XJW01 Auto LCR meter review ($120 bench top LCR meter)

[dannyf]

Nice project to study.

Nice project. A very innovative design.

- Cx@K3 produces the voltage across the DUT; Cy@K3 produces the current through the DUT.
- sine wave produced through pwm;
- the most interesting part is OP07 and V1/V2 as a phase detector: I have used similar designs in the ESR meter. It is a digitally controlled phase detector without resorting to analog switches - the typical implementation.
- I didn't look at the software but he must be doing some fairly sophisticated math there.
- He used a 8051. With a faster chip, he can simplify the design considerably and go much higher in frequency.
- the calibration procedure is a little bit too complicated. But the fact that it is implemented at all is a huge plus.

[nctnico]

I'm starting to like this little critter more and more. The good old HP4274A I bought a couple of months ago could not match it's capability to measure low resistance (tens of milli-ohms). Too bad it only supports a few frequencies.

[quantumvolt]

Remote 'comparison' of XJW01 with Hameg HM8118 (new, in calibration - https://www.rohde-schwarz.com/us/product/hm8118-productstartpage_63493-44101.html).

   ... --- ...

Since I saw the XJW01 meter on eBay, I knew I would buy one (https://www.eevblog.com/forum/testgear/xjw01-auto-lcr-meter-review-%28$120-bench-top-lcr-meter%29/msg429607/#msg429607).

In general I like it a bit untidy, sloppy, dodgy ..., and I thought that if it is all junk I can always follow the safety seeking herd and buy a DE-5000 (I will buy that one too  ).

I have now had it about a year or so. I have had a lot of fun. To make a long story short I have collected quite a few quality caps (+- 0.5% and a few better). I have built oscillators and Wien bridges using Vishay precision resistors and 2x Agilent 34970A (AC range, interchanged bridge legs with relays) and a precision frequency counter. And a lot of math and statistical estimation / regression. All to determine the 'real' value of capacitors - and hence the errors of my XJW01.

When I thought I knew the errors (for 'all' ranges, values and temperatures) of this box, I put in this ad https://www.eevblog.com/forum/buysellwanted/%28wtb%29-measurement-of-capacitors-%28from-calibrated-lcr-meter%29-%28paypal%29/.

I got an answer from a very nice member here that has a Hameg (Rohde&Schwarz) precision LCR bridge at work. So I sent him the test caps to England and got the results back in email. Thanks a lot for his kindness.

This comparison (3 weeks between measurements) of my XJW01 is anecdotal in the sense that it cannot in any way be generalized, but on a collection of high quality caps (180 pF to 33.3 nF - almost zero loss and almost zero temperature coefficient), my USD 99 (Best Offer) XJW01 meter was never more than 0.155 % off compared to the USD 1500+ Hameg lab instrument.

 

[Carrington]

Does anyone know where one can buy this other?
And how much it cost?



Source: http://www.crystalradio.cn/thread-373788-259-1.html

Thanks.

[happydad]

Can someone explain how the design calculates ESR?  I assume that the actual V & I measurements give us impedance, which is not ESR.  What am I missing?
Thanks.

[nctnico]

You can calculate the capacitance and resistance (ESR) from the complex impedance. Google 'lcr meter formula' and you'll find a document from HP.

[happydad]

OK, so I am assuming that the generator is putting out sine wave so we can get phase information from reading back V & I?  A lot is hidden in the software, and I'm still trying to figure out how you extract phase information.

[happydad]

From what I've read, in order to separate out the real and imaginary components from what was measured, we have to multiply the V and I by the signal generator, at both 0deg and 90deg.  For education purposes, where is that implemented in the attached schematic?  If I were to do a brief walk through, we have 6th-order LC filtered DDS sine wave ac coupled and buffered generator driving the DUT.  We then measure the differential complex voltage (and current) through TL084 opamps, gained up by 2 and rc filtered through 1kohm and 470pF, and then running it through a programmable gain stage, then another 3x buffered stage to another differential TL082 with PA0 and PA1 ports.  Here is where I am not sure what is going on.  Are we trying to flip the sign here?  If so, it's 180deg and not 90.  Maybe the 180deg flip is to get rid of the reactive component?  Very confused.  Please help.  Thanks.

