Fixing a multimeter based on VC9805 schematic

[madan1]

Hi there, recently I got a not that cheap chinese multimeter based on VC9805 schematic.
Unfortunately it doesn't measure correctly.

Here is an example:

Range | ~33mv | ~100mv | ~503mv | ~2V | ~3V | ~10V
200mv | 33------ | 101-------| - ---------| -------| ----- | -
2v------| 0.030---| 0.98------ | 504----- | 1.94 | - ---- | -
20v----| 0.01-----| 0.08 ----- | 0.48----- | 1.7-- | 3 ---| 10
200v---| 0 --------| 0 -------- | 0.03 ----- | 0 --- | 2.8 | 9.8
1000v-| 0---------| 0 -------- | 0 --------| 0---- | 1 --- | 8

range ohms
200---- | ---------| --------| ------- | ---------| 1.2
2k------ | --------| --------| ------- | 0.981 | 0
20k---- | 4.89---| --------| 2.15 | 0.96----| 0
200k-- | 4.7-----| 29.5--| 1.9-- | 0.8------| 0
2m---- | 003--- | 0.027 | 0------| 0-------- | 0
20m-- | 0------ | 0.01-- | 0------| 0-------- | 0

Both V and Ohm measures go through RS2 RS3 RS4 and RS5. All that resistors measure as shown on the schematic - 1k, 9k, 90k and 900k. RS6-9 are 1.75m, 2.95m, 1.75m and 2.95m or total of 9m ( while it should be 10m, but even with added 1M resistor, the measure pattern still stays the same ).

I also have measured the resistance of the paths for all V ranges between Rin and the corresponding resistor ( for instance range 20V - Rin <-> RS5xRS4 ) and all paths have almost the same resistance ( +/- 0.2 ohm ). These are the paths which go through the range selector knob. Also, the multimeter is almost brand new and no damage is visible on the contacting surfaces.

Rin is 1M

Any ideas what else to check?


p.s.

range | 1uf | 22uf | 10- | 1 ----| 0.47 | 10
200u--| 1.1 | 23.4 | 10.9| 1.1 --| 0.5 -| 0
20u---| 0.94| 19.3 | 8.52| 0.93--| 0.44 | 0.01
2u----| 0.95| ---- | ----| 0.945 | 0.448| 0.009
200nf-| ----| ---- | --- | ----- | -----| 9.1
20n---| ----| ---- | --- | ----- | -----| 9.15

The current measurements seem fine ( of course need calibration ).

[petehall347]

no idea but good luck with it . still have 3 dmm.s to look at again all same model different faults .
i tried working backwards from what caused the faults such as measuring volts with the leads plugged into in amps sockets and a bypassed ultra fast fuse .

[eccerr0r]

Meters are usually rated ±x% + y digits...

https://www.elektrovadi.com/VC9805-D...RE,PR-470.html

Assuming your voltages are from a calibrated source (and if you do have a calibrated source, I want to know how because all these years I still have yet to get a source that's traceably calibrated despite having many multimeters that could benefit from calibration), feeding a 503mV signal into 20V, 0.5% down is 501mV => 0.50, and three counts error would be 0.47, so 0.48 is within specifications. A quick glance at the voltages don't look way too far off.

The resistances you have to go through a similar calculation for error - this time it's 0.8% + 1 digit. Now I don't know what your resistors are and their calibration is, but you can do the same calculations.

You would have to muck with the trimmers to make them closer, perhaps the VRref 1K pot for everything if it's really that off.

[redwire]

I would check the protection circuits and maybe the rotary switch because the DMM is measuring off numbers only on some functions/ranges?

There are a bunch of diode-connected transistors for clamping overloads. These can get damaged (leaky) which makes low values read low or zero. Or they can turn on early and make higher readings stuck at a mid value.

I would check transistors QC3, QC4 which protect between the two grounds.
TR 9103 (by CR 104) seems to be protection for the ohms (output) current source.

QC1, QC2 for overloads on capacitance measurement.
Tr L6 is protection for thermocouple input.
DI 1, DI 2 protection for current shunts.
Tf1, Tf2 protect the Freq. input

The ICL7106 A/D is a bit hard to understand because DMM input COM is to A/D (-) IN LO, not A/D COM or V-.

