5 dollar voltmeter with 5 digit (0.1mv) resolution

[LabSpokane]

I don't have a metrology lab, so for temp, you get what you get.  These tests were run between 60 and 61F.  To really get a voltage standard to consistently run to the 0.1 mV benchmark, one really needs a feedback mechanism to handle temperature compensation.

 I'll put up a simple schematic later.  The 3.3K resistor was placed in line with the zener reference.  I put a 1Mohm in front of the DMM measuring the source voltage to try and reduce the loading on the reading.  Is that "correct"?  Probably not, but its what one has to do without any feedback compensation. It's a dance between keeping enough current flowing through the zener to get it past the knee, not heat it or the resistor significantly, and keep it at as close to one current as possible.  See attached diagram.  The IV curve to the left of the knee looks like a cliff, but its not.  Once you're in the uV neighborhood, the slope of that line matters.

This is pretty much all I have time for at this time. I actually need to get back on track on another project. 
on't have a metrology lab.  These tests were run between 60 and 61F.  To really get a voltage standard to consistently run to the 0.1 mV benchmark, one really needs a feedback mechanism to handle temperature compensation.

Good luck with your project!!

[paulie]

I don't have a metrology lab, so for temp, you get what you get.  These tests were run between 60 and 61F.  To really get a voltage standard to consistently run to the 0.1 mV benchmark, one really needs a feedback mechanism to handle temperature compensation.

 I'll put up a simple schematic later.  The 3.3K resistor was placed in line with the zener reference.  I put a 1Mohm in front of the DMM measuring the source voltage to try and reduce the loading on the reading.  Is that "correct"?  Probably not, but its what one has to do without any feedback compensation. It's a dance between keeping enough current flowing through the zener to get it past the knee, not heat it or the resistor significantly, and keep it at as close to one current as possible.  See attached diagram.  The IV curve to the left of the knee looks like a cliff, but its not.  Once you're in the uV neighborhood, the slope of that line matters.

This is pretty much all I have time for at this time. I actually need to get back on track on another project. 
on't have a metrology lab.  These tests were run between 60 and 61F.  To really get a voltage standard to consistently run to the 0.1 mV benchmark, one really needs a feedback mechanism to handle temperature compensation.

Good luck with your project!!

It's unfortunate you are not in a position to help me with basic meter and temperature readings but the graphs and lectures were interesting. Would you do me a favor and return the references in the stamped envelop I provided? A friend of mine has a Keithly that is uncalibrated but indicates he may be getting that done sometime in the future. I would like to be prepared just in case.

ps. None of the references I sent were zener based. They were bandgap based integrated circuits with somewhat different characteristics. I do have some zener based references here like the one in the page one photo. It was kind of you to reiterate how those work for anybody not familiar with the technology.

Thanks again for your offer to help. Sorry it didn't work out. Good luck with your new project.

[LabSpokane]

Apologies.  I skimmed the data sheets very very briefly.  They do show a zener symbol, so it was an easy mistake to make. I am returning the samples do you tomorrow.

And honestly, it didn't work out because you didn't provide a completed device. None of them had a current limiting resistor.  I flat couldn't test one because it was an adjustable source and required bias resistors. As provided, the unit was a dead short.

[paulie]

And honestly, it didn't work out because you didn't provide a completed device. None of them had a current limiting resistor.  I flat couldn't test one because it was an adjustable source and required bias resistors. As provided, the unit was a dead short.

I think there might be other reasons it didn't work out. As anyone who skimmed even very very very briefly the first post here can see these are complete voltage references. The resistors are clearly visible. Also the value and type of this resistor and how to connect the reference was described in detail in several of my recent replies.

There were a few individuals besides myself who hooked these same ones up and found they performed perfectly as stand alone references. No 3.3k, 1meg, feedback circuit, or adjustable source required. Apply power and read. The only problem being no calibrated measurements available to compare.

In any case it would be greatly appreciated if you could drop those off at earliest convenience so I'll have a rough idea when to expect them at this end. Thanks again for your efforts.

ps. from page one:

[JohnnyBerg]

For testing things I build myself i nice little voltage standard:



It uses a TI REF5025A, base accuracy 0.1%

[janaf]

$1.75 for the 5 digit ADC, then processor, PCB, display, housing, a few small components, development, production cost & margin, seller work & profit selling small quantities. If they can do that, fine with me. If they can do it all legally, I'm happier.

If it's the MCP3421, they are listed at $1.74 ea @5000 pcs. No way they can make a complete meter and sell and make a profit at $5 each.

Microchip's prices may be different in China, but even if not... I just bought some 4 digit DPMs for $1.50 each.  Adding $1.74 to the BOM and $3.50 to the price doesn't seem unreasonable.

Hmm, looking again, here's a source for the 5 digit single range voltmeters for $3.95, quantity 15:

http://www.aliexpress.com/store/product/Red-LED-display-Color-DC-0-33-000V-0-36-Digital-Voltmeter-three-Wires-5-Digit/835957_1320727454.html

(They also have the 5 digit dual range voltmeters for $5.20, and 5 digit current meters for $4.48.)