[nctnico]

That schematic isn't from the XJW01 !

This thread has some info on how an LCR meter works (an LCR meter actually has a lot of similarities with an impedance analyser):
https://www.eevblog.com/forum/projects/impedance-analyzer-build-and-experiments/

[timofonic]

I wonder why aren't there a DIY high end bench DMM. Are they so difficult beasts.

[lukier]

I wonder why aren't there a DIY high end bench DMM. Are they so difficult beasts.

Our forum member Mickle T builds 6.5-8.5 digits multimeters

See here
https://www.eevblog.com/forum/metrology/lm399-based-10-v-reference/msg589601/#msg589601

and http://bbs.38hot.net/ forum.

I suppose going above 5.5 digits is not impossible with DIY, it is just very expensive. Aging the references, buying Vishay resistors in small quantities, implementing and testing multi-slope ADC and then in the end you'll need 3458A / SR104 / Fluke 732B / KVD to test the performance. So at this point it is much cheaper and easier to buy 6.5 digit meter, even a new one (e.g. 3446XA) or look for a second hand on eBay.

[Vgkid]

Running off memory. Remember we are talking about lcr meters. Which work rather easily if you break it down to a dimpler model. Lets use components in series. The basic portions to a auto balancing bridge, what almost all (digital)bridges use. To calc the unknown impedance compare the voltage on the unknown to the voltage on the known resistor. Using dome math, we can ratiometrically determine the the impedsnce on the unknown resistor(this scored me bonus kudos in my physics class). Then to calc the parameter(phase shift) we measure the timeit tskes for the waveform to either lead/lag/stsy the same. When compared to the reference resistor. Now talking components: when making precise(relatively speaking) with good resolution. The main things that need to be reasonably stable is the amplitude of the sine wave.  If you tbink about this, since we are dealing with ratios, an lm399 is overkill. Heck the GR1689 used used a 3 terminal voltage(adjusted to 2.5v) feeding the 8-bit dac. Wich was filtered to produce the sine wave. This signal was thrn passed through a high speed aml(lh002) before going through the reference and thd dut. For the reference resistor you needed stability, and low tc. Some units used flatt mica eound pww resistors. The lower end genrads used those. While the higher end oned used s102's.  The esi's and Wayne Kerrs used a similiar system. Though I  know that it seems like every ESI unit used a rather complex csl procedure. With frewuency high end ocxo's are not needed. Most units used a plain xo. With either a msnual frequency adjust(if it even has one), or a frequency offset stored in memory 1689/93.
* I can go into more detail if you guys want it.
** Im tired, and on my phone, so ignore spelling/grammar issues.

[nctnico]

I also have some experience with building LCR-ish circuits and one of the parameters you forgot about is noise. Also even when using ratios you'll still need a known-good reference. In my case I used a 0.1% resistor to achieve better than 0.5% accuracy.

[dannyf]

Are we trying to flip the sign here?

The instruments essentially measure sthe voltage across the DUT, and then the current through the DUT (via a I/V conversion).

PA0/PA1 is a simple synchronous detector: I suspect that the mcu alternately shorts PA0/PA1 to ground and float to measure V's and then I's phase relative to the control signals on PA0/PA1. With that, you get the relative phase between V and I.

I haven't looked at his code but I have used an identical approach myself on a LCR build with an AVR (AT90USB1286 to be precise). It is not very efficient btw and given the number of analog switches he is already using, it puzzles me who he didn't go down the route of a real synchronous detector route.

[dannyf]

Are they so difficult beasts.

