TH2830 vs TH2832

[nctnico]

I feel an LCR meter upgrade coming up... But I really need a new laptop first 

[Martin72]

My upgrade will take a while, but then I won't need another one:
https://www.eevblog.com/forum/repair/hioki-3532-50-lcr-schematic-service-manual-wanted/msg4969691/#msg4969691

[KungFuJosh]

And then with open/short calibration...

I did the open/short after the patient 30 seconds of waiting. 😉

Here's a warm up time comparison. Open/short correction applied after the specified time. Note: it's a THT fixture, the resistor is a 1% CMF, but it was not in the same position for both tests.

For my needs in general, the 30s time is plenty accurate.

[KungFuJosh]

It´s itching in my fingers to take the file and.....No.

When does your calibration expire? 😉

[luudee]

I did the open/short after the patient 30 seconds of waiting. 😉

Here's a warm up time comparison. Open/short correction applied after the specified time. Note: it's a THT fixture, the resistor is a 1% CMF, but it was not in the same position for both tests.

For my needs in general, the 30s time is plenty accurate.

Hmm, looks like 200 KHz is still off ...

luudee

[KungFuJosh]

Hmm, looks like 200 KHz is still off ...

luudee

Yes, but overall significant improvement. When I do a full test we'll see how everything else is. 100KΩ accuracy isn't expected to be great at 200kHz, but that is worse than it should be.

What do 100KΩ measurements look like across the range with your TH2832?

Thanks,
Josh

[luudee]

Yes, but overall significant improvement. When I do a full test we'll see how everything else is. 100KΩ accuracy isn't expected to be great at 200kHz, but that is worse than it should be.

What do 100KΩ measurements look like across the range with your TH2832?

Thanks,
Josh

Josh,

please look at post #165 ...

I will redo the tests. I am ordering some 0.1% resistors from mouser, 100 ohm, 10K, 100K, 499K & 1M

That should give a basic feel for it's performance ...

luudee

[KungFuJosh]

please look at post #165 ...

I will redo the tests. I am ordering some 0.1% resistors from mouser, 100 ohm, 10K, 100K, 499K & 1M

Oh yeah, that is much better.

I look forward to seeing the extended results.

Thanks,
Josh

[mawyatt]

A simple RC model using the 200KHz value predicts a shunt C of ~2.14pF for the R, which seems large. This value C would produce a 100KHz reading of ~97.7K.

So it seems at 200KHz this is not utilizing a proper internal calibration value?

Agree, luudee's test should revel much!!

Also a quick look at the ratio of the 30 second and 1 hour warm up values across all frequencies and DC hints that the 200KHz range is suspect, only the 200KHz range has significant change!!

Best,

[KungFuJosh]

So it seems at 200KHz this is not utilizing a proper internal calibration value?

My thoughts exactly. However, ST seems to think that there's only open/close correction, and no other means of adjustment. I don't know if that's true.

If there is a way to make adjustments, then that would be awesome. Figuring that out is another story.

There is manual open/close correction that can be done, but I assume that will get reset with ever new correction...and I don't remember if you can actually adjust the values. I'll mess with it.

[KungFuJosh]

I got my test board in and populated it. What a time saver! I got the whole test done on the ST2832 in just under an hour. Setting up the SMD stuff on the fixture one at a time before took hours to get the test done, and I didn't even test everything. Testing with the ST42 took under 20 minutes.

Here's my test board:


Test setup:


Results attached, along with the BOM of the parts on the board.

[zrq]

Is it possible to try finding the factory calibration routine in the meter (hope it exists) and do it if you have some parts that you trust?

[KungFuJosh]

Is it possible to try finding the factory calibration routine in the meter (hope it exists) and do it if you have some parts that you trust?

I have no idea! But I love the idea. If it's as easy as running an inbuilt routine and attaching some stuff, I can do that. 😉 I don't know if such a routine exists, and I'm certain Tonghui or ST won't share any service manuals. Of course, now I want to try and find a service manual anyway.

There is a "spot correction" option in the open/short correction screen. I think that might work along the lines of what you're suggesting for specific frequency correction. Only 2 problems with that so far: 1. I haven't figured out how to set it yet (haven't really tried though, didn't RTFM for that either), and 2. Every time you do open/short correction, it clears the spot correction.

