Cheap SMD LCR Fixture, the Good, Bad and Ugly!

[mawyatt]

However, I do not agree with the procedure you use for the "Short Circuit Compensation".
For this step you must not use any thickness between the measurement points, otherwise you lose an important parameter of the DUT which substantially coincides with the equivalent series inductance associated with its physical length. (as example, think to the series resonance of a capacitor... ).
More, during SC Trimming the effects of the fringe capacitance to the fixture interior case is absolutely negligible (due to the voltages near zero...).
Even leading companies like R&S and Keysight require short circuit trimming with the tips touching each other.

Personally, in some specific cases, in order to avoid including in the measurement the contact resistance between the tips and the DUT,
I carry out the short circuit trimming by interposing between the tips a thin layer of tin (of negligible thickness compared to the length of the DUT).

Different ideas that I hope will help both...

Depends on what you are measuring, if tiny inductance then yes don't use a Zero Ohm Shunt SMD as the length invokes inductance, however if measuring small SMD resistors of small values then using the Zero Ohm Shunt SMD allows one to partially emulate the plunger contact pressure and resistance better.

Edit: Using Terman Formula as an estimate of the self inductance of a rectangular conductor of width and length of a SMD (est conductor thickness of 0.1mm) we get as an approximation:

0603  40pH
0805  54pH
1206  94pH

(Results similar with Grover formula for circular wire of dia as width and same length)

Contact pressure and resistance is a built-in issue with these plunger type SMD fixtures with the ones we have and varies quite a bit, especially with SMD length. When the SMD device is soldered into a PCB the entire end surface is electrically engaged, however when measuring with the fixture only the plunger tiny contact area is measured, so our thought was mimicking this with a same size Zero Ohm SMD would better compensate for such during the Short Calibration.

Of course YMMV.

Best,

[mawyatt]

I have one of these test fixtures too, the cheapest I could find on Aliexpress, and with some tender love, it works quite well on my HP 4276A LCZ meter.

I improved the grounding inside, cleaned out the overspray from inside with acetone to allow proper contact to the case for the BNC's and the divider panel, tightened everything up, and stuck it all back together with stainless steel screws just for a bit more pretty.

I am wondering though, is it worth using coax (shield connected at the BNC end) for the internal wiring?
Maybe using some RG316 or similar. Would this help improve things, or not worth the effort?

Interesting thought, not sure if there would be much benefit for using coaxial cables, since the entire interior space is surrounded by a ground shielded case, thus any slight wire movement moves within a grounded enclosure and invokes a tiny shunt parasitic capacitance change. However, the coaxial cable would also move but the internal capacitance would likely change less than the open wire, so might be worth a try.

Best,

[mawyatt]

As a side note, in the SMD test fixtures, when one does the "Open" calibration the calibration should be done with the fixture plunger spacing the same as the expected DUT. The reasoning is that the plunger fringe capacitance should be the same as when the DUT is evaluated, and the internal two wires on the plunger movement side (L-Cur and L-Pot) slightly move and the fringe capacitance to the fixture interior case (ground) slightly changes. Same goes for the calibration "Short", use a Zero Ohm Short of the same spacing as the expected DUT, here we usually use Zero Ohm SMD devices.

I was recently measuring some single digit pF 0402 caps on my CKT5000 LCR meter with Tonghui SMD fixture, and I also realised I needed to space the fixture tips in a consistent manner for the open calibration.
I used a small offcut of 0.030" (0.75mm) thick styrene sheet and got the high precision measurements that I expected for those tiny caps that I subsequently lost... 

I didn't use a 0R SMD part for the short calibration as I thought the imperfections in that part and the extra contact resistances would be more significant than getting the spacing right - but I didn't waste time on testing that theory.   