With a build-in 24bit adc, it can be done fairly easily with a mcu (C8051F350 for example).

Using discrete parts, external adcs, it is beyond most people's capabilities if you want to go over 20-22 bits.

[Yansi]

There have been a few threads about these chinese meters, or the principles of their functionality. I have also kindly asked a member Blueskull to make a translation. If you find the thread, there is the translated schematic.  Might be interesting for you interested guys :-)

[Yansi]

Sorry, I haven't refreshed the thread, when I was writing.

[nctnico]

Are they so difficult beasts.

With a build-in 24bit adc, it can be done fairly easily with a mcu (C8051F350 for example).

Using discrete parts, external adcs, it is beyond most people's capabilities if you want to go over 20-22 bits.

You don't need that kind of ADC resolution. The input frequency is fixed so crank up the gain of the measured signal enough to get to a reasonable ADC resolution. IIRC I used 500x in my circuit to measure in the tens of milli Ohms range using a 10 bit ADC in a microcontroller. At some point noise and non-linearities bite you in the ass.

[dannyf]

The schematics (2 versions) were translated, and posted here.
https://www.eevblog.com/forum/projects/chinese-lcr-meter-schematic-with-stm32-questions-can-someone-translate-please/msg832297/#msg832297

This version, built around a 8051, differs from the STM32 version posted earlier.

1) the 8051 version uses a square wave + filter to generate the excitation signal; the STM32 version uses a dedicated DDS chip;
2) the 8051 version has a dedicated AFE, the STM32 version reuses the AFE for V and I;
3) the 8051 version has a different way of shorting the input signal in the phase detector.

Overall, I think I like the STM32 version better. I do wish it had the AFE of the 8051 version.

I suspect the STM32 version is newer.

[nctnico]

I would like to have the newer version as well since it can go to 100kHz. Having the source code would be excellent; I assume it will be easy to let it make frequency sweeps so you can also use it as some kind of impedance analyser.

[dannyf]

Here is how the phase detector works, in a nutshell.

By alternatingly shorting the input to ground, here through two switches and in the xjw lcr meter through two pins (PA0/PA1 in the STM32 version, K7 and P2.0 in the 8051 version), you are essentially controlling how the two rc filters are going to be charged up.

The output trace is the differential voltage between the upper rc filter and the lower rc filter. Traces are plotted with the input signal in 45 degrees apart, starting from 0 degrees at the top.

The traces are just as you would have expected.

So in his program, I assume he is using this approach to measure the phase differential between V and his reference signal (that drives the shorting of the input signal), and then measure the phase differential between I and his reference signal. Once you have the two, you get the phase difference between V and I.

[Vgkid]

I also have some experience with building LCR-ish circuits and one of the parameters you forgot about is noise. Also even when using ratios you'll still need a known-good reference. In my case I used a 0.1% resistor to achieve better than 0.5% accuracy.

You do need relatively stable resistors for that. In my .1% gr1658, 0.02% resistors are used. When it comes to higher end/ more advanced lcr meters. The internal resistor reference is not as important, ,but the stability is. For those, you calibrate them using a known external standard.+|- 10ppm for the 1689/93. With a known dq value to be entered as well.

[happydad]

Hi Danny,
Thanks for the sim on synchronous detector.  I speculate that he is switching between PA0-high/PA1-low and PA0-low/PA1-high at 1/4 the signal rate, thus producing 0  and 90deg sampling of both voltage and current for a total of 4 readings.  That will give you all you need to know about ESR and C.  Sounds about right?

[Yansi]

Yes, you measure the voltage and currect accross the DUT, both as complex variables (in phase and quadrature), so you can use the standard formulas for complex impedances to calculate any impedance.

[happydad]

OK Thanks.  So to recap, I am picturing sine waves of V and I  that are 90 degrees apart right up to the synchronous detector.  After the synchronous detection, there is an rc filter at 16Hz?  Isn't that a little low?