LCR Cal Test PCB

[KungFuJosh]

A bunch of us have been comparing the accuracy and ranges of our LCR meters, and to compare to the "official" cal cert Martin72 received with his ST2830, I made my own test report and a test PCB to go with it.

I've been through 3 out of 4 versions of the board (skipped the 3rd after learning how to test dissipation factor). I'm now getting ready to make a 5th, and hopefully final version of the board. I could use some ideas and a little insight in finishing the last design.

There's some extra space on the R side, as I recently removed 1GΩ and 100MΩ test points. If somebody things I should keep one or both, let me know why you feel that way. I'm also considering adding 1pF back onto the C side, but it's lower than typically useful on these tests, but I do know the Shannon ST42 can test it fine.

Currently, my biggest question is finding a more reliable method for kelvin clip test points. What I have now is 3mm pads on top and bottom of the PCB running to the test components. For my kelvin clips to get a good connection, I have to add solder to both points around the board, and the resulting connection isn't amazing. It requires a little wiggling to get the best connection. I want something more stable.

Here's what the board design looks like currently:

[KungFuJosh]

Here's the most recent example of my test done with V4 of the board.

[Martin72]

That doesn't look too bad.
Personally, I would stop at 10mH, 100µF, 1M Ohm.
And start at 100nH, 10pF and 1 Ohm.
While finding very small tolerances for resistors is not a problem, I had not found anything below 2% for capacitors, for example, and nothing below 5% for most coils.
What did you use, where can you find it?

[KungFuJosh]

That doesn't look too bad.
Personally, I would stop at 10mH, 100µF, 1M Ohm.
And start at 100nH, 10pF and 1 Ohm.
While finding very small tolerances for resistors is not a problem, I had not found anything below 2% for capacitors, for example, and nothing below 5% for most coils.
What did you use, where can you find it?

The first reason I got into testing inductance is for making Wah Wah pedals. Those inductors range from 500mH to 700mH, so I'll always include them. How accurate they are on their own doesn't matter. Once the board is finalized, it can be calibrated and then the calibrated value is the only thing that matters. I also test guitar pickups, and the (standard) ones I've done range between 3 to 5 Henries so far.

I'll put together a BOM including the parts. Everything except the wah inductor was from Mouser. Most, if not all (remaining) resistors were 0.1%. The rest is probably along the lines of what you said. High values tend to be worse, but I'll buy extras (including higher values) and find what tests to the desired value, and go with that in the end. Seeing 75µF or so when it's supposed to be 100µF is annoying.

I still think finding a better method for connecting kelvin probes is high priority. Or scrapping the kelvin probes entirely, whichever works better. 😉 I think having the lower values isolated might be enough on their own. I can test this by using the kelvin clips only touching the board top test points. Still though, I would prefer a better kelvin option.

I've seen some smaller pins that might work, but I dunno if that's the best way, especially if they get bent over time.

Thanks,
Josh

[Martin72]

The pins that DMMCheck Plus uses are quite stable, I can still reach them easily with Kelvin clamps.

[tautech]

Currently, my biggest question is finding a more reliable method for kelvin clip test points.

SMD test points ?
https://www.keyelco.com/category.cfm/Test-Points-Tips-Probes-Clips/Test-Points-Surface-Mount/id/523

[KungFuJosh]

The first version of my test PCB had similar style turrets as the test points, but people complained about that not being a kelvin connection. For a kelvin connection, there needs to be two connectors for each side of each test point.

Doing a top mounted pair of pins or other connector is the idea. But the single connection per test point is not ideal.

Here's the old V1 board:

[KungFuJosh]

I should mention my plan is likely to change to two parallel traces run to each test point (instead of top/bottom traces). The question is the best connectors to use with them. Or if somebody has a better idea, I'm open.

[Martin72]

Hi,

For a kelvin connection, there needs to be two connectors for each side of each test point.

 
Doesn't make sense in my eyes.
You can only connect 2 points with normal Kelvin test leads anyway.
The four-wire measurement is used to eliminate the cable lengths.

[KungFuJosh]

Hi,

For a kelvin connection, there needs to be two connectors for each side of each test point.

 
Doesn't make sense in my eyes.
You can only connect 2 points with normal Kelvin test leads anyway.
The four-wire measurement is used to eliminate the cable lengths.

