Sensitive LCR Meter Measurements

[mawyatt]

Decided to post this after doing some evaluation measurements over here:

https://www.eevblog.com/forum/testgear/lcr-cal-test-pcb/

When doing sensitive High Z LCR measurements (for example, small C at low F) one must be aware of the impinging E fields.

Kelvin Clips like the TH26011CS and others generally do a good job but can fail at higher impedances often encountered during small capacitor measurements. If one studies the TH Kelvin Clip construction carefully, the entire length of the handle half is unshielded right to the clip tips. This allows these conductors to radiate and be susceptible to localized disturbances such as your hands, affecting any attempted quality measurements. Since most don't have "Calibrated Hands" this effect can't be totally eliminate by careful calibration unless one wants to glue their hands to the Kelvin Clip Handles

Tonghui specs the TH260011CS to 100KHz which seems reasonable, they have a version that specs higher but we don't have such so can't say what is different, altho likely the shields from the cable extends closer to the Kelvin Clip tips. There may be others from quality sources that also a specified to higher frequency use.

This is one reason we prefer the use of "dedicated" fixtures like the TH26048A and SMD types (TH26008) which attach directly to the LCR meter BNCs, since these have a shielded enclosure for the conductors with keeps most of the fields contained and repeatable.

The SMD Fixture types do have "exposed conductors" without shields in the SMD Plungers, altho the wiring to BNCs is all within the case shielding and the plungers are somewhat fixed just above the case ground shield so the E fields are sorta "bent" down towards the case cover shield. One can see the effects by moving your hand close to the plunger to observe the effects, however not nearly as sensitive as the TH Kelvin Clips mentioned.

There are also some nice SMD Tweezers cables like the TH26009B for use with LCR meters, these are rated to 15MHz and carry the shields right to the tweezer tips. 

There's also the issues at the opposite end of impedances like the low Z stuff, but we'll pause on that for the moment.

Anyway, we've known this for some time (shame on us for not posting earlier     ) but never considered posting until making the measurements mentioned, hopefully saving others from additional LCR measurement uncertainty.

Edit: When posting below we recalled an illustration that Hioki had which showed a conductive work base connected to the "Guard Terminal" on the LCR meter for sensitive measurements (it's in the IM3536 users manual). We found our PCBite plate is conductive and connected it to the LCR Meter "Guard Terminal" as shown below. Results improved considerably wrt to TH260011CS Kelvin Clip hand location sensitivity with the Guarded Conductive Plate, altho close by the handle still showed high sensitivity as would be expected since the plate can't have much effect for the near field of the handle conductor within

Best,

[KungFuJosh]

So are we gonna add some shielding to our kelvin clips now? 😉

[mawyatt]

We're not!! Just going to be careful of use and understand the sensitivity of the measurements at hand

Later we might investigate Kelvin Clips with higher rated frequency use, but even then one is still stuck with the cable length effects, so best to just stay at low frequencies and be aware of the actual Kelvin configuration  limitations.

Honestly was surprised to find the Tonghui Clip handles totally unshielded, that was something we didn't know until evaluating your PCB with the 1pF test cap, and trying to determine the source on measurement uncertainty and noting how sensitive hand location was. I mean within ~100mm of the clips, and began to show effects as we would move our hand to write the measurement value only to see a different value on the IM3536!!

Operating on a conductive workbench plate would likely help, but we don't have such. Hioki has an illustration of using a conductive plate attached to the "Guard" terminal which we should probably do. Need to find such tho!! Edit: Found one, the PCbite base is conductive, added above!!

Best,

[KungFuJosh]

Did the plate help with the 1pF test?

[mawyatt]

Yep, sure did

See edit in OP.

Best,

[Sensorcat]

... the entire length of the handle half is unshielded right to the clip tips.

I'm not sure if I understand correctly: Does that mean the full path from the little box, where the big and the four small cable meet, to the tips, unshielded? If yes, that's quite disappointing! Raises the question if other manufacturers do better...

[mawyatt]

... the entire length of the handle half is unshielded right to the clip tips.

