5 dollar voltmeter with 5 digit (0.1mv) resolution

[paulie]

Well that's interesting. No more $4.70 meters on Ebay. Since this thread started prices have almost doubled. Typical of sellers who probably spend more time trolling the net than tending to business. Fortunately still available on Aliexpress for 5 bucks shipped. Thank god for language barrier.

FYI beware Ebay seller bicoins_de. Notice the low rating. Several complaints for shipping the old low res meter in place of the new one. I suspect most didn't even notice. A few settled for $2 partial refund but not happy. AFAIK the other Ebay sellers and Aliexpress are kosher.

[LabSpokane]

I've got the first of 4 voltage references that Paulie loaned me running now, the 4040a-4.1. The short story is that I don't think it'll work as better than a +\- 1-2mV voltage reference. In other words, this one meets the spec of +\-0.1%.  Will post charts later of the runs and what the voltage refs look like on the scope. Right now my just being in the room was adding noise to my power supply output.

When I'm done, I'll run the same trial in my bargain voltage standard.

If nothing else this was a good opportunity to see what I really have in the way of stable power supplies.

[paulie]

Thanks for the favor and for "coming out" here so more people can benefit compared to PM.

I think there might be some misunderstanding of the goal like Alex30 on page one. Or maybe it's me. My requirement is to get some high accuracy readings for those references ASAP. Not adjust them to some to some precise voltage or see how far off they are or do long or short term drift studies. Only need to record their current voltage level along with Vin and temperature and get them back here for meter calibration.

Or is it the case that these are generating wild amounts of noise and therefor impossible to characterize? The specs for these bandgap type references indicate otherwise and my scope shows very little problems with stability. I was simply using 5v from a usb port an also a 9v battery and saw only a few microvolts of noise.

I simply need a transfer reference not an absolute voltage standard like in the other threads. Am I barking up the wrong tree or missing something here?

[edavid]

Well that's interesting. No more $4.70 meters on Ebay. Since this thread started prices have almost doubled. Typical of sellers who probably spend more time trolling the net than tending to business. Fortunately still available on Aliexpress for 5 bucks shipped. Thank god for language barrier.

Look again, the one range meters are still that price.

FYI beware Ebay seller bicoins_de. Notice the low rating. Several complaints for shipping the old low res meter in place of the new one. I suspect most didn't even notice. A few settled for $2 partial refund but not happy. AFAIK the other Ebay sellers and Aliexpress are kosher.

All the "BuyInCoins" userids are the same seller, and have about the same rating.  If they ship the wrong thing, just insist on a full refund.

[LabSpokane]

The issue with these references is that they depend on a specific current for their voltage.  And with a 100 ppm/C temperature coefficient for my resistors, I'm basically just measuring the temperature coefficient of the resistor.  There's no way to measure these accurately without a current limiting resistor as shown in the data sheet. Otherwise you're just cooking the device.

I'm documenting the setup and will send you the reference resistors I used. I'll also put up the chart of the references over time. Some settle in. Some don't. This is all on the 100uV level.

This is the 4040A-2.5V

I'm using a 3.3K to get the current down to 0.7mA to avoid heating the part. It has stabilized and is a good 4 digit source. The fifth digit just isn't going to happen, unfortunately. It's more than just reading all the digits off the DMM.   

[LabSpokane]

This is a $40 reference standard from www.voltagestandard.com after it has a chance to settle down after 30 mins. I have the previous version of the DMM check. It's just a thermometer now. The drift is largely thermal. . Everything is an interferer at this point. Just being in the room causes drift since all the components are exposed. But this is the next step to get another digit.

The standard was stated as being 5.0000V and was calibrated in May or June. So, that's what it will take as a reference for the 0.1mV meter. And even that standard really needs an enclosure. Everything becomes an antenna as well. 

[paulie]

Well this is proving to be quite an education for me. I see now just hooking up to the 0.1mv meter and watching those last few digits is nowhere near as useful as I thought. And I was way off base thinking 5 digits was free from the kind of influences they go on about in those other threads. So this whole story just gets curiouser and curiouser.

I did notice things went down the tubes fast when connecting mains devices or anything but battery sources for both the reference and the meter (post #39). But never suspected complications due to current and medium term drift. How foolish to think hooking up a 2k thin film and watching the meter for a few seconds was all it took. I was in a real rush to get those "simple readings" back but now I'm grateful to take advantage of your experience and equipment to get real information. Thanks again for your generous willingness to spend time on this.