[madan1]

Originally Posted by eccerr0r

Meters are usually rated ±x% + y digits...

https://www.elektrovadi.com/VC9805-D...RE,PR-470.html

Assuming your voltages are from a calibrated source (and if you do have a calibrated source, I want to know how because all these years I still have yet to get a source that's traceably calibrated despite having many multimeters that could benefit from calibration), feeding a 503mV signal into 20V, 0.5% down is 501mV => 0.50, and three counts error would be 0.47, so 0.48 is within specifications. A quick glance at the voltages don't look way too far off.

The resistances you have to go through a similar calculation for error - this time it's 0.8% + 1 digit. Now I don't know what your resistors are and their calibration is, but you can do the same calculations.

You would have to muck with the trimmers to make them closer, perhaps the VRref 1K pot for everything if it's really that off.

At the moment the DMM is calibrated within +/-0.25V for the 20V range in reference to another mid-range DMM. Indeed accuracy of +/-0.05v is absolutely enough for my needs.
For the above measurements I used a bench power supply. It is "roughly calibrated" to +/- 0.2v. I also used batteries for additional tests. The OHMmeter also needs calibration, but it will be in vain if the main problem is not fixed.
And the main problem is that let's say in "2V" I have a steady reading of 1.5V, but in "20V" it goes down to 1.3V and in 200 it shows 0.
In comparison, another really cheap chinase DMM ( ~$3 ) shows constant 1.5 in all ranges.
The same thing happens and to the ohmmeter values ( again the $3 DMM shows the same value in all ranges ).

[eccerr0r]

You can test against another DMM in the same range, but a bench PSU is an awful reference unless it's a quality model that has been professionally calibrated. You can't trust a CCC PSU.

You have to keep in mind, without a real reference, maybe your other DMM is wrong too. Cannot tell.

Also please type out all the digits. Your numbers don't tell me enough unless all significant digits are shown (e.g. if you say 1.5V in the 20V range on a 3½ digit DMM, I have to assume anything from 1.45 volts to 1.54 volts which is ten possible numbers. Even worse in 2V range, it could be anywhere from 1.450V to 1.549V when rounded, which is 100 possible numbers. The former, 3 digit rule is violated but very close; the latter, the 3-digit rule is clearly violated.)

[madan1]

Right now the accuracy is not that important. I just want to get the same value in all ranges.
Here are some new readings.

AAA battery
200mv - 1
2v - 1.486
20v - 1.46 ( -0.26v )
200v - 01.2 ( -0.26v )
1000 - 000 ( -0.2+ )

18650 battery
200mv - 1
2v - 1
20v - 4.01
200v - 03.8 ( -0.21 )
1000 - 001-002

another 18650
200mv - 1
2v - 1.057
20v - 1.03 ( -0.027 )
200v - 00.8 (-0.23 )
1000 - 000

AA battery
200mv - 192.2
2v - .189 ( -3.2 )
20v - 0.16 ( -0.029 )
200v - 00.0 ( -0.16+ )
1000v - 000

Looks like the difference between 2v and 20v is biggest.

I gonna test again with resistors.


Edit:

leads only:
200 - 00.0
2 - .000
20 - 0.00
200 - 00.0
2 - .000
20 - 0.00

.22R - 0 in all ranges

1.6r
200 - 00.8
0 in all other ranges

330r
200 - 1
2 - .320
20 - 0.30 ( -0.020 )
200 - 00.1 ( -0.20 )
2 - .000 (-0.1+)
20 - 0.00

1.2kr
200 - 1
2 - 1.182
20 - 1.16 ( -0.022 )
200 - 01.0 ( -0.16 )
2 - .000
20 - 0.00

2k
200 - 1
2 - 1.970
20 - 1.95 ( -0.02 )
200 - 01.7 ( 0.25 )
2 - .000
20 - 0.00
the same R reads as follows on the other cheap DMM:
200 - 1
2k - 1.991
20k - 1.99
200k - 02.0
2m - .002

[eccerr0r]

But the numbers appear to be in spec, if the divider stack is accurate, only thing left is leakage...

Maybe even the capacitor connected to pin35 of the 7106...

[stj]

did you check the resistors in the divider circuit?

[eccerr0r]

Apparently the OP did check them:

Originally Posted by madan1

Both V and Ohm measures go through RS2 RS3 RS4 and RS5. All that resistors measure as shown on the schematic - 1k, 9k, 90k and 900k. RS6-9 are 1.75m, 2.95m, 1.75m and 2.95m or total of 9m ( while it should be 10m, but even with added 1M resistor, the measure pattern still stays the same ).