[paulie]

For testing things I build myself i nice little voltage standard:
It uses a TI REF5025A, base accuracy 0.1%

Way cool! Looks a lot like those $100 internet references. Not set for round absolute voltage but if measured with a lab meter should be just as precise and useful for calibrating DVMs. I see there is resistor and current standard too which is also way cool.

ps. That chip seems to be available on Ebay for around $10. I wonder what advantage over the somewhat more common and less expensive 4040a which is speced for same accuracy.

[JohnnyBerg]

For testing things I build myself i nice little voltage standard:
It uses a TI REF5025A, base accuracy 0.1%

Way cool! Looks a lot like those $100 internet references. Not set for round absolute voltage but if measured with a lab meter should be just as precise and useful for calibrating DVMs. I see there is resistor and current standard too which is also way cool.

ps. That chip seems to be available on Ebay for around $10. I wonder what advantage over the somewhat more common and less expensive 4040a which is speced for same accuracy.

Thanks.

The advantage is drift and stability. The REF5025A (that is what I am using) has a MAX drift of 8ppm/K. Also the long term stability is very good. Have a look at the datasheet.

Actually, I ordered 10 boards, and use only 1 for myself. I build 7 pcs and sell them for around $15. As the chip cost around $8 I am practicly giving them away 

I also sell the 3 empty board, for $3 

[paulie]

Oh yeah... 8ppm is a lot better than 4040a 100ppm/1000hr. Maybe sometimes you do get what you pay for.

BTW poking around with some very low voltages it seems there is no need for "isolated supply". I tried a divider off the same power rail as the meter and also off a separate battery source and in all cases it maintained very stable 0.1mv resolution. No difference isolated or not.  Or maybe I don't understand what they mean by isolated.

Looking at the schematic I also don't see why there is benefit to isolation either. Ground for power and input is the same. What am I missing?

[janaf]

Take a look at the results I published pictures of a page back. Those are the common ground type, 3-wires only. Results are imo great. Cant see why common ground should not work. Before the Rail-Rail op-amps it was impossible to measure down to zero with common ground. There where even devices (panels like these) that would self-destroy if used with non-floating power. Had to do with nulling circuit. Earlier, the common ground type where usually a bit more expensive than the floating type. Not sure why.

With the 5-digit modules it seems the ones with an extra decimal at low voltages require separate supply, while the ones with common ground don't show the extra decimal at low voltage. Don't know if there is a connection (sorry) there.

[edavid]

With the 5-digit modules it seems the ones with an extra decimal at low voltages require separate supply, while the ones with common ground don't show the extra decimal at low voltage. Don't know if there is a connection (sorry) there.

Here is what the AliExpress seller says about the dual range/4 wire type:

"Flexible supply: There is no need to isolate power supply, you can use the system common ground.(Note: If you need 0.1 mV resolution please use isolated power supply. Otherwise the accuracy drop an order of magnitude."

I think this is mainly due to the supply current flowing in the common ground lead.

[JohnnyBerg]

Oh yeah... 8ppm is a lot better than 4040a 100ppm/1000hr. Maybe sometimes you do get what you pay for.

Actually, 1000hrs drift of the REF5025A is typ. 45 ppm.

Looking at the datasheet of the 4040A on page 5 I see 100ppm/K (vs 8 of the REF5025) temp. coef. and 120ppm / 1000hrs.

When you pay some extra money for the REF2025 High grade version you get 3ppm/K drift. So yes .. you get what you pay for. Prices have a tendency to rise exponentially with accuracy and stability.

[paulie]

I think this is mainly due to the supply current flowing in the common ground lead.

Yes, that makes sense now. So a long thin common ground wire for both voltage source and for power would cause trouble. The meter draws very few milliamps but when you are dealing in the microvolt range even tiny offsets like that can throw readings off.

I didn't experience it because my setup had 2 separate ground wires meeting at the meter.  So that's why it didn't matter whether I used the 3 or 4 wire connections.  The 4 wire and 3 wire boards are identical but separate connectors on the 4 wire forces using separate grounds thereby preventing current offset. So by isolation they really just mean use both connectors.  I see said the blind man.

[paulie]

Photo and more details of my calibration mod attached to post #1:

[splin]

Is it possible to reprogram the STM8 on these meters?

[paulie]

Is it possible to reprogram the STM8 on these meters?

These can be programmed with the SWD pin seen in upper left corner of the photo. I tested it and it works but since there is no firmware for this I chose to use an external Mega8 for the input expander data log adapter. The circuit and program already existed for a previous ICL7135 project so it only took a few minutes to update the BCD table to 7seg.

ps. I ordered two more of these meters on sale at Aliexpress so maybe when they arrive I can afford to play with flashing one. I do love that SDCC no nonsense C compiler.

[paulie]

Couple developments. The references from LabSpokane arrived back and I want to thank him again for all the hard work. It's unfortunate confusion over the circuits resulted in more effort than necessary. He was nice to include the same resistor used in his tests so maybe I can build a jig with that to garner useful info from the graphs.

