Cheap chinese AD584 voltage reference - legit cal data? Let's find out!

[usagi]

I bought one of those cheap chinese voltage ad584 voltage references off ebay. A "Volt Ref.   Hao Qi Xin" model.

I've seen claims that the cal data included with them is pure fiction. I decided to find out for myself if this was true or not.

Found access to a HP3458A, and compared the readings to the supplied cal data.

Supplied cal data:

• Agilent 34401A   2015   21C
• 2.500V: 2.50090
• 5.000V: 5.00239
• 7.500V: 7.50235
• 10.000V: 10.00384

Tested values:

• HP 3458A   2017    20.6C
• 2.500V: 2.500847
• 5.000V: 5.002338
• 7.500V: 7.502800
• 10.000V: 10.003795

So at least for my module purchased from ebay, the supplied cal numbers look legit. And will surely be adequate references for my 4-1/2 and 5-1/2 meters 

[CustomEngineerer]

With so many of those being sold on ebay, would you mind sharing who you bought from? Thanks

[technogeeky]

Also of note:

• There are at least two versions of this device: one charged by USB, one is charged by arbitrary 5V input.
• There is a design flaw if you have some kinds of probes: it is possible for your probe inserted into the right side connector to contact two terminals of a chip underneath it, shorting them out. It didn't appear to cause damage, but it still should have been prevented.

[SvanGool]

I think it is this reference being referred to: http://www.ebay.com/itm/Voltage-Reference-Module-AD584kH-High-Precision-4-Channel-2-5V-7-5V-5V-10V-/271955314814?hash=item3f51ccc87e:g:L0cAAOSwPcVVywg5
According to: https://www.eevblog.com/forum/testgear/used-multimeter-buyer's-guide/msg1144200/#msg1144200

[mtdoc]

Also of note:
There is a design flaw if you have some kinds of probes: it is possible for your probe inserted into the right side connector to contact two terminals of a chip underneath it, shorting them out. It didn't appear to cause damage, but it still should have been prevented.[/li][/list]

Ha!  I did that and it blew out the UN8HX Liithium battery charge control chip underneath. Fortunately it was cheap to order a few of those chips on Aliexpress and an easy fix. Now I have a piece of insulating foam in place to prevent this.

Also note - the OPs reference has the lower spec'd AD584kH chip. Some of these are sold with the higher spec'd AD584LH chip.

[tronde]

Also note - the OPs reference has the lower spec'd AD584kH chip. Some of these are sold with the higher spec'd AD584LH chip.

It seems the people behind L-type are cheating. All units have the same printed test report

https://www.eevblog.com/forum/metrology/are-cheap-ad584-units-worth-it/msg920924/#msg920924

[mtdoc]

Also note - the OPs reference has the lower spec'd AD584kH chip. Some of these are sold with the higher spec'd AD584LH chip.

It seems the people behind L-type are cheating. All units have the same printed test report

https://www.eevblog.com/forum/metrology/are-cheap-ad584-units-worth-it/msg920924/#msg920924

Mine, with a well aged AD584LH chip came without any test report and I am fine with that. It wouldn't have been trustworthy anyways.  It's been rock steady since I got it a few years ago. For my hobbyist voltage transfer standard use it is perfect.

[Retep]

It seems the people behind L-type are cheating. All units have the same printed test report

About a year ago I bought the AD584LH version about a year ago from ebay seller 201377jie. It came with a paper with cal data very similar to the paper discussed in the thread you linked to, with the exception that the voltages were handwritten. I checked it against my HP 3455A and after having the room heated up to 23^oC (like indicated on the cal sheet) and having the meter turned on for about an hour the voltages measured by the HP3455A were surprisingly close to what was on the cal sheet, only a few counts difference at most.

That being said the HP 3455A is really old and hasn't been calibrated for ages, so it might have been just coincidence.

I noticed the AD584LH I got was quite old (date code 0742 -  salvaged part?) and seemed to be quite sensitive to temperature; when the room temperature was 15^oC the measured difference was quite a bit larger.

	Cal sheet	HP 3455A (Vin=18.0 V, T=23.5oC )
2.5V	2.50194	2.50197
5.0V	5.00326	5.00327
7.5V	7.50664	7.50666
10V	10.00736	10.00736

[Housedad]

I don't think they calibrate anything either.   Like everything on Ebay, your mileage may (and will) vary in what you get.

[crazyguy]

Also of note:

• There are at least two versions of this device: one charged by USB, one is charged by arbitrary 5V input.
• There is a design flaw if you have some kinds of probes: it is possible for your probe inserted into the right side connector to contact two terminals of a chip underneath it, shorting them out. It didn't appear to cause damage, but it still should have been prevented.

I bought two AD584 Voltage Ref from the original seller on Taobao in January 2017. The updated version should be charged by Micro USB port. I think the arbitrary 5V input is old stock
https://item.taobao.com/item.htm?spm=a1z09.2.0.0.x4Crsq&id=12553765969&_u=4jeeudi4e13

There is another cheap one without Li-Po battery/charger/control electronic circuit
https://item.taobao.com/item.htm?spm=2013.1.20141001.2.eYdGrg&id=22177471816&scm=1007.10115.70237.100200300000000&pvid=bac279b4-cd97-48e1-a0c2-55ef45aa1f6b&idnum=0

[dav]

What is the usefulness of a battery in a voltage reference? 

