DIY 4 wire shunt based on 12AWG solid Cu wire. All DMMs were set to Auto Range, Auto Zero and 10NPLC. ~10 minutes allowed to stabilize, since don't have time for longer evaluation.

Temp 77F, RH 40% and no shielding from AC, me or anything else. So please consider this just a quick and dirty comparison, sorry just don't have the time for a more in-depth measurement session.

DMM6500         ~1.75uV with SD ~0.06uv
KS34465A        ~ 0.98uv with SD ~0.12uv
SDM6500X        ~1.8uv  with SD ~0.15uv 

Edit: Longer time >30 min to stabilize, from images below.

DMM6500         ~1.70uv with SD ~0.13uv
KS34465A        ~0.64uv  with SD ~0.18uv
SDM6500X       ~2.1uv  with SD ~0.22uv

Best,

[2N3055]

Rigol DM3068:

[Johnny B Good]

I must say I'm confused a bit because you mention Ohms measurement and voltage measurement....

If I understood correctly,  you are saying that you connected copper short on the input and measured accuracy and stability zero on 200mV range.
And measurements were flickering + - 2 uV.

For these measurements I use proper Fluke Zero Ohm plug. It has low thermal mass design.

If you don't have one, try wrapping a cotton towel or something around input sockets to block air circulation. Cotton is good because no static..

Also look at using histogram function. It should plot distribution of results.

You didn't mention how fast was acquisition.  Try 10NPLC and 100NPLC.  1NPLC and less won't suppress 50Hz noise and also faster sampling has separate noise and resolution specification.

 Thank you for your swift response. The ohms measurement reference was just an instance of where fractional microvolt thermal emfs (Seebeck effect) become a significant confounding effect when trying to measure sub ohm resistance values with just a mere one milliamp of test current.

 Initially, I did use the supplied test leads and croc clips to clip onto an inch long SWG 14 or 16 solid copper wire with the leads gently twisted together to minimise any temperature imbalance and inductive noise pickup but I got the same result (actually slightly better) using one of the cheap Mestek meter's test leads so I made up a short test loop with 5 cm of similar gauge stranded wire and a couple of ex-GPO banana plugs but this was slightly worse, possibly because I'd neglected to thoroughly clean the several decades' worth of atmospheric contamination they may have accumulated (they'd looked clean enough at first sight).

 Reversing test leads and the short circuit loop did produce some shift in offset but not enough to definitively prove these were the sole source of such offsets. I concluded that if temperature gradients were responsible, it seemed more likely that the major contributor was in the meter itself. However, I'll try your suggested remedy to minimise the effect of air circulation the next time I can take a break from my current test (monitoring the output voltage of a μ7805 bolted down to the heat spreader bolted to the LPR101's baseplate which is temperature stabilised to within +/- 0.1 deg C by an attached fan cooled heatsink controlled by a nano MCU reading a thermistor/fixed resistor divider embedded in the heat spreader plate.

 I tried the 100PLC setting but it was no more stable than the default 10PLC setting I normally use, just slower in spitting out readings. Using auto-zero helped reduce but not eliminate the offset suggesting that it was (essentially) eliminating the error within the meter but not that coming from my 'zero volt reference' so reducing random air currents around the terminals and the short circuiting link may well address this uncertainty.

 The readings weren't flicking between +/2μV so much as slowly hovering around those points, varying by how hard the croc clips were biting into the copper wire or small twists of the banana plug shorting loop (and from any reversals of polarity I tried out). I was aware that the heat from my fingers when handling the croc clips to gain the lowest possible contact resistance would likely induce some thermoelectric disturbance so did allow what I thought would be ample time for thermal equilibrium to return. I suspect I'm simply underestimating the thermoelectric effect at this new to me nano volt level of measurement.

 I have played briefly with the histogram and graphing options but the UI is terribly clumsy and unintuitive. Even the barograph settings are tricky to set up (although a useful aid to monitoring the slow voltage drift - ~100μV these past 48 hours, probably from both the μ7805 and the meter). I might have another look with those graphing tools later on but I'm mainly interested in whether this "within spec" offset from a perfect zero is typical or not.

[TheDefpom]

FYI I have observed jumps around the 0v reading due to calibration resolution imperfections, this may or may not be present on a factory calibration, I just know that I have observed it after playing around with calibration of my unit.

This occurs if the calibration source had a small DC offset or rounding error, the meter doesn’t know how to correct for the offset, and so basically jumps from what it’s -0v and +0v values are, so if the calibration source had a 5uV DC offset it would result in a 10uV step between-0v and +0v., this is because the meter has two linear calibrations done, one for the + range, and one for the -.

At the 0v point it is basically switching between two ranges, the + range and the - range. It isn’t a single range for -200mV to +200mV, it is a +0V to +200mV range and a -200mV to -0V range.