In other news I am now the owner of one of those Ebay AD584 voltage standards. A friend of mine gave it to me because it did not come with precision readings so not of much use for his purpose. I did determine that the Ebay meter is at least 7ms accurate based on the AD584 initial voltage spec. We probably already guessed that from reviews published earlier in this thread. Probably a lot better in fact based on those. Of course I can't be sure unless I can get some reference readings from somebody who owns a precision meter. My buddy is considering getting his calibrated and got quotes from some standards labs but that may be some time off.

Meanwhile I'm logging a set of the DIY references using the Ebay meter with my 16 channel input expander. Hopefully once it's calibrated I can adjust the data and get useful info. Even if not at least a chance to check out my logger and oven hardware which hasn't been operational since last year.

[JBeale]

I got a few of these ebay 0.1 mV resolution panel meters (they cost $7 for me) and on my samples, the DC accuracy looks like +/- 0.5 mV in the low range, compared with a Keithley 196 DMM and some voltage refs. The behavior of the meter when the input voltage has a step from 2.500 V to 5.000 V is not so nice. The decimal position update lags behind the digit update, so it briefly shows 30V and 40V before settling at 5.000 V.
See also: http://youtu.be/9r6liwNnXaY
In the video, the actual voltages (from an AD584KH-based box) are: 2.50047, 4.99957, 7.50225, 10.00065 V.

[paulie]

Yes, the glitches at that transition point are annoying. I also wish the transition was at 9.9999v instead of 4.3000v. What would be really nice would be 0-9.9999v then 0-99.999v like real auto range voltmeters. That would probably require external circuitry though so might be too much to ask. No matter what these are quite a bargain compared to brand name 5 digit meters even if it is just a panel meter.

BTW Thanks for that 24 bit ADC project. Building my own version helped me understand why 24 bits is never really 24 bits.

[electr_peter]

Customised SW could improve transition in ranges.
As ADC is controlled from MCU via I2C bus, ranges transition goes like: ADC->max value in range; MCU-> saturation, change range(possible fail here); ADC-> different reading (possible fail here).
4.300 range uses greater amplification in ADC, maybe it over-saturates/fails during this change and MCU SW simply shows erroneous reading.

[paulie]

Recently needed to measure HV so another mod to increase meter range. More photos and info attached to post #1.

[paulie]

So here's the question... can we upgrade the 0-33.000v meter to perform like the one that costs twice as much by replacing the 20k input R with higher value? Imagine 5 digit 0.1mv resolution meter for $3 and change. As added bonus increase input impedance to more closely match commercial DMM in the process.

ps.I've just ordered a few more of those just to find out.

[paulie]

So the answer is YES! I couldn't wait for the latest meters to arrive (damn chinese new year) so yanked  the old 0-33.000v meter from existing equipment and gave it a try.

No problem boosting resolution from 1mv to 0.1mv by simply changing that resistor. Apparently it's also possible to increase even further to 0.01mv.  However range would only be 0-0.33v which is not that useful for my current applications. I need to measure 2.5v voltage references so 0-3.3v does the job. Either that or get involved with resistive dividers which is a whole other project.

When used with my x10 opamp/multiplexer circuit I do end up with 0.01mv resolution but with VERY restricted range (~2.35-2.65v). So effectively a 6 digit budget voltmeter that's more than stable enough to track 1ppm TC and long term drift.

BTW another beneficial side effect to this modification effect is no range change artifacts because there is no range change on the cheaper meter.

[MartinX]

Those meters are terrific! I saw this thread and thought that would be perfect to put in my boat to monitor the battery voltage, the resolution is really handy, you can see if the battery is discharging really well. I bought some on Ebay and now that I got them the accuracy is better than I thought. I think I will add some reverse polarity protection and over voltage protection to better cope with the automotive environment, I also plan on taking the board out of the holder and put some conformal coating on it.

[paulie]

When I first started working with these meters, specially the new 0.1mv one, I thought it was crazy overkill to use it for PSU or battery monitoring. Now I'm smarter and see big advantages in those applications. For one thing the current monitor for my own PSU design benefits big time because there is no need for fancy op amp circuits and expensive shunts. It's insane how some of these guys chase their tails over technical issues when really there is a much simpler solution. A cheap 0.1ohm 1/4watt resistor in the ground lead will give readings accurate from1ma to 2amps. A DPST Switch from that point to the V+ line allows using the same meter for voltage and current monitoring. Ultimate efficiency and cost effectiveness.

I also see that as you suggest it's very helpful in keeping track of battery discharge. Instead of changes every few hours or days a sensitive meter keeps you updated on a time frame of seconds or minutes. Proved very useful in my solar array. Several more of these meters are on order now and will scatter them all over the place.

As far as reverse polarity there is already a diode built in for that as can be seen in the post #1 schematic. Two friends are using these for automotive which sounds similar to boat and all they needed were low cost 7805 type regulators and caps for noise suppression. One fellow measured 80v transients which bricked several meters before but no problems since.