Bias Network for LCR Meter

[mawyatt]

Studying the effects of bias voltage on capacitors usually requires an LCR meter with a build-in bias network, or configuring an external network to isolate the DC bias from the LCR meter terminals which has no built-in bias or the LCR bias doesn't have sufficient range.

With a brief break in other work we decided to create an external network to support external DC bias with the bench type LCR meters we have and use. After some careful experimenting we found the Tonghui TH2830 to be receptive to AC coupling the High Current and High Potential terminals, our other lab bench meter wasn't tho!!

With a relatively large range of potential test capacitors, which includes ceramic, electrolytic, tantalum, and polymer, and a large voltage range we decide to investigate the HP16065A (service manual kindly provided by Thomas) External Voltage Bias Fixture. This provided a baseline schematic and methodology to move forward.

Here's our first cut at an external bias fixture to use with a standard bench type LCR meter. The general idea is to insert the fixture between the LCR meter terminals and the actual component testing fixture, wether it be a set of Kelvin clips, tweezers, SMD holder or leaded device fixture/holder. With the standard spacing of the BNC terminals the bias fixture can be directly connect to the LCR terminals with 4 BNC male to male adapters, or with quality BNC cables (either 50 or 75 ohm). The component fixtures attaches to the other side of the bias fixture.

In the schematic the LCR High Current and High Potential terminals are isolated with series capacitors. The design is flexible to allow paralleling film type capacitors for greater capacitance, and the Current terminal also supports high capacitance electrolytic types as shown. The LCR terminal are protected with back to back diodes, back to back Zener diodes, and anti-parallel series Zener diodes with a GP diode. Also the Low Potential is protected with triple back to back diodes and shown. These various diode protection schemes are similar to what HP 16065A has, although they may not be required with certain LCR meters (it's always easier to remove or not install a component than add one!!).

The DC comes thru a couple resistors (10K and 100K) and a shunt external noise reduction capacitor, Diode D19 is optional to prevent reverse bias if a polarized capacitor is used (C9 and/or C11). A sense line is provided, also with an 10K and 100K resistors and shunt capacitor to allow monitoring the DUT voltage. Unipolar overvoltage monitoring is provided by Zener D17 and an LED, the Zener can be selected based upon the desired voltage range of the fixture, and if bipolar operation is required these should not be installed.

Note that the High Current series capacitor(s) are outside the High Potential Sense terminal, which also has series capacitor(s) but is a much higher impedance than the High Current terminal. The general idea is that when a "Short" calibration is performed, the calibration includes the effects of the series capacitors and reduces their measurement effects. A drawback with larger capacitors is the longer time constant to charge the DUT and the fixture capacitors, if one tends to measure larger capacitors, then reducing the 100K charging resistor might be beneficial.

One of the reasons for interest in this fixture is to help understand the subtle voltage effects on certain ceramic capacitors, and later polymer and electrolytic types. Note that the PCB is designed for all sorts of variations to allow customization as required.

Anyway, thought others might find this useful, we plan on constructing a few types with different components for various voltage ranges and capacitors types.

One word of CAUTION, this has NOT been tested on any LCR meter (yet), and could cause LCR meter damage, so please proceed with care and YMMV!!

We'll provide PCB design files and such after we've had time to evaluate the concept and verify the performance. Also likely will design a 3D printed case to hold the PCB.

Comments, suggestions, and critique are welcome.

Best,

[Jay_Diddy_B]

Hi,
HP has 1uF in the Hpot connection and 5.6uF in the Hcur connection:



The HP instruments do open and short calibration.

I believe the 20 Ohm resistors and the switch are to discharge the DUT. The switch is operated by a safety cover.

Jay_Diddy_B

[jonpaul]

Characteristics of capacitors vs bias is well documented since 1960s.

Many app notes and papers from HP, Wayne Kerr and old line capacitor manufacturers.

We used the HP and WK LCR and bridges with built-in or external bias.

Jon

[mawyatt]

@Jay-D,

Yes this is the basis for the design, HP used a 5.6uF series cap in the force line. This limits the lower frequency response and we decided to use a larger effective capacitance by paralleling film capacitors as shown.

One could also use a unipolar capacitor like a Al, Ta or Polymer and some experiments show this also works with the Tonghui TH2830 we have, so we included the ability to use. However, we couldn't get the Hioki IM3536 to work with any of the series capacitive techniques, film or otherwise! This LCR meter must have a different type architecture than the TH2830. For our case this isn't an issue since the area where the DC bias measurements are required is in the lower frequency ranges.

We just did some more elaborate tests with the TH2830 and having a drive series film capacitance of ~13uF and a sense series cap of 1uF allows reliable measurements of up to ~500uF at 50Hz after a short calibration at the DUT terminals. We setup a test where the DUT was first measured with the IM3536 at 50Hz (and other frequencies) then measured with the TH2830 with the added series capacitors mentioned. We had excellent agreement up to ~1KHz with various capacitor types like ceramic disc, film and electrolytics up to 500uF. The wired setup contributed to errors above 1KHz tho, which should be reduced with a proper fixture.