Doesn't make sense to me either, except that other members here said the same thing, but also for eliminating the traces going to the test points. There was discussion about it when I first posted the picture of the V1 version of the test board in the 2830/2832 improvement thread. If I misunderstood something, or I'm missing something else, I'd love to be educated on the subject. But not enough to go to you know where. 😉😉

[mawyatt]

Here's an idea where the Kelvin Clips "clamp" to Plated Thru Holes in the PCB by "Splitting" open the Kelvin Clips this allows the Sensing Port to take place where the Force Current enters the Thru Hole DUT contact lead pad, or the SMD DUT pad, thus "sensing" very close to the DUT leads/termination.

You should include an "Open" and "Short" for LCR Calibration and would expect to give good results down into the tens of milliohms impedance regions but haven't verified such.

Best,

[shabaz]

Nice-looking board.

Here's an idea where the Kelvin Clips "clamp" to Plated Thru Holes

Great ideas. Also the trace lengths should be similar-ish from component to component, and (if there's space) a U.FL pad could be interesting per component, so that users could optionally use the board with VNAs too.

Another thing could be to split the SMD pads (so that there are four pads for each component), i.e. have the kelvin connections separated all the way to the pad.

Also, instead of the silkscreen legends, maybe silkscreen white filled rectangles could be interesting, so that people can populate with whatever components meet their needs, and a permanent pen can be used to label the components.
Another idea could be to retain the current text silkscreen, but ensure there's a faint box outline where a user could apply a printed sticker if they wish to change the value.

Anyway, I have no idea if these extra suggestions are any good, or they might be overkill, possibly too much effort if the board is already mostly designed, for the small additional benefit.

[jchw4]

The pins that DMMCheck Plus uses are quite stable, I can still reach them easily with Kelvin clamps.

I found that I get different readings on the 1uH inductor depending on how I put the clips. The mounting screw in the corner doesn't allow symmetric clip positioning for the most common clip size.
Looking at your picture, I guess that the correct way for that part is to put the left clip "from the front" and another one "from the right side" kind of "behind the screw".

BUT the vendor does not specify that and we have no information on how they make connections during the calibration.
You may suggest using smaller clips for that part, but the funny part is that I am using BK Precision TL8KC1 clips with my DE5000.
And I guess that BK Precision is shipping exactly the same clips with all their LCR meters including the one that DMMCheck Plus is using for the calibration...

As you now have quite a few LCR meters  , you probably have a few different clips too. Could you measure that inductance with different clips and share the result?

[Phil1977]

Very nice PCB - if there is a possibility to get one in Europe I´d like to buy one.

From my point of view it´d be perfect if there would be two additional "Sense"-testpads directly next to each reference component. That way you could connect your 4W-sense-probes as perfect as possible and everything else on the PCB stays the same.

For usage of the 4W-Kelvin-crocodile clamps the vertical test pins should be fine. Then they only compensate the test cables and not the PCB influence, but for most of the values that should be okay.

[KungFuJosh]

Here's an idea where the Kelvin Clips "clamp" to Plated Thru Holes in the PCB by "Splitting" open the Kelvin Clips this allows the Sensing Port to take place where the Force Current enters the Thru Hole DUT contact lead pad, or the SMD DUT pad, thus "sensing" very close to the DUT leads/termination.

I love it, and not just because I was hoping somebody smarter than me would suggest it. 😉 PTH vias shaped appropriately I think would work great, and I much prefer that over using pads as I am currently.

You should include an "Open" and "Short" for LCR Calibration and would expect to give good results down into the tens of milliohms impedance regions but haven't verified such.

I love this idea too. I hadn't thought much about it, but I recall your advice on doing similar for the SMT fixture on our LCRs.

Another thing could be to split the SMD pads (so that there are four pads for each component), i.e. have the kelvin connections separated all the way to the pad.

That's an interesting idea. However, I think if it has any benefit, it's beyond what's needed for this specific test. The idea gives me two concerns: 1. that it would reduce the solder pad area; 2. That the 2 pads being soldered to the same end of a component doesn't seem like there would be any significant difference than connecting both kelvin connections to the same pad (as they are now). I could be wrong of course, but for the purpose of this PCB, regular pads seem fine.