It's a little shock to see how touchy things are even in the 5 digit world but also kind of thrilling to find I'll get to deal with similar issues as the "big guys".

[LabSpokane]

I need to make a clarification. I in no way meant to say that the voltage standard I referenced previously was in any way of poor quality. What I will try to show is how once we get to the 100uV range, procedures matter. And the "test" I ran basically illustrates what happens when one ignores spec sheets and temperature coefficients.

The voltagestandard.com reference is a tremendous bargain and a great tool.  I have absolutely zero complaints about a 5 digit standard for $40.  I was the one who cheaped out by not putting it into an enclosure. 

What I was trying to say, which came out badly, that getting to 1 mV accuracy isn't too hard for most of us on a non-professional basis.  Each dependable digit after that comes at a price and requires a reasonable amount of attention to procedures, data sheets and environmental conditions. 

This, of course, begs the question of the value of a 0.1 mV or more precise reading for everyday usage. 

More to come this evening. 

[paulie]

0.1mv precision is critical for some of my applications and the main reason I started this thread. For example it's required in order to get 1ma resolution current readings from the 0.1ohm cheapo 1/4w shunt resistor in my PSU project. I've been using the old style meter with PCB traces and ground wires to get crude current values but time to improve that by taking advantage of the new more sensitive meter. No precision current sense chips or complicated op amp circuits here. I tend toward the most simple and cost effective solutions.

That first graph of yours is interesting to me for a few reasons:

1. After the first hour it seems to settle down to less than 0.1mv variation which is quite useable for my purpose. Or maybe I'm not interpreting the voltage scale correctly. Not being a great math wiz those oddball volts/div are hard for me to get a handle on.

2. Is that initial warmup drift due to self-heating, component aging, or did it just come in from the cold? What happens if it's disconnected and similar test run again?

3. I'm guessing Vin is probably 5v but what about temperature? I'm hoping you are taking note because, in view of your comments, this may be even more important than I first realized. I do have a few temperature controlled ovens of various sizes from day job and can accommodate just about any reasonable ambient.

4. It would be very interesting to see what difference there is between your 3.3k 100ppm and the penny 2k resistors I started with. I used those because their TC and the delta Vin/Vout of the reference chips did not seem that critical. I do have a few thousand 10ppm Vishays (again left over from day job) but apparently underestimated the importance of TC here so did not use them. Maybe time to assemble another batch of reference modules.

Another reason 1ma was chosen for across the board current is that anything less violates the minimum current spec of the TL431. I.E. 0.7ma not good for that part. The zener too. I was hoping to keep self-heating issues constant.

Sorry for all the questions but these tests have got me really excited.

[Kalvin]

How about adding a push button which would increase the shunt resistance temporarily from 0.1 ohm to (let's say) 100 ohm. The resulting voltage drop will be 0.1V at 1 milliampere. If you know that the device will consume only a milliampere or so in certain operation state or mode, this might be a simple & usable solution to increase the resolution and accuracy.

[LabSpokane]

The resistors have a 100 ppm/C *temperature coefficient*. I only stock generic 1% and 5% penny resistors as well. That 100 ppm/C is a guess based on what I think I ordered from Mouser. I bought on price for general work, not on specs.

My "lab" runs between 58 and 61F. I didn't turn on the heat because I didn't want to introduce a thermal cycle into the results. All the components had reached ambient temp before test.

The 0.7mA is just from what I had in stock for resistors. The data sheet specified 3.6k and I have 3.3k on the shelf.

The settling you see on that one chart is the resistor reaching thermal equilibrium.

My opinion at this point is that without ovenizing, trimming, low temp coefficient resistors etc., that these voltage references will not be reliable for 0.1mV.  I would strongly recommend going to voltagestandard.com and getting a multi output reference. He uses components that are up to the task and all you have to do is hit the correct ambient temperature and it will work without guessing.  The cost is pretty minimal given the performance and will definitely get you the digit you're looking for.

Here's the standard I would consider:  http://www.voltagestandard.com/PentaRef.html

It's $78.

[paulie]

I appreciate your advice but unfortunately another one of the main purposes for this thread is economy so I'm not in a position to spend more for a reference than the cost of these meters. Certainly not 10x more. In fact even if those were available for $5 it would not be a consideration for me at this time.