[Johnny B Good]

FYI I have observed jumps around the 0v reading due to calibration resolution imperfections, this may or may not be present on a factory calibration, I just know that I have observed it after playing around with calibration of my unit.

This occurs if the calibration source had a small DC offset or rounding error, the meter doesn’t know how to correct for the offset, and so basically jumps from what it’s -0v and +0v values are, so if the calibration source had a 5uV DC offset it would result in a 10uV step between-0v and +0v., this is because the meter has two linear calibrations done, one for the + range, and one for the -.

At the 0v point it is basically switching between two ranges, the + range and the - range. It isn’t a single range for -200mV to +200mV, it is a +0V to +200mV range and a -200mV to -0V range.

 That's an interesting insight which I'd been blithely unaware of. I'd witnessed a similar phenomena with those cheap Mestek 9999 counts DMMs which I'd simply assumed to be a binary to decimal rounding error in the zero to 0.2mV range, which afflicts both meters, when using the 1000mV scale.

 However, since I'd never seen any such discontinuities when monitoring the EFC voltage in my homebrewed GPSDO against a +2000mV offset to allow me to use the 1000mV scale on a 2.3713 (+/-0.3mV) EFC voltage, this separate plus and minus voltage range function offers a much better explanation for my own observation of the 0.2mV gap in those close to zero volt readings with the Mestek meters.

 Other than for a couple of 7 or 8 hour overnight breaks, I've kept the meter powered 24/7 over the past 13 days (it only consumes 13 watts - just 10 watts more than its standby consumption, some 9 watts less than the SDS1202X-E I've had running 24/7 for the past month or so during my Rb reference oscillator test regime, making it a more economic option than the SDS 2104X+ with its 54 watts consumption figure ).

 I'm ageing in its reference (whatever that is - no one has yet published a YT teardown video to reveal what Siglent have actually used in this model so it remains a mystery yet to be revealed). Another 30 days should see it top the magic 1000 hour mark with hopefully the worst of its initial ageing drift behind it.

 As with all things involving precision measurement, stability trumps absolute accuracy every time (you can always go back and recalibrate previous measurement datasets provided the stability of the originally inaccurate readings could be trusted to have stayed the same or at least followed a well understood and predictable drift curve).

 I'd seemingly had no need for better absolute accuracy than those cheap Mestek DMMs were offering until I discovered that they actually did have a tempco that otherwise didn't show in the typically temperature stable environment of my workroom (23*C +/- 1*C). It was only when I was deliberately making wild excursions of room temperature to test the temperature stability of my temperature stabilised LPR101 that I finally realised I could do with a more temperature stable DMM and started checking out affordable 4 1/2 digit bench DMMs.

 I had been tempted by Bangood's special offer on the Owon XDM1041 for only 95 quid (now a whopping £103.70) but was swiftly put off by Owno's... er, Owon's (Freudean slip?) total silence in regard to tempco specs hence my looking at the Siglent offerings, landing up pushing the boat out for the SDM3056X in preference to the slightly cheaper SDM3055X since it had the best tempco specs even if it was only offering a 'mere' fourfold increased resolution.

 Voltage sources that I'd previously considered to be as good as lab standards (the Vref pin of my AE CQE branded OCXOs (ex-Symmetricom) for example) now look terribly unstable using the SDM3065 and I have to keep reminding myself of the 2 1/2 orders of magnitude increase in resolution that allows me to see what had been totally invisible to my best handheld DMMs (Mestek DM91As). It's a whole new world of microvolt resolution out there for me to explore and come to grips with.

[TheDefpom]

They use the LM399 reference :-)

Also, this is one of my older videos which shows the insides:

[Johnny B Good]

@TheDefpom

 Thanks for that enlightenment.    As for blowing multimeter fuses, I did a similar thing with one of my mesteks back in March (possibly the original DMM looking at the order date). I have a set of 9 x 5x20mm 630mA and a set of 10 5x20mm 10A ceramic fuses in my spares drawer now as a consequence.

 My 'accident' wasn't quite so dramatic since it involved using the 1000mA scale and careless current adjustment of a 30v 10A bench supply which took out the 630mA fuse. I think I replaced it with a 500mA glass fuse initially to tide me over whilst I waited for delivery of the specified fuses I'd ordered from a UK based ebay seller. I didn't have long to wait, less than a week afaicr.

 I could see the LM339A quite clearly in the opening still even if I couldn't read the part number until I paused the video at the most stable and clearest view (I had to use half speed to pick my moment though). Even with foreknowledge, I still struggled to interpret the number using a magnifying glass viewing it in full screen HD.