I don't know if these were tested in or out of circuit, but it looks like if the selector was removed (which most multimeters it's guaranteed because they have to be pulled apart to access the board) then there is no interference with measurement. I recall a multimeter I fixed with inaccuracy issues in ohms mode only that was part of that resistor stack, but it was that lowest one. However it uses different resistor stacks for ohms and volts mode because there's no constant current in vacuum tube voltmeters...

The 1.75/2.95/1.75/2.95 resistance stack totals 9.4MΩ which is not quite right, but the error is linear with respect to the main measurement stack (900K/90K/9K/1K) so it can be balanced out by VRref tuning. The main measurement stack of 900K/90K/9K/1K is the only one that must be accurate with respect to each other. The 1K would be most suspect because it seems that the lower the reading the higher the error, but regardless, it still seems within the specs of the meter...

[redwire]

Originally Posted by eccerr0r

But the numbers appear to be in spec, if the divider stack is accurate, only thing left is leakage...
Maybe even the capacitor connected to pin35 of the 7106...

If the A/D linearity is poor, the Integrator capacitor (pin35) might just be a cheap ceramic part.
Some DMM's have the required big through-hole film cap, and a tiny X7R might explain the poor linearity seen here.
Otherwise, the divider resistors are out or the A/D is noisy/bad, or maybe the power supply rails are noisy if they cheaped out on capacitance.

[eccerr0r]

I don't know why I assumed all DMMs used a film cap there (I made sure I used a film cap for my 7116 on-breadboard experiment!), but you're right, the cheapening of multimeters probably forced people to use a ceramic disc and god forbid a ceramic SMT MLCC for that capacitor... Yow. I guess I have to start looking for that integrating capacitor and shun those DMMs that use a ceramic there.

Then again I don't know about the stability/leakage of newer ceramic MLCCs, as I have been surprised that the average SMT resistor seems to have better tolerance than a typical gold band (5%) through hole resistor. But perhaps it's just because there's no room to print on the tiny resistor what the tolerance should be...

[stj]

smd resistors are metal film, mostly 1%
this is because you can handle more current in a smaller space.

as for caps, you can get smd film caps.

[eccerr0r]

Oh my... film capacitors are pricey and huge...
I recall the int cap for a 7106 needs to be around 0.22µF, these are like at least 50 cents a piece in the 1000s quantity. A ceramic of the same capacitance is 1/10th the cost and probably 1/4th to 1/10th the size!

[redwire]

Look at pics of the vulnerable DT830B DMM which is like OP's.
Some use a big through-hole film cap 0.15uF and others use ceramic X7R for the integrator cap.
SMT filmcaps (PPS dielectric) still melt easily in reflow and I rarely see them used.

[eccerr0r]

I would imagine that leakage would be a bigger problem than capacitance reduction due to voltage. The capacitor must be low leakage and not have microphonic characteristics. I don't think the voltage across the capacitor should reach more than a few volts (3V would be my upper bound guess) so it shouldn't be a huge concern if a high-ish voltage part was selected.

Yes, I was wondering how the heck SMT Polyester caps would survive reflow hence assuming that they didn't exist... but they sure do and how the heck would these be safely reflowed?

[madan1]

Sorry guys, I didn't have the time to work on this. Probably tomorrow will be able thoroughly check your replies and do some more tests.
Meanwhile here are some photos to get an idea about the particular DMM.


p.s. 6R3 = Rin

[eccerr0r]

Eiew COB, mark of cheapness

At least it looks like it is probably using a film capacitor 6C6 for the int cap, though I'm not sure how the board correlates to the actual schematic, they look a bit mismatched. I don't know why there's a resistor in series however, that seems a little out of place from all the 7106 series chip circuits I've at least seen...

[madan1]

I have traced the V line from the probe to the IC and it exactly matches the schematics ( the components are marked differently though ). Tomorrow will trace the Vref lines and will check the components pointed out in the above comments.

[madan1]

Well, I guess I killed the analog part. I shorted 6R5 and after that I'm not able to get a zero ( even after I removed the bridge wire )... not it just loops between -1 and +1 ( for V ).

edit False alarm... It turned out the selector knob is causing the issue..


edit 2:
I shorted again 6R5 and this reduced the difference between 20V range and 200V range to ~0.14V ( battery 7.74 in 20V -> 7.6-7.7 in 200V -> 007 in 1000V ). Still to test it with other voltages and resistors, but I guess the results will be similar. Any ideas how to improve even more the accuracy between the ranges?

The capacitance of 6C6 is OK.