Again looking at your graph it seems the variation once settled is around 0.04mv which is considerably more than my requirement for calibrating these meters. Am I missing something here?

As far as resistors I determined from their specs and the reference chips change in Vin/Vout that the cheap thin films I installed in those references were adequate. And the 2k value was chosen so that all of them would have similar current. Not being an expert it's possible this was not the best choice however I don't see a huge difference between 0.7ma and 1.0ma in terms of heating. My goal was not so much to minimize heating because that can be accommodated with warmup but instead to standardize and not violate the minimum for the TL431. 0.7ma does not work for that part.

I assembled those with resistor built-in to facilitate quick and easy readings but have no objection to you running your own experiments. Do you think it would be possible to do a run with the 2k built-in resistors? Hopefully with similar scales so we can compare?

Also since temperature is so important here it would be very helpful if we knew exactly what that is. Apparently 1 deg can make quite a difference. I can reproduce whatever on my end ±0.5 deg as long as I know the original conditions.

[janaf]

I bought six of these some time ago for a bench supply to be. When reading this thread, I had to pop the package open and put them to the test. Mine are the older type, power and measurement share ground.

My "reference" is a 7.5 digit NI meter. Power from an adjustable PS. At room temperature approximately 20C.

I'm impressed by these cheapos!

[LabSpokane]

Here's what I got for the 4040A-2.5:

[LabSpokane]

4040A-4.1V:

(The spikes in the power supply are *me* being near the test setup)

[LabSpokane]

LM336-2.5:

The spikes in the power supply this time are actually a VHF transmitter nearby. 

[LabSpokane]

And lastly, for comparison, VoltageStandard.com's DMM Check at 5.0000V:

The extended drop in the voltage reading is the heat going off at night.  If I had run this test at a constant 68F, I'm certain I would have 5.0000V, not 4.9999X.

[paulie]

Thank you for taking time to run those insightful tests. It's obvious the $44 reference is an excellent standard and far better than the 10 cent specials. It would be great if I had access to one, even briefly, and all my troubles would be solved.

So I understand now that the first ones are for Vin and the second Vout. One thing looks promising, most seem to settle to a fraction of a millivolt after a while so there's hope I can get some useful results from this. So if we took a reading say an hour after power up and noted the temperature this might just do it for me. At least hopefully a little better than the absolute default spec of the 4040a which is all I have now.

Are these done with your 3.3k resistor and in each case the middle pin is left floating?

[janaf]

Some previous posts asked / commented about refresh rate. I estimate mine update at 4-5 hz.

Current is around 8mA / module, with spikes about 1hz, maybe some zeroing going on. I didn't try to measure what it looks like.

 

[paulie]

The older version does seem to have a slightly faster update rate. The 4.3000v one appears to be around 3hz which might be expected considering the higher resolution. Maybe some averaging or oversampling going on.

BTW thanks for the readings on your units. It's amazing that even the 33.000v version is so precise. What a miracle of Microchip engineering to get that amazing performance out of such a cheap and simple circuit.

[janaf]

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.

Makes me think of another thread here about fake FTDI chip. Its quite feasible that someone has the production docs for the MCP3421 and making knock-offs, selling cheap.

[paulie]

It is my personal belief that the whole fake craze is an urban myth, wolf under the bed, flying saucer, romantic notion. It happens but rare. Maybe 0.1% of suspected cases actually verified.

Note that there is a difference between "fake" and "clone". FTDI for example shot themselves in the foot over clones. Sometimes completely different part numbers that tried to take advantage of FTDI drivers which is where all the real development costs lie.

[janaf]

Yes, agree on the Fake/Clone. My bad.

But still, I'm very skeptical if it's possible to produce & sell legally at that price.

[paulie]

Hey, you don't know those Chinese cookie makers. They will produce things like that for 3 cent margin and work for pennies an hour. Unlike western with sense of entitlement who stick their nose up at $20/hr. That's how China won the cold (economic) war.

BTW I did order 5 pcs of this ICs  for $20 on Ebay (another slow boat from China).. Almost not worth it with the meter only a buck more but I didn't want to risk bending the reference by unsoldering. I still can't believe such a simple circuit actually PERFORMS.

[edavid]

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.)