I'm not sure if I understand correctly: Does that mean the full path from the little box, where the big and the four small cable meet, to the tips, unshielded? If yes, that's quite disappointing! Raises the question if other manufacturers do better...

No, the 4 wires are shielded but at the Kelvin Clip handles where the wire enters the handle, the wires are not attached to any shield there, just at the box. The center conductor goes to the long metal "blade" that becomes the Kelvin Clip tip at the handle. So the entire length of there handle is not shielded and why your hand has such an influence on the measurement.

See image below.

The popular cheap Kelvin Clips you often see, are no better. Same unshielded conductor thru the handle to the Clip Tip.

Also, see OP for information on using Guard Terminal to help suppress external measurement influences for high Z measurements.

Best,

[mawyatt]

Using LCR Meter Guard Function for SMD Fixture, see:

https://www.eevblog.com/forum/projects/different-type-lcr-smd-fixture/

Best,

[nctnico]

Did you see this thread: https://www.eevblog.com/forum/testgear/zurich-instruments-mfia-impedance-analyzer-(z-1m-1t)-review-teardown/ ? There could be some interesting information to be obtained from the review and teardown.

[Martin72]

I'm not sure if I understand correctly: Does that mean the full path from the little box, where the big and the four small cable meet, to the tips, unshielded? If yes, that's quite disappointing! Raises the question if other manufacturers do better...

That's why I'm a friend of measuring components with as few cable lengths as possible, the higher the measurement frequency.
So it's no surprise, for example, that the ET4410/4510 LCRs perform rather poorly at 100kHz if they are supplied with cables that TongHui/Sourcetronic specify as max. 10kHz.

[jonpaul]

See HP Impedance Measurement Handbook of 1980s.

https://www.keysight.com/us/en/assets/7018-06840/application-notes/5950-3000.pdf

Depeneong on the Z range either 4 term (kelvin) or guarded measurement is used, but not both at one time.


J

[mawyatt]

Did you see this thread: https://www.eevblog.com/forum/testgear/zurich-instruments-mfia-impedance-analyzer-(z-1m-1t)-review-teardown/ ? There could be some interesting information to be obtained from the review and teardown.

Did not see this until now, that's a really nice instrument, wish we could afford such!!

Thanks for the note

Best,

[mawyatt]

See HP Impedance Measurement Handbook of 1980s.

https://www.keysight.com/us/en/assets/7018-06840/application-notes/5950-3000.pdf

Depeneong on the Z range either 4 term (kelvin) or guarded measurement is used, but not both at one time.


J

Actually the technique shown in OP by Hioki works quite well indeed with Kelvin Clips and using the Guard Plate. We did so with Hioki IM3536, Tonghui TH26011CS Kelvin Clips and PCBrite Conductive plate as mentioned.

We really should look into getting some Kelvin Clips rated higher than the TH26011CS tho, they are rated only to 100KHz.

Best,

[Martin72]

I don't know how well Tonghui accessories are available in the states, but Sourcetronic ships worldwide:

https://www.sourcetronic.com/shop/en/kelvin-clip-terminal-st26011bs.html

Rated up to 5Mhz(pic shows the 300kHz version).

[KungFuJosh]

I don't know how well Tonghui accessories are available in the states, but Sourcetronic ships worldwide:

https://www.sourcetronic.com/shop/en/kelvin-clip-terminal-st26011bs.html

Rated up to 5Mhz(pic shows the 300kHz version).

I would ask them to confirm that. Every TH version that looks like that, including the same model number as the URL maxes out at 100kHz.

https://www.tonghui.com.cn/en/product_list_192.html

[Martin72]

Jeez, this is a lot they offer...

And you're right, not a single cable goes over 100kHz(except the ones with a smd tweezer) - that needs to be clarified, I'll write to sourcetronic.

[mawyatt]

Yeah, those are ones we've been looking at, rated to 5MHz. They look exactly like the CS versions we have (rated to 100KHz).

Even the Clips look identical and from the images looks as if the Kelvin Handles are the same with the "exposed" unshielded conductor blades.

Begs the question, what's the difference, maybe the cables....or maybe they are just tested at 5MHz??