What we use for the Open Calibration to get the spacing right is just an insert in the lever arm behind the plunger where the spring is between the plunger case and the lever arm. Just a Bourns plastic blue trimmer tool works well, slide back and forth to get the approximate desired DUT spacing. This keeps everything away from the plungers and they then produce the parasitic capacitance that gets normalized during the Open Cal. We also have achieved some good results with tiny SMD small valued capacitors, and get consistent repeatable results with quality C0G/NP0 capacitors using this technique.

Regarding the Short, Cal. Those Zero Ohm DUT imperfections, and the plunger contact variations is what we wanted to normalize during the Short Cal. Our feeling was (and certainly questionable) that these variations and SMD Zero Ohm imperfections were likely to mimic the actual small resistance DUT parasitic resistance under measurement than just doing a Short Cal with the plunger tips touching, of course as always YMMV.

Best,

[mawyatt]

Just did a quick test using our IM3536 LCR meter, this meter can deliver higher DUT test currents and why we prefer it for low impedance measurements.

Using the cheap SMD Fixture measuring a 1206 1 ohm 1% resistor:

DMM6500 using Kelvin Clip leads 990.7738 mohms with 32 uohm SD
IM3536 with fixture direct plunger Short Cal  971.50 mohms
IM3536 with 1206 Zero Ohm Short Cal 981.85 mohms

Also a 1206 2.49 ohm 1% resistor:
DMM6500 2.489757 ohms
IM3536 direct Short Cal 2.46736 ohms
Im3536 using Zero Ohm 1206 Short Cal 2.48366 ohms

Think this shows that we shouldn't rely on the cheap SMD fixtures for low impedance measurements regardless of how the Short Cal is performed, altho in our case the use of a Zero Ohm Shunt seems to help a little.

If you study the fixture design and construction this should be no surprise as the "Kelvin" type connections don't actually make it to the plunger tips where ideally they should be, but are terminated at a long brass headed thin bolt that is screwed into the bottom of the plunger. So from the threaded bolt end to the plunger tip where the DUT makes contact is a single connection and has considerable impedance.

This is one reason we developed the Lever Arm based LCR meter fixture shown above which allows the "Kelvin" connections closer physically and electrically to the DUT, and the PCB test pad has a much larger surface area for the SMD DUT to get pressed against. Also you can create a higher contact pressure with the lever arm adjustment. We anticipate this will provide a better "fixture" for low impedance DUTs and physically larger SMD DUTs with improved contact impedance and area with better repeatability for low Z components. The PCBs went off for fabrication today, so we should have some answers soon.

Edit: As a note, we haven't been able to get good repeatable low impedance measurements with these cheap SMD fixtures and would not recommend considering them for low Z measurements. We've experienced considerable uncertainty with low Z measurements due to the above mentioned shortcomings, however they have provided excellent performance and repeatability with higher Z measurements.

Best,

[ExaLab]

These test fixtures are not created with the aim of measuring very low impedance smd components.

In the Keysight one for example (whose cost exceeds a thousand dollars), the connection for the "sense" return path is made directly on the BNC connectors of the case, thus leaving uncompensated the part of the cable between the BNC and the mobile measuring cylinder. Evidently, their intention is to favor other aspects of the measurement tool...

If you wish to cancel the contact resistance, i.e. make it almost irrelevant by the pressure applied on the component, the only solution is to avoid the direct connection between the Force and Sense signals. If you haven't already done so, I would try to divide the PCB contact area with the component into two isolated sections, one dedicated to the force and the second to the sense. During measurement, the metallization of the DUT will connect them together...

[mawyatt]

Have seen the mentioned Keysight fixture with the plate across the Force and Sense BNCs, never understood the thinking behind this.

Understand the issue of separating the Force and Sense contact points and trying to figure out if splitting the DUT contacts in half would help solve this and having the DUT make the Force and Sense connections, but decided against that for the PCB because of the issue how to do the Open Cal. Even considered using resistors between the Force and Sense that would allow an Open Cal and the DUT would short them out.