 It seems to be located in a position that leaves it exposed to the flow of cooling air, testing its thermal shield to the limit. I suppose it must be a case of "Damned if you do and damned if you don't" when it comes to optimum placement. I guess that location must have been deemed the least susceptible to temperature extremes. I had hoped they may have used something like an LTZ1000A or better for this model.

 Just for a 'laugh' (and to satisfy idle curiosity) I ordered an AD584L 4-Channel Voltage Reference Module from a Chinese ebay seller last week. I'm not expecting too much by way of absolute accuracy, especially if it's supplied with a printed out calibration slip. Hopefully, it will be a genuine 'L' version rather than a 'J' type disguised as an 'L' type. Whatever it is, I'll be recalibrating it using the SDM3065X. Provided it's reasonably stable, it might still prove to be worth the £9.34 I shelled out for it.    
   

[Kleinstein]

The LM399 shell is quite effective. A little flow of air is not that bad. A little more thermal cover may be a good idea. For a meter of this class the LM399 is well good enough. The problem is more that the ADC chip can not directly work with a 7 V reference. So the reference is divided down to some 2.5 or 5 V for the ADC. The divider can contribute more to the dirft than the actual reference.  The more natural refrence for those SD chips is a bandgap reference like the max6325 used in the SDM3055. However these often show more long term drift.  Ideally would have hoped for a dual reference, like LM399 for the long term and a low noise bandgap for the normal ADC use. However this is likely too expensive for a meter in a price sensitive range.

The dual fuse form is a nice idea, if the fuses are choosen so that normally only the external acessible one blows. It still comes at the price of high burden for the higher currents.

[Bad_Driver]

@TheDefpom

 Just for a 'laugh' (and to satisfy idle curiosity) I ordered an AD584L 4-Channel Voltage Reference Module from a Chinese ebay seller last week. I'm not expecting too much by way of absolute accuracy, especially if it's supplied with a printed out calibration slip. Hopefully, it will be a genuine 'L' version rather than a 'J' type disguised as an 'L' type. Whatever it is, I'll be recalibrating it using the SDM3065X. Provided it's reasonably stable, it might still prove to be worth the £9.34 I shelled out for it.

Hi John, nice to have you here again with the Siglent fan boys 
I got one oft this chinesium voltage references, the „certificate“ seems to be a simple copy and was three years old. And it was good enough for a 4.5 3.5 digit DMM but not for any more.
 At the end I used the reference for one of my double oven TRIMBLE OCXO where because of aging the internal voltage reference was to low for the „sweet“ spot.

[2N3055]

The LM399 shell is quite effective. A little flow of air is not that bad. A little more thermal cover may be a good idea. For a meter of this class the LM399 is well good enough. The problem is more that the ADC chip can not directly work with a 7 V reference. So the reference is divided down to some 2.5 or 5 V for the ADC. The divider can contribute more to the dirft than the actual reference.  The more natural refrence for those SD chips is a bandgap reference like the max6325 used in the SDM3055. However these often show more long term drift.  Ideally would have hoped for a dual reference, like LM399 for the long term and a low noise bandgap for the normal ADC use. However this is likely too expensive for a meter in a price sensitive range.

The dual fuse form is a nice idea, if the fuses are choosen so that normally only the external acessible one blows. It still comes at the price of high burden for the higher currents.

MAX6325 should be buried zener type, same as AD586/587, LT1021 etc..

[bateau020]

Anyone got a list of outstanding firmware bugs on this series (the SDM3055 is my main interest)?

Haven't found one. Maybe it is already known, but this is what I found (latest firmware):

* when in dual, do not use autorange, as otherwise the values either come out way wrong, or it may simply hang the meter (a manual reboot is needed).
* when switching primary and secondary measurement (press "dual") it goes back to auto. Which may immediately hang the meter.

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.

[PartialDischarge]

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.

mine does not hang up. It reads slower but that's all, Its in auto with sw 5.01.01.06

[bateau020]

That must be on a SDM3045X, on a previous fw. Sorry, don't have that meter.
I found that problem on a SDM3055 with 1.01.01.25. Extra info: It is more repeatable with leads plugged in, and after a fresh boot. Also: DCV + ACV has no problem.

[PartialDischarge]

So that means it is a sw issue...

[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).

[Neper]

Still the rater odd volume settings of the continiuty beeper on the 3045.

[CDN_Torsten]

For what it's worth, my 3055 does not hang...slightly older SW.

SW: 1.01.01.22R1
HW: 03-00-00-13-00

[tautech]

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).

Yet one I have here is not afflicted however 3065X with latest FW is ! 
No 3055 stock to check.

Reporting this to the Siglent private forum.

[Calvin]

Hi,

I can confirm the hang-up of the SDM3065X with SW: V3.01.01.10 at DCV/Dual/DCI 

regards
Calvin

[TheDefpom]

Would the need to measure DCV and dca at the same time be a real situation?