Not sure if all the precaution measures HP took with 16065A are necessary (none were used with the wired setup), but these were included in the PCB since it's easy not to use them!

Anyway, the external DC bias technique with the TH2830 seems to work very well!!!

Note we included the ability to have a much larger force series capacitance than 5.6uF HP utilized, because we anticipated using the IM3536 which goes down to 4Hz and thus requires a larger capacitance. However after discovering the IM3536 doesn't work properly with this external DC bias technique, the additional capacitance isn't required.

Best, 

[mawyatt]

@ Jon,

We are aware of the documentation of capacitors vs bias, we even had a full featured materials lab that did very detailed analysis of all sorts of capacitors, dielectrics, materials and such before retiring, and even detailed characterization of highly toxic BeO that was used in the Electrolytic Suspended Gyro (ESG) technology for electrostatic levitation of a Be hollow ball.

Our needs are different now, without going into fine details we need the characterize various ceramics, capacitor types (film, disc, MLCC, Polymer, Tantalum, and Electrolytic types) and such for a specific use. Things like capacitance vs bias voltage are only one attribute, others are capacitance vs waveform type, polarity, offset, capacitance relaxation vs waveform period and more, all this is to produce a waveform that has very little added distortion/aberrations from the capacitance. The waveforms are completely arbitrary in shape, amplitude (+-100V), offset, and frequency/period.

With the various capacitor types & materials available today, this requires a careful study to understand these characteristics. Hopefully this is all leading to a custom chip design that may employ on and off chip capacitance in a very precise waveform application, where a possible hundreds per chip and total thousand individual precision arbitrary high voltage waveforms are required.

As of now we don't know the detailed effects (short and long term) of these waveform aberrations on the system performance, other than it's very sensitive to certain waveform characteristics like a couple millivolts error in a +-100V waveform!!!

Best,

[jonpaul]

Dear MawWatt, Rebonjour, mille mercis, comprends totalement....

I can highly recommend the 1980s HP/Yokogawa 4195A network /Spectrum Analyzer 0..500 MHz with the fine impedance test set 41951A and appropriate SMD, TH and binding post adapters. 

The test set is rated 100 kHz to 500 MHz, with external bias input 0..40V. 4195A has DC bias sweep 0..40V

The voltage rating is very conservative to stay within the 42.2 V SELV classification.

The HP 4195A used cost can be ~ 2k plus similar for the test fixtures and adapters.

We were very lucky to acquire a mint Analyzer, all test fixtures, adapters and original manuals, from an HP or NITS related in CO.

The Zo is expressed in rectangular or polar and easily scalability or storage.

We use the set-up since 1992 in characteristics of transformers self capacitance.

Bon chance

[Conrad Hoffman]

Thus my love of the old GR bridges like the 716C. The (separate) detectors are AC coupled and it's easy to bias the DUT to hundreds of volts. Inductance, not so much. For that it might be useful to look at the GR Incremental Inductance Bridge and companion 1308-A high voltage/current audio oscillator to see how they did it.

[mawyatt]

Update!

The custom PCB has been received and now awaiting the components. In the meantime we've been able to get the Hioki IM3536 to work with the external high bias voltage concept shown. The High Sense terminal requires a much higher series protection capacitor, fortunately the PCB allows room for this larger cap and we already have some on order. Also the IM3536 mode of operation and other specifics should allow operation down to 4Hz

Best,

[mawyatt]

Components received finally, and we were able to assemble the first bias fixture test PCB this morning. Initial results are very good and everything works as expected! For example, a 0603 test cap measured 4.672pF with the adapter, 4.683pF without on the TH2830, and 4.694pF on the Hioki  

Of course more tests are in order and will be completed when time permits.

Only issue encountered in PCB assembly was we didn't have the 4.7uF film capacitors, so substituted a 10uF for C10. This has a larger footprint with 22mm lead spacing and we had to bend the lends to fit the 15mm spaced holes.

The assembled design values for the fixture are R1 and R3 10K, R4 and R2 90K, C1 and C2 (and C10) 10uF 250V film, C3 and C6 0.1uF film, D8,9,10,13,14,15 !N4004, D11,12,16,17 BAV116, D2,3,6,7 18V Zener. D1, D4, D5 not used, and C4,5,11,12,7,8,9 not used.

Here's the HV Bias Fixture details. Note that this includes many of the protection features from the old HP which may or may not be required. We've done a number of tests and the HV Bias Adapter works very well with the Tonghui TH2830 and Hioki IM3536.

CAUTION: This may cause damage to your LCR meter without proper use and knowledge, if you not familiar with how your LCR meter operates and how the LCR Input protection behaves then you risk potential meter damage. YMMV!!

If folks want more detail I'll be happy to provide such.

Edit: Schematic updated.

Best,

[mawyatt]

Here's some images of the PCB and setup. We still need to cut the Al extrusion case and drill the end plate holes.