For now the cheap Fixtures have served well, but not applicable for low Z measurements, while hopefully the Lever Arm type fixture will perform better for low Z components.

Might be worthwhile to rethink this Force-Sense splitting for the next version, if we ever do another version.

Best,

 

[ExaLab]

If your main goal is to measure low impedance components, the "open" calibration assumes a marginal role and in some cases can be avoided.
Anyway, it is always possible to use a high value resistor to perform the open calibration.
As example, if you intend to characterize resistors with a value lower than 10 ohms and you carry out the open cal using a 10 Mohm resistor, the induced error on the measurement should be less than 1 ppm...

[mawyatt]

That's a good idea just using a much larger Z device that "bridges" the gap.

Curiosity got to me and I got a couple 100 ohm 1206 resistors and placed them in the "teeth" of some Kelvin Clips, this separates the Force and Sense functions by 100 ohms. Did a quick Open Cal on our TH2830 (can't use the IM3536 at this movement, fan makes too much noise and bothers family!!), then normal Short Cal and measured 0.1 and 1 ohm Dale 1% resistors (large aluminum cased type), and a Caddock 10K 0.01%. Then repeated this using a normal Open Cal.

This wasn't anything other than to "see" how the Open Cal responded to the Force and Sense being separated by 100 ohms, and suspected little variation, which is what we got!!

So now we got the "bug" to roll a version with separate Force and Sense lines to the DUT and shunted by a couple resistors. Since our PCBs are in the production "Que" at the fab over the weekend, we might sneak a quickly prepared version of the LCR Fixture PCB with separate Force and Sense DUT PADs. Maybe a version without the on PCB BNCs, where these would be in a box rather than on the PCB (thinking here is better mechanical stability when pushing the Lever Arm). Might even split the DUT PADs uneven to give emphasis to the Force side which handles the current.

Anyway, thanks for the dialog

Best,

[ExaLab]

This wasn't anything other than to "see" how the Open Cal responded to the Force and Sense being separated by 100 ohms, and suspected little variation, which is what we got!!

I have no idea what the input impedance of the sense line might be. If this were ideally infinite, the error that emerges by carrying out the open cal using two 100 ohm bridge resistors would essentially be linked to the capacitance of the measurement cable and should decrease reducing the test frequency.

If the discrepancy you observe is also present in the DC measurements, this means that the instrument's equivalent input resistive component has its weight!

[mawyatt]

This wasn't anything other than to "see" how the Open Cal responded to the Force and Sense being separated by 100 ohms, and suspected little variation, which is what we got!!

I have no idea what the input impedance of the sense line might be. If this were ideally infinite, the error that emerges by carrying out the open cal using two 100 ohm bridge resistors would essentially be linked to the capacitance of the measurement cable and should decrease reducing the test frequency.

If the discrepancy you observe is also present in the DC measurements, this means that the instrument's equivalent input resistive component has its weight!

Exactly, and we got what was expected. Little variation in DC and a small variation at 10 and 100KHz due to capacitive (Kelvin cable) effects. Since the intent of these custom LCR fixture(s) (working on PCB layout of the split Force/Sense DUT Pad version now) is for low impedance devices and the fixture doesn't have the long cable and capacitance of the Kelvin Leads we expect the overall results to be better than our cheap LCR SMD fixture for low impedance SMD devices.

If this behaves as expected, then "tempted" to cannibalize one of the cheap LCR SMD fixtures to get those nice male BNCs with the lever arms (can't find these anywhere) and maybe use the fixture base with the plunger mechanisms removed as the base for the low impedance SMD fixture, or just mount the fixture PCB on top of a metal case with female BNCs and use BNC male to male adapters to interface with the LCR Meter female BNCs.

Anyway, we'll see how this progresses as time permits, as we have a "paying" project with PCBs in fab now which must take precedence.

Best,

[mawyatt]

Update.

