Siglent Bench DMMs SDM3065X, SDM3055 and SDM3045X

[tautech]

Chances are the input switching could use a mechanical relay, and this would also be a candidate for the error.  Mechanical switching a low level signal can be tricky with oxidized contacts. So one may even consider alternating between the 20 V and 200 V range and still have the 10 V test signal. Some excercise may help to clean the relay contacts.

Certainly some range changes are switched with relays.

You might be onto something here.
It might also be interesting to know if all models use the same relays and if they have been changed in different HW versions.
Calling Detective Sherlock Holmes .........

[Kleinstein]

I don't think the ADC is a likely candidate for the fault. This would be a rather large error for the ADC. The gain part within SD ADCs is usually relatively simple a change in the sampling frequency for the input. The input is sampled in a a switched capacitor way and gain for the input just increases the sampling frequency for the switched capacitor. This way one gets gain steps very close to integer factors like 2,4,8 and a rare cases also other numbers in between.

The more relalistic candidates for a failure are the divider resistor, relay contacts or a bad solder joint. A test while pressing different parts of the PCB could check for contact problems that react to mechanical movement.

[hpw]

Or use simple a freeze spray, good quality & without dust...

[oz2cpu]

it just did it again
when looking at the two small pulses, they look about same duration ??
so i zoomed it, and read the cursor values .
first long pulse is 83.5 mins,
first short pulse is 21.8 mins,
last short pulse is 20.2 mins,
Blue (the new SDM3055) is still slowly climbing, aging is begun :-)
is there an hour counter we can read from the unit ?

tautech ? do you have access to tech files ?
is it possible to tell me, what parts numbers,
are in the signal path 20V DC range, from the banana and all the way to the ADC,
this way my search for the jumpy problem will be much easier,
I guess you cant send me a schematic ? or a pcb layout signal path explainer ?

[tautech]

tautech do you have access to tech files ?

None at all, sorry.

[oz2cpu]

who will cry with me ??

I opened it, just to take a look

[tautech]

  What's happened there ? 

[oz2cpu]

remember my SDM3055 with the odd sparking or clicking sounds ??

https://youtu.be/hCBbnuBiqwc

it was send back to Siglent for Repair.. this is how they repaired it,
I assume they did nothing, the sparking came from a cap near the ADC that is halve way off the pcb
and its pad is pulled off !! some one dropped a screwdriver into my unit while they assembled it,
or something like that, how the heck will you pull of a 0805 cap this much, together with its pad,
you can even see the prepeg on the bottom side of the pad,
there is no way i can repair this my self.. i never had one happy day with my SDM3055
and now over one year passed !!
Tried to take a few better pictures

[tautech]

Yes it seems this is physical damage from PCB mishandling ....but when.
Are any impact marks visible ?
PM me with the SN# and I'll get its records researched.

[Johnny B Good]

remember my SDM3055 with the odd sparking or clicking sounds ??

https://youtu.be/hCBbnuBiqwc

it was send back to Siglent for Repair.. this is how they repaired it,
I assume they did nothing, the sparking came from a cap near the ADC that is halve way off the pcb
and its pad is pulled off !! some one dropped a screwdriver into my unit while they assembled it,
or something like that, how the heck will you pull of a 0805 cap this much, together with its pad,
you can even see the prepeg on the bottom side of the pad,
there is no way i can repair this my self.. i never had one happy day with my SDM3055
and now over one year passed !!
Tried to take a few better pictures

 I'm a little puzzled as to why you mentioned "and now over one year passed !!". The video you made is dated 28 July 2021, just a little over 6 months ago and Siglent offer a three year warranty (for us Brits at least) on, as far as I've ever known, every one of their test instruments from new. You seem to be implying that you'd originally purchased this just over three years ago. Is this the case?

 Warranty repairs are typically warranted for another 3 or 6 months or the remaining from new warranty period, whichever is the the longer. Even if it really is just a few days "Out of Warranty", in view of what seems to be "repair damage" sustained whilst it was still within its original warranty period, I think it may still be worth asking Siglent to effect an "in warranty repair".

 You seem to have enough documentary evidence to support such a claim and with Siglent having demonstrated that they do care one bit about their customers (unlike the likes of Feeltech/FeelElec for instance), they may well 'do the right thing' if only for the sake of turning a disaster into a PR coup, especially if tautech manages to find evidence in support of your case.

[oz2cpu]

before we blow this sideways,
latest info : Siglent and my Sales rep are on the case, all say they will send me a new one,
so i am as happy as can be again, i do admit i felt like a fuse blown last night..
Yes Johnny the unit was purchased a few days over 1 year ago, it took 6month of research to dig down the strange sparking sound,
and a few month of sending and repair delay, i was lucky to be able to loan another DMM from my sales rep, so yes, i know they do care about me and my work.
Bad things can happen, you need to measure people, on how they handle their customers, when it does,
and so far i have no problems, i expect to get a shipment info within a few days :-)

[mawyatt]

Good news, and Siglent has again stepped up and taken care of the customer

Best,

[tautech]

Good news, and Siglent has again stepped up and taken care of the customer

Best,

This also shows the value in purchasing only from authorized dealers. 

Most of the wheels had been greased already so barely a additional push from here got things moving. 

[oz2cpu]

by the way : since that SDM3055 is open, i took a tweezer and poked a little bit to the cap, and it fell off,
then i repeated my 10V long time super zoomed test, and now the odd jumps are gone :-)
funny to know that was the source of the problems.

anyone know the type number ? datasheet for this Resistor Network ?
google know nothing

[hpw]

@anyone know the type number ? datasheet for this Resistor Network ?