Best,

[Martin72]

I'll find out, the guys from Sourcetronic are even in the same city where I live...
Maybe it's actually the cables.
10kHz/100kHz is still more “LF”, where the properties of coax cables still play a subordinate role.
But what do you do in the Mhz range, where cables with their pFs already play a parasitic role....

[Martin72]

I have now written to them and asked for a data sheet, together with an explanation as to why.
The things look so similar, but it may well be that the cables from Sourcetronic do not come from TongHui or are “different” and are manufactured by TongHui for Sourcetronic.
We will know soon.

[mawyatt]

I'll find out, the guys from Sourcetronic are even in the same city where I live...
Maybe it's actually the cables.
10kHz/100kHz is still more “LF”, where the properties of coax cables still play a subordinate role.
But what do you do in the Mhz range, where cables with their pFs already play a parasitic role....

It's not so much the cables capacitance as the exposed cable signal fields that are an issue. If the cable shields were carried up to the Kelvin Clip tips like the Tonghui TH26009B Kelvin Tweezers, this would increase the over cable capacitance but produce better results because the cable signal fields are no longer exposed along the handle. So the Kelvin Clip Handles wouldn't have exposed signal fields which your hand changes and thus changes the measurement underway, only the Kelvin tips would have exposed fields but these are near the DUT, and far from your hands and not influenced by such.

Honestly quite surprised they are made this way with the long exposed conductive blade to tip instead of a short conductive tip connected to a shielded wire going thru the handle, as it wouldn't be difficult to carry the cable shields right to the short conductive Kelvin Clip tips.


All the Kelvin Clips we have, and we have quite a few expensive and cheap, are not shielded up to the Clip tips with exception of the Tonghui Tweezers, probably reason they are expensive and rated to 15MHz

Best,

[mawyatt]

Following up on the issue with the Kelvin Type clips we decided to do a quick experiment to see how well the quality Tonghui TH26009B Shielded Tweezers would compare to a direct connected LCR Meter SMD fixture shown here:

For some reason link doesn't work, so search for:

"cheap-smd-lcr-fixture-the-good-bad-and-ugly!"

Honestly, we don't have a quality SMD fixture (know we should have, but don't), however we do have a pair of cheap clone types that have been modified (see above) that produce highly repeatable results. Like < 0.01pF with small capacitance SMD chips, so no fear in utilizing these SMD fixtures as a reference to compare the TH26009B Tweezers with.

First off we setup the SMD fixture and did the usual Short and Open Cal with the IM3536. For the Open we pried the plunger Opening ~ same as 0603 device, to Cal out the parasitic capacitances at that spacing (important if one is doing small capacitance measurements like 1pF!!). Precision C0G type 0603 values of 1, 10, 100 and 1000pF were selected for comparisons.

After the SMD fixture measurements we setup the TH Tweezers on top of a conductive plate (PC Bite) connected to the IM3536 Guard Terminal. Operating with the IM3536 Guard Enabled has shown better repeatability with all cable type clips with high Z measurements, so we use this when necessary. We also spaced the Tweezers ~ same as 0603 spacing for Open Cal for same reasons.

Here's the IM3536 results at 1MHz:

Capacitance             SMD Fixture                   TH26009B
1pF                         1.0004pF                       0.9937pF
10pF                       9.9692pF                       9.9303pF
100pF                     99.9551pF                     99.7126pF
1nF                         992.236pF                     992.838pF

The results produced were stable and little variations.

This shows that quality Tweezer Kelvin Types like TH26009B which employs shielded conductors right to the tweezer tips can produce good results with small DUT capacitances with carefully attention to setup and measurement details.

However, because the Kelvin connection is not made at the very tip end, one would expect results with very small impedances to suffer. Here even the SMD fixtures mentioned above aren't going to be optimum as the Kelvin connection on these types is at the plunger body and not the plunger tip. This was one reason we developed the experimental fixture type shown here and it's various refinements:

https://www.eevblog.com/forum/projects/different-type-lcr-smd-fixture/msg5061028/#msg5061028

Anyway, seems the Quality Tweezers like the TH26009B which carry the signal shields to the tips are capable of quality, repeatable measurements with small capacitance DUTs (high Z), even at higher frequencies.