We got a change to try the split contact area concept for the Lever Arm type SMD fixture. Still needs some work, as this was just a kludge setup using a split BNC cables for the Force and Sense leads to the LCR meter. Later when time permits we may roll a direct BNC to LCR meter fixture with the split contact area.

Here's some results with low valued 2512 resistors using a zero ohm 2512 for short cal. on the TH2830, and for comparison using a SMD fixture on the IM3536. Number is () is repeated measurement by removing DUT and later reinserting such.

Type                                    TH2830                                       IM3536

2512 1% 10mohm                9.949mohm (9.951)                     -9mohm (6.68) unstable readings
              40mohm                39.840mohm (39.848)                 21.38mohm (31.54) somewhat erratic
              50mohm               49.993mohm (49.994)                  42.09mohm (45.39)
              100mohm              100.317mohm (100.332)             81.52mohm (93.57)   

Getting good repeatable results is much easier, altho still a task, with the Lever Arm Fixture for these low valued resistors.

Here's what things look like and we'll be working with this later when time permits, and likely rolling a direct BNC connecting PCB.

Anyway, just an update, hope some folks find this interesting as we've found getting good repeatable low Z measurements with SMD is not trivial.

Edit: These are DCR measurements. Also no disrespect for the standard SMD fixture, these aren't intended for low Z measurements. If one studies the design the Kelvin type connections are terminated on a long thin brass bolt and screws into the bottom of the brass plated plungers, so everything after the thin brass bolt head is not part of the Kelvin measurement. This includes the plungers, and the plunger to DUT contact surfaces, which can only be compensated for by the Short Cal. Whereas the Lever Arm structure with the split contact area, the DUT contact "ends" make the Kelvin termination, and thus should improve results for low Z components, which apparently works is some cases as we've demonstrated.

BTW, if some folks want to participate we might roll the next PCB with gold contacts, which should improve contact area performance.

Best,

                     

[M Harris]

You fixtures look absolutely identical to the Rohde and Schwarz LCX-Z3 SMD fixture I have... which is US$1700ish. Do you have a link to where to get yours? Would be very interesting to put both on the R&S ZNB VNA and see if there is any difference....

[mawyatt]

The R&S is very likely a rebranded of either a Tonghui (TH26007 or 8 ) and other source SMD fixture. There are many sources available from AliExpress and range in $ from $70~200 with slightly different characteristics. Study the images as they have slightly different configurations and BNC interfaces, but any like below should work. We have two types, but this is the only one we can find we ordered, the other is so old we have no record.

https://www.aliexpress.us/item/3256801507452996.html?spm=a2g0o.order_list.order_list_main.426.758318022GzISR&gatewayAdapt=glo2usa4itemAdapt

You may need to check inside the case on the AliExpress types, one of ours the thin brass bolt that attaches to the plungers was loose and giving erratic readings.

Best,

[mawyatt]

Ok, this is getting really interesting indeed!!

While waiting for Amazon to deliver we decided to use the Hioki IM3536 with the kludge fixture above, 1m cables included. These cables are just 2 wire with shield cable we had, and have the wires to 2 BNCs and shield to BNC shell. The wires are not heavy just small #22~24 or so.

After some thinking about the Short calibration and attempting to remove the effects of DUT contact area, the method we were using was to use a Zero ohm 2512 SMD as the Zero Ohm DUT, however we have no means to "know" how close this SMD is to 0, and thus just "assumed" it to be 0.

Also remember we are using a split contact concept with the fixture to help deal with the SMD end terminations, and the Force and Sense lines are not brought together by the fixture, the idea is the DUT end terminations make the Force Sense connections on each end of the DUT. This helps reduce the effects of contact to the SMD device by making the DUT perform the Kelvin connection.

Back the unknown Zero ohm SMD, we decided to use a quality 1 milliohm 1% 2512 as the Zero Ohm DUT and do a short Cal with this "knowing" it's 0.001 ohms. Then we just add 0.001 ohms to the readings, or just RTFM and use the reference and scale functions, no time for the manual now tho.