[nctnico]

I got my test board in and populated it. What a time saver! I got the whole test done on the ST2832 in just under an hour. Setting up the SMD stuff on the fixture one at a time before took hours to get the test done, and I didn't even test everything. Testing with the ST42 took under 20 minutes.

Here's my test board:

Why aren't the traces to the components split to have Kelvin connections all the way to the parts? I see a 10nH inductor there for example.

[KungFuJosh]

Why aren't the traces to the components split to have Kelvin connections all the way to the parts? I see a 10nH inductor there for example.

Two reasons:

1. I didn't know I should do that? 😉
2. It's perty good/accurate the way it is.

[Martin72]

The idea itself is good. 
But when measuring the very low inductances and capacitors you could get "problems", because every mm way counts.

[KungFuJosh]

Well, now I know. If I ever design something like this again, I'll do the split traces. Of course, I ordered 10, and made 2, and I only need 1, so... 🤷🤣

[Martin72]

If I may "steal" your idea, I would design it something like DMMCheck had done with the L/C option board..

[KungFuJosh]

If I may "steal" your idea, I would design it something like DMMCheck had done with the L/C option board..

🤣🤣🤣 That's where I got the idea from! But I went for an extended range.

[Martin72]

Yepp, but with a little difference..

[KungFuJosh]

Yepp, but with a little difference..

Oh, yeah, I didn't want them in series though. My tests apply to the ST2832 and the ST42. If my board was dependent on series results, then I couldn't test with tweezers.

Why do you like the other way better?

[nctnico]

If I may "steal" your idea, I would design it something like DMMCheck had done with the L/C option board..

With such traces you can end up in tens of nH. That would already add more than 1% of uncertainty to the 1uH inductor on the DMM check board. Not to mention the stubs you get. IOW: IMHO the DMM check style board is very unsuitable for measuring small component values accurately. There is a very good reason LCR meters have these fancy adapters for measuring components.

[Martin72]

Nico...
I wouldn´t start with nH, because like I said before in this value range every mm counts..
The LC-Optionboard: Values were measured out by a frequency of 10kHz, it was "wise" from them not to go higher.

[KungFuJosh]

You're both missing the point. The actual value of the target isn't that important. What matters most is the value you get from the test device at the testing points.

For example, if you send the loaded PCB out for calibration, the cal lab will certify the values they read at each test point. That calibration is your reference, not what values showed up on Mouser's shopping cart. The specific component value doesn't matter at all, as long as it can be calibration verified. Then you check against the calibration values to see if your meter is in spec. Is this inductor 1.0039345mH or 1.003955mH? If that difference matters to you, you need to get it calibrated anyway. For me, if it says 1.00mH, the rest of the digits aren't that interesting.

For my needs, all the values were close enough to see where there are and are not issues with my LCR meter. I doubt I'd ever have the PCB calibrated, but I can refer to my saved cal data and see if my meters are still consistent in a year or 10.

All that said, the values between the ST42 and the ST2832 were close enough. The ST42 tweezers were used directly on each test device. I suppose, if you actually care enough to get it calibrated, you can have a direct tweezer test and a separate test point test done for reference of both testing methods.

I think you're both thinking too much like this is a functional circuit where the component value actually has an effect on something.

[nctnico]

You're both missing the point. The actual value of the target isn't that important. What matters most is the value you get from the test device at the testing points.

That doesn't make sense because at the point where the board itself starts to influence the component value significantly (which is definitely the case for the low component values) you won't have a sensible measurement to begin with. Especially with the large antennas attached to the parts that will pickup whatever noise is in the air.

If you want to make an LCR calibration standard on a board, then make a 4 layer board that has 4 BNCs that can plug into an LCR meter, use the outer layers as shield and run the signals from the BNCs on the inner layers seperately to pads to solder a part. IOW: the traces connect at the part only. This board can only hold 1 component. But even then you can only test parts that have a significantly higher value compared to the variations between the boards. Adding a shield can over the the component helps to improve the stability of the readings even further. BTW: you will need several of these boards including an open and a shorted one.