Best,

[mawyatt]

This effort created another where we wanted to see how well things worked with low Z SMDs, so selected some precision low ohms 2512 SMDs.

After a few futile efforts we gave up on the TH26009B Tweezers, too difficult to create a good repeatable Zero Ohms Cal, and the readings were unstable jumping around too much for our liking. We needed calibrated hands

Maybe if we had more patience, but were somewhat in a hurry so proceeded to the SMD Fixture mentioned above in "Good, Bad and Ugly" which for some reason link didn't work, so you'll need to do a search for the thread. This fixture directly mounts to the IM3536 and should produce better results, which it did but far from what we would have liked!!

Some difficultly with good Zero Ohms Cal, remember we're talking about sub-milliohms now!! Readings took awhile to stabilize also, so room for improvement!!

Then we engaged the Special SMD Fixture approach indicated here:

https://www.eevblog.com/forum/projects/different-type-lcr-smd-fixture/msg5061028/#msg5061028

This technique works with the Split Kelvin Connections Force-Sense right up to the DUT SMD and allowing the DUT SMD end caps to make the Force-Sense Kelvin connection. Getting a good Zero Ohm Cal is still tricky, but nowhere as difficult as with the Standard SMD Fixture, or the Tweezers, and getting good stable readings is a breeze comparatively

We first did the Special SMD Fixture, then the Standard SMD Fixture. Afterwards we decided to reinstall the Special Fixture, do a quick Zero Cal, and rerun the test DUT Resistors, note the repeatability in the Special SMD Fixture results.

Resistor Value
in Milliohms       Special SMD Fixture  (Repeat see text above)                      Standard SMD Fixture     

1                       0.80                           (0.80)      {0.82}     <0.81>                       ~1.23
5                       4.84                           (4.82)      {4.82}     <4.86>                     ~5.45
10                     9.90                           (9.87)      {9.86}      <9.92>                     ~11.98
15                     14.70                         (14.71)     {14.72}    <14.73>                     ~13.29
20                     19.83                         (19.84)     {19.85}    <19.85>                    ~17.84
40                     39.87                         (39.88)     {39.88}    <39.88>                      44.87
50                     49.87                         (49.91)     {49.91}    <49.93>                       48.29
100                   100.21                       (100.23)    {100.21}  <100.27>                       99.28 
   
"~" indicates some variability in readings

Anyway, the Special SMD fixture seems to be a good candidate fixture for low Z SMD LCR Meter measurements and curious to see if the Guard Plate under the SMD measurement provides additional benefit (see link above).

Edit: 4~5 hours later redid the measurements shown as {#} for additional results.

Edit: From <#> post 24 below.

Best                       

[exmadscientist]

This seems like a good thread for dumping one of my old observations:

When using an Agilent E4980A LCR meter or its predecessor the 4284A, with leads longer than I'd like but difficult to make better, doing the open and short corrections twice (so open-short-open-short or vice versa) gave significantly better results, especially in terms of repeatability. I can only guess that it's easier to dial in a small correction than a large one, and the first round correction makes the second round smaller?

Either way -- I've done that ever since when using an LCR meter and trying to get any level of precision.

[mawyatt]

Good point, will give that a try next time and see if it improves the Calibrations. Not sure how these Calibration Measurements are utilized internally within the LCR Meter, but doing it twice certainly can't hurt!!

Best,

[mawyatt]

Just turned power on to the IM3536, no Calibration (using previous Calibration from yesterday), and started measuring the pervious 2512 Precision Resistors.

This is with no warm up, no Calibration, and just quickly making measurements.

Note results added in <#> in above post number 21.

Here's what we use for Zero Ohm Calibrations, short lengths of #12 Cu wire with ends filed smooth for Standard SMD Fixture. For Special SMD fixture is a short flat rectangular conductor slightly bent and ends rounded, also shown is an "L" #12 Cu wire we use for Kelvin Clips.

2512 test resistors are 1,5,10,15,20,40,50 & 100 milli-ohms.

Best,