So does this work? You bet, stunningly good we might add, absolutely superb repeatability in the milliohm regions!! Here's some additional 2512 1% low valued resistors measured with the Hioki IM3536, set to Rdc, 16 average, 0.1 ohm range. Repeatability is excellent too!!

Value (milliohms)        Reading (milliohms)       Added for DMM6500 (milliohms)
10                             10.04                                10.026
15                             14.97                                14.868
20                             20.06                                20.016
40                             39.90                                39.916
50                             50.03                                50.065
100                          100.25                              100.242

Later we'll shorten the leads, as they a much too long at over a meter, and maybe spend so time RTFM

BTW this doesn't require a heavy scaled ultra low impedance fixture with massive terminals and wires, just some simple thin cable, few BNCs and a kludge PCB with a $1 Lever Arm used for holding flat items, a 2512 resistor is a flat item...so why not?? Well, we must thank Sir Lord Kelvin again 

Anyway, if you have a quality lab grade LCR meter, might want to give this Lever Arm with Split Contact SMD Fixture a try, doesn't cost much and it just works


We tried to measure these low ohm resistors with the nice Tonghui $200 tweezers, repeatability is hopeless below ~0.1 ohm.


Best,


Edit: added DMM6500 readings above, had to go from center point BNC to single banana on all 4 BNC centers and no ground connections. Another kludge setup but giving good repeatable results!!

[hpw]

We have two types, but this is the only one we can find we ordered, the other is so old we have no record.

https://www.aliexpress.us/item/3256801507452996.html?spm=a2g0o.order_list.order_list_main.426.758318022GzISR&gatewayAdapt=glo2usa4itemAdapt

Well, what model do you really use : Quad BNC to lock or Quad with only 2 to lock.

See pictures... as Ali comes up with two different models.

[mawyatt]

We use both types with 2 and 4 locking BNCs. The one shown below with the 2 lock BNCs we replaced the plastic DUT positioner with a 3D printed version as the supplied plastic positioner didn't do an acceptable job with smaller SMD devices.

There are various types with 2 and 4 locking BNCs with different plastic positioners available and different cases, note the case ends where the tap holding the spring is part of the top lid or a seperate "L" shaped bolted to top lid.

Anyway, these are the two we have, and both types work fine with our TH2830 and IM3536 lab LCR meters, acceptable for our needs with the previous mentioned interior details, modifications, and such.

Best,

[KungFuJosh]

In general with those fixtures, 4 tabs usually denotes a higher quality fixture. 2 tabs is usually generic / lower budget option.

However, some sellers on Ali show the 4 tab version in their ad, then send you the 2 tab version. Somebody tried that with me, and I filed a complaint with Ali and got a partial refund. My SMD fixture cost me less than $20. 😉

[mawyatt]

Here's the final rendition of the Lever Arm SMD fixture. This is mounted on a cheap AliExpress LCR Kelvin Fixture with the Kelvin cables removed, and a small PCB mounted to the top lid as shown below. At first we were going to use coax from the BNCs to the PCB, but decided to just use unshielded wire since this is intended for larger SMD of the low impedance type and not high Z components, nor high frequencies.

The cheap fixture makes a good host for this, and the Lever Type BNCs are nice and why we wanted to mount the PCB on a fixture base with these type BNCs.

Here's some images of this version.

Best,

[mawyatt]

Couple more.

Best,

[hpw]

The R&S is very likely a rebranded of either a Tonghui (TH26007 or 8 ) and other source SMD fixture. There are many sources available from AliExpress and range in $ from $70~200 with slightly different characteristics. Study the images as they have slightly different configurations and BNC interfaces, but any like below should work. We have two types, but this is the only one we can find we ordered, the other is so old we have no record.

https://www.aliexpress.us/item/3256801507452996.html?spm=a2g0o.order_list.order_list_main.426.758318022GzISR&gatewayAdapt=glo2usa4itemAdapt

You may need to check inside the case on the AliExpress types, one of ours the thin brass bolt that attaches to the plungers was loose and giving erratic readings.