[kcbrown]

What is the usefulness of a battery in a voltage reference? 

Supply voltage stability.  Nothing beats a battery for that.

[1audio]

Actually, disconnecting from an external power source/ AC line removes leakage, noise and uncertainty sources so batteries are a good feature. The important question is long term drift variation. The only way to know is to give it time. Let it burn in for two weeks, get initial readings and then check it again after a few more weeks and you're still looking at early changes. For 3 1/2 digits these solutions can work well. beyond that you learn that the meter works but not how accurate it is.

Now that a hobbyist can get an 8 1/2 digit meter the reality of elusive accuracy comes home. If you have your own JJA maybe you can be confident. Otherwise you will be in perpetual anxiety about what the real value is. I have 6  1/2 and 7 1/2 digit meters and a 732A and even if I had the 732A calibrated I would not trust it past 5 digits. Too many other things to degrade the accuracy. I have come to accept the limitations and move on but the understanding of all the possible error sources gives me lots of respect for those shops who can really meet the requirements for 6 1/2+ meters.

[Wytnucls]

My Volt Ref AD584LH 'Hao Qi Xin'
Obtained in 2014 with hand-written calibration figures.  Date code: 8609
Tested today against Keithley 2000 and Metrahit 30M 6 1/2 meters, both with original calibration 2012:
23C
REF        2.50070  5.00005  7.50049  9.99912
K2000    2.50070  5.00005  7.50050  9.99916
30M       2.50069  5.00002  7.50052  9.99915

[usagi]

so it seems the AD584LH Hao Qi Xin references are good for hobbyist use and cal data is real.

[AlessandroAU]

Here is mine:

[dav]

There are at least two versions of this device: one charged by USB, one is charged by arbitrary 5V input.

So what is the newest revision, the USB one or the 5V input one?

[MrW0lf]

Did get 2 of these devices:

AD584kH (5V input): https://www.amazon.de/gp/product/B00Z033Y6Y/
AD588 (Micro-USB): http://www.ebay.co.uk/itm/142146232442

Just to check it with best DMM I have would be plain boring, so here they are - friendly family of colorful DUTs:



Results:



Conclusion: Time to take apart some of my oldest DMMs ...and buy 6.5digit DMM just to be sure... 

[boggis the cat]

Ambient temperature is important.  Make sure that your tests were done at the same temperature before you go trying to adjust any of the multimeters.

(They're quite likely all in spec anyway.  Not surprising for Agilent / Keysight, but good for cheap instruments.)

[Housedad]

Interesting results.   Those chips are rather accurate.  I bought a AD588 version on ebay last month to check out the two Instek GDM-8251's I bought.  Maybe their handwritten calibration is not so far off after all.
I posted the results in another thread: 

I think I lucked out.   I ordered and received a battery powered AD588BQ voltage reference from China (yeah, I know)  Anyway, it came with a supposed readout from a Agilent 34401a.  ( whoopy doo  a measurement that may or not be calibrated)   They showed a measurement of 10.00039 and 5.00057.

Going by the Ad855 Datasheet,  It should not be any more than +/- 1mv off of 10v and 5v.    I hooked up the +15v and -15v to the board from my power supply  (Instek GDP-3303s) and let the AD588BQ  and DMM's warmup for two hours and  checked the voltages.   Meters set to slow mode.  Using standard Probemaster 8008 leads 60" long. 

unit 1:

AD588   5v:     4.9999v
AD588 10v:     9.9993v


Unit 2:

AD588   5v:     4.9999v
AD588 10v:     9.9990v

I know this is only a relative comparison with nothing verifiable, but even given the GDM-8251 tolerance of .012% +5, I'm pretty happy with it. 

edit:  I had originaly mistyped the name of the chip as AD855BQ  when it is actually a AD588BQ reference chip.  Sorry for the confusion it may have caused anyone.

https://www.eevblog.com/forum/testgear/gw-instek-gdm-8251a-dmm-calibration/msg1134038/#msg1134038

[amirm]

I bought this one from Amazon: https://www.amazon.com/gp/product/B00Z033Y6Y/ref=oh_aui_search_detailpage?ie=UTF8&psc=1

On my Keithley it is almost a perfect match which was very surprising to me.  All my other handheld meters deviated a fair bit so it does the job.

[MrW0lf]

Ambient temperature is important.  Make sure that your tests were done at the same temperature before you go trying to adjust any of the multimeters.

It was about 21...22C. So might affect some meters, in general ones most off are oldest and most used. 2010DMM is old but very little used. As for scopes AD2 was calibrated some days ago at same temp, using official procedure (first cal internal gen with DMM, then gen cals scope). Built in "DMM" panel was used to measure. With Pico I just picked single most suitable range, applied DC offset + 12bit, entered cal formula (matched to middle point in range) in math channel. Results were repeatable over some hours. It's interesting that you can beat cheap DMMs with scope taking advantage of smoothed ADC noise boosting effective accuracy.

As for AD5* modules - quite good, especially to keep cheap DMMs at bay. But first thing I did was special cable not to short stuff inside AD584kH module. Seems it was made to match some very cheap "chinese mainstream" bananas, which are bit shorter than quality ones.