13K resistors may 1..2W so as 8x in series as 104k 

[Kleinstein]

The resistor network in question is th IRC labeled one.  IRC (now part of TTelectronics) seems to be the manufacturer and chances are this is a custom one.
Similar types are available as standard parts with TaN resistance elements.
The KS 34465 and similar seem to use similar resistor networks.

From the position in the circuit my guess is that it is part of the current source for the ohms ranges. The overall similar, lower cost SDM3045 looks like it uses more normal SMD resistors for this.

[Kean]

As Kleinstein says these are probably a custom part, but there is similar ones on the TT IRC website with datasheets

https://www.ttelectronics.com/products/categories/resistors/resistors/?pageNumber=1&orderBy=0&Series=CHC%20Precision

[oz2cpu]

Correct RN500 is for OHM i just double checked it,
I did discover the signal path, from jack to ADC,
I am now mapping temperature dependend parts in the 20V range,
in this quest i find a few things of interest
the Analog muxes are very sensitive to temperature, i did not expect that.
the large resistors R302-R309 are the voltage dividers for 200mV and 2V range.

the Resistor network RN1 is in use 20V-200V-1000V range, and is also very sensitive to temperature,
and it is very sensitive it is to anything comming near it, maybe a good idea to shield it ?
and to temperature isolate it from ambient ?

the zero drift opamp U406 is both the impedance buffer for ADC and LO ref drive.. it is also sensitive to temperature.
Refferance chip and ADC it self are also sensitive, but not as much as i expected,
right next to the ref chip, there is a TL072 running at +/-15V it is running hot, i did not figure out its function yet.
I am preparing to explain my findings in more detail soon.

signals explain
in 20V DC range, 10.0V input
GND of RN1 is+2.5V (buffered refferance)
LO black terminal for voltage input is +2.5V same buffered ref.
divided voltage from RN1 (pin3 is selected via ADG1408 U number unknown
then the voltage is buffered with left side U406 AD8629 zero drift opamp
right side of U406 is also a buffer, for the ref, both outputs go via high value resistor to a cap just over U406
this differential signal is 2.6V and 2.5V (the measured 10V input, is now only 100mV dif signal)
zero input, is 2.5V on both.
on the ADC, pin11 is the 2.6V, voltage measure
pin 12 is the buffered ref 2.5V
pin 13 and 15 is a direct connection to ref ic pin 6 = 2.5V

i find it a little bit odd, 20V range, and 200V range is 100% the same hardware configuration,
the only differance is internal ADC gain,
this means the 20V range, with a 10V signal, will have 10 times as much noise,
compared to the 200V range with a 100V input,
i would have loved it to be the other way :-)

[oz2cpu]

and here is the other picture, explaining the signal path

[Kleinstein]

R302 to R309 should not be a divider, but just a acts as a single resistor that can withstand a high voltage. It is there for protection.

The MUX chips react to temperature gradients, but not very much to the absolute temperature.

Using the same divider for 2 ranges is relatively common. In the usual high end meters the 1000 and 100 V range use a 1:100 divider and than measure the divider output in 1 or 10 V range. How the noise in the ranges behaves depends as there are multiple sources: the divider resistor (can be a mayor noise source compared to modern amplifiers), the amplifier and the ADC.
The divider configuration explains the relatively high noise in the 20 V range, though still good for a 5.5 digit meter.

[oz2cpu]

yes, i belive you are right about the protection resistors, NOT voltage dividers, i double check tonight, and update my post..

YES correct, all this about noise, and temperature drift, are all stuff, way below the normal usage, and normal specifications,
this 5½ digit do more than its specifications allready as it is,
i just try to take it up one more level, for the hobby of it..

next try is to experiment with ADC noise, my idea was to short the differential input near the ADC,
and see if i get a flat line.

[Kleinstein]

Before connecting anything internal, one could to the "normal" noise characterization of the different ranges. The 2 V range should be the best working range of the SDM3055 (and SDM3045), as there should be no divider or gain. Besides the ADC there should be mainly known noise sources: the protection resistor chain and the buffer amplifier. So the 2 V range should already be relatively close to the ADC noise. At least noise wise it should be well in the 6 digit range.

For the input divider it is very convenient to use the 1:100 step, as this allows to use electronic switching after the divider. The divider has a 1:10 step, but with 1000 V at the input this could have up to 100 V after the 1:10 divider and would thus need a relay for switching. So the setup with the 20 V as divide by 100 and than like the 200 mV range is a pragmatic solution and at the 5.5 digit level also sufficient.
The usual 6 digit meters use a similar setup, but with a 10 V range for the buffer/ADC and thus the 100 V range as divide by 100 and than like the 1 V range (e.g. gain of 10 before the ADC).  So with these meters the 100 V range is a bit on the weak side.

[oz2cpu]

great idea, but i can only create a clean and stable voltage at 10.000 V :-) using my voltage standard project,

my normal powersupplies are less clean, and less stable, compared to this DMM
so this is not the way to reveal noise in the different ranges,
I got some 4 in 1 matched 1k thinnfilm resistor networks,
could be an idea to use alot of them in series, to create 1V and 100mV clean and stable, using 10V as input

[Kleinstein]

The initial noise test is with a short at the input, so a stable 0 V that fits all ranges.
A test with a votlage like 2 V or 10 V is mainly there to also check the DMM internal reference.

[oz2cpu]

ZERO volt short:
the last few days i did exactly that, on another DMM 6½ digits,
and find it not that easy to do,
some banana connectors add noise, also i find screw is better over soldering,
and a wire that is thick, and as short as possible is better too.