So have this one now 

1) The new series have thick painting as it gets
2) one BNC connector do not have any ground
3) the isolation / separation wall even too no grounding
4) only the connection shifting parts 4 screws have ground

Good:

- SMD parts remain now on the plastic parts as before, any artillery books for small parts to read 
- 4 BNC to nonnect

The measurements as using coax cable as from a different fixture/tweezers. The a single cable about 80pF (as equal to RG178, RG316).
A lower cable cap. and may shorter to use.

Measurements (currently without the fixing the grounding issue) using new calibrated my HANTEK 1833C:


1) 10K SMD 805 0.05% is almost at 100Hz/100kHz equal  ... OK

2) 1pF +/- 0.25pF SMD 603 as 8..9pf as TH26009C tweezers ... NOK 

So the question rises, what lowest cap value as in xx pF is possible accurate to read.

Hp

[mawyatt]

That depends on how well you've configured your SMD fixture & setup. You need to make all those ground connections and make sure you have a stable mechanical as well as electrical environment. The case cover, all BNCs, the internal shield all need to be well grounded. Sounds as if you have some prep work ahead on the SMD fixture. We used a large drill bit to "clean" the cover holes for bare metal contact, as well as shield mounts.

Best repeatable results are with direct LCR connection, external coaxial cables add uncertainty, so keep them short, with quality stable cables, and when comparing measurements don't move the cables and/or fixture (as coaxial cables flex the internal capacitance can change). Also the E Fields around the open section of the fixture can affect readings on low cap values, just measure a 1pF 0805 C0G/NP0 and wave your hands around near the fixture top. This will convey how far the fields reach outside the fixture.

To direct answer your question, we have a 0.3pF & 4.7pF 0603 C0G/NP0 that we use with other caps as in-house references. For example, we measured with TH2830 0.3159pF, 0.3174pF (with DC Bias adapter), then another time 0.31614pF and DE-5000 (using BNC adapter) 0.32pF, then 4.6938pF for TH2830 and 4.6942pF with Hioki and on and on scattered throughout the years. All within a few FF over the years, so very good repeatability and instrument agreements IMO.

With all this in mind we have no problem achieving repeatability well below 1pF (actually into FF ranges for small values), as confirmed by our Tonghui TH2830 and Hioki IM3536, even achievable with short cables

Anyway, the key message if one wishes repeatable measurements, then a stable mechanical and electrical environment is paramount

Best,

[mawyatt]

Also should mention, when doing the "Open Cal", use something to space the fixture plungers the same distant apart as they will experience when engaged with the DUT. We just use something that keeps the Lever Arm at the proper spacing, a pencil works!!

Best,

[Kean]

Also should mention, when doing the "Open Cal", use something to space the fixture plungers the same distant apart as they will experience when engaged with the DUT. We just use something that keeps the Lever Arm at the proper spacing, a pencil works!!

I was just about to write this myself.  But a pencil??? 

[KungFuJosh]

So have this one now 

Can you post a photo? How many retainer arms does it have? 2 or 4?

[hpw]

That depends on how well you've configured your SMD fixture & setup. You need to make all those ground connections and make sure you have a stable mechanical as well as electrical environment. ....

Best repeatable results are with direct LCR connection, external coaxial cables add uncertainty, so keep them short, ....

Anyway, the key message if one wishes repeatable measurements, then a stable mechanical and electrical environment is paramount

Yes, did some mechanical work, not at the best and now after open & close calibration things looks any better. Bingo!

Now the height of the two mechanical connections seams to be a bit off, as I had to open those parts.

Attached my copper work, so all grounding should now covered in once

Yes, the cables could be shorten, as not RG178 = silver conductor, current multi strand copper.

Thank you for your help

hp