Also, instead of the silkscreen legends, maybe silkscreen white filled rectangles could be interesting, so that people can populate with whatever components meet their needs, and a permanent pen can be used to label the components.
Another idea could be to retain the current text silkscreen, but ensure there's a faint box outline where a user could apply a printed sticker if they wish to change the value.

Anyway, I have no idea if these extra suggestions are any good, or they might be overkill, possibly too much effort if the board is already mostly designed, for the small additional benefit.

I haven't decided what I'm going to do with this when it's done done. I'll at least share the Gerbers so others can have them made. I think if people want to do stickers, it doesn't really matter what I leave on the silkscreen.

All ideas are appreciated! Thank you.

Very nice PCB - if there is a possibility to get one in Europe I´d like to buy one.

From my point of view it´d be perfect if there would be two additional "Sense"-testpads directly next to each reference component. That way you could connect your 4W-sense-probes as perfect as possible and everything else on the PCB stays the same.

For usage of the 4W-Kelvin-crocodile clamps the vertical test pins should be fine. Then they only compensate the test cables and not the PCB influence, but for most of the values that should be okay.

Thank you! As I mentioned above, I'll at least be sharing the Gerbers when this is done. Then anybody can order 5 boards for much less than the cost of shipping to Europe.

Yes, others have mentioned that they would prefer having separate connections for each component. Personally, I don't like that idea...because I'm lazy and testing already takes too long. For small values where that might have an effect, I agree (and it's already implemented that way). But for most of the stuff, it makes no difference.

If you check my cal test results I posted above, you'll see that the tweezers and the bench LCR were in excellent agreement for most items. It's probably useful to mention that the tweezers tested components directly, and did not use any of the kelvin connection test points. For the results being as close as they are, I think the logic behind my connection scheme is good enough for me. 😉

Thanks,
Josh

[KungFuJosh]

Here's a question: Does anybody else want a couple additional values?

I was thinking about adding 1pF back to the board.

Also was thinking whether what I have for resistance is enough. It used to have 100MΩ and 1GΩ, but those values aren't that useful. I'm thinking the real estate would be more useful for adding 1pF in.

ETA: I know it doesn't matter if people relabel it, but maybe somebody wants more things available in a specific category.

[coromonadalix]

unless an huge demand,   i think your kit is fine,  people will never be 100% pleased, your not down to metrology levels

if you are satisfied,  that's what matters,   turret test points are good enough ... 

my 2x kelvin test clips are shorting only on the same connection once they are closed  ...  no need for 4 separated clips ....

others  will be able to do their test rig ...

thks 

[KungFuJosh]

Thanks! I have no interest in getting to metrology levels, that's why I posted in this section. 😉

I am going to convert to Mike's idea of PTH vias for the kelvin connectors. I like that it's more accurate, less work, and costs less since there's no need for expensive turrets.

[Martin72]

As you now have quite a few LCR meters  , you probably have a few different clips too. Could you measure that inductance with different clips and share the result?

Yes, there are a few now, but they are basically the same because they always have the same clips.

[mawyatt]

The Tonghui Kelvin Clips are slightly smaller than the generic types (we have a bunch of these also), and of much higher quality and have embedded Hcur, Hpot, Lcur and Lpot Labels on the Clip handles.

Best,

[Martin72]

The ET3502 makes a good impression in the few measurements I have taken, and if this continues to be the case, I will keep an eye out for other cables.
Maybe I'll buy some from Sourcetronic.

[KungFuJosh]

Does anybody think I need bigger than a 2mm x 5mm hole for each of the kelvin tips on their random kelvin clamps?

[EggertEnjoyer123]

Make sure you consider tempco.

Most ceramic capacitors have bad temperature coefficient, and you should probably get C0G/NP0 capacitors or film capacitors. Your standard X5V or whatever capacitors won't be accurate if the temperature changes.

I'm not sure what the temperature coefficient for inductors is. Might have to check that too

[KungFuJosh]

Thank you. They are all C0G/NP0 MLCC for that reason, except where the values are too high, then they're X7R.

[KungFuJosh]

optionally use the board with VNAs too.

I'm curious, as I've never had a VNA. What about this board would be useful to VNA users?