Best,

[free_electron]

cool. here is a schematic for a hioki. at first glance it looks identical to hp's

[mawyatt]

One of things we were concerned about is when using Kelvin clips. When the clips fingers are open on the low side then the LPot and LCur are not connected together. This might allow a path for a charged cap to attempt to discharge thru the LPot input before the clips makes the complete contact to LCur. Not knowing if this could damage the LPot input is why the back to back diodes D16 and D17 are present which allows a path to the LCur input. Also the diode string of D8-10 and D13-15 allow a current path back to chassis ground which is where the DC bias input is referenced. Since it takes +-2V for the string to conduct they normally don't matter and only provide the path in case the return currents can't get back to chassis easily.

We also included the ability to use back to back electrolytic caps for cases where one needs to use high currents at low frequencies. Our IM3536 goes down to 4Hz, so the 20uF film caps look like ~2K there, and you can't pump much current to the DUT without having a large voltage drop across the DC Block. With a couple 1000uF back to back electrolytics then this looks like ~80 ohms and larger AC currents can be supported (see C4,5,11,12, the schematic shows 33uF, that was just for the footprint). The EC jumper allows the electrolytics to be engaged/disengaged when needed and you don't have to suffer the higher leakage under normal use. If you only are concerned about unipolar DC bias then D1 can be included, and you don't need back to back electrolytics, and single one is fine.

Anyway, everything seems to work very well!! Just did some quick diode capacitance measurements with reversed and forward bias

Edit: Just noted that D4 and D5 should be moved to the DUT side of the DC Blocking Caps on HCur and HPot, schematic updated above 

Best,

[mawyatt]

Update,

We've probably tested well over hundred devices now and can report that we get good agreement with and without the DC Bias Adapter fixture and between the Tonghui TH2830 and Hioki IM3536 LCR meters. Devices ranged with various types of Aluminum Electrolytics from 10uF to 1000uF, Tantalums from 10uF to 100uF, Polymer from 10uF to 1000uF, chips from 4.7pF to 47uF, film types from 1nF to 10uF, and some measurements down to 4Hz (IM3536).

No ill effects to report, things work as expected and we get reliable results. We also did some DC measurements with diodes to observe the junction capacitance vs voltage behaviors.

When time permits may try and get some software up and running to allow DC sweeps of various components.

Anyway, hope this helps some folks that have LCR meters that don't have built-in DC Bias capability, or limited capability, and/or don't have access to the commercial DC Bias add on fixtures which tend to be expensive, even used!!

Best,

[mawyatt]

Latest Update,

The LCR Meter Bias Adapter aluminum enclosure has been prepared and assembled along with custom 3D printed end plates. Results are quite promising, even at 1MHz and up to 8MHz, which is limit for Hioki IM3536 

Here are some images of the assembly and setup with the IM3536 which is directly connected (with support) with 4 BNC adapters, and to the Tonghui TH2830 via 4 BNC cables. The LCR Meter Bias Adapter is compatible with all our fixtures and Kelvin cables (BNC type) 

Best,

[mawyatt]

Few more images.

Best,

[mawyatt]

Completed some test comparing various capacitance measurements with and without the Bias Adapter.

Here's some capacitance measurements completed on the IM3536 with and without (#) a using SMD fixture at 100KHz.

4.6864pf (4.6826pf), 9.9537pf (9.9498pf), 101.043pf (101.039pf), 993.052pf (993.026pf)

Even at 1MHz. 4.6726pf (4.6814pf), 9.9385pf (9.9464pf)

With a Kelvin Cable at 10KHz.

1.00856uf (1.00934uf), 474.80nf (474.81uf), 2.231nf (2.232nf)

This gave us some confidence and we made some diode capacitance measurements for SD340 (3a SB), 1N5817 (1a SB), 1N4007 and 1N4148. These were made with Kelvin Cable at 1MHz (per data sheets) using the Bias Adapter to bias the diode under test.

Anyway, hope this helps folks that are in need of a DC bias adapter for their LCR meter as the Bias Adapter doesn't seem to degrade the LCR measurement.

Best,

[TimFox]

Have you tried your fixture with a DE-5000 LCR meter?

[mawyatt]

No have not tried it with the DE-5000. Since the chip set in the DE-5000 is based upon a synchronous mod-demod it should work just like with the TH2830 and IM3536.

If we had a spare adapter for the DE-5000 I could make a BNC type to DE-5000 adapter and check it out. Also note on the Bias Adapter PCB design, this has separate small HX style 2 pin connectors at the Hcur, Hpot, Lpot and Lcur Meter side I/O, one could use these with a DE-5000 adapter to work with the DE-5000. Suspect the DE-5000 has a lower voltage threshold on the inputs/outputs, so lower voltage protection zeners should be considered.

If anyone has a spare DE-5000 connector, will be happy to build up a custom cable and check it out with the Bias Adapter.

Edit: One potential issue with the DE-5000 might be during Short calibration, if this requires a DC type galvanic connection between the Hcur and Hpot Inputs then the Bias Adapter won't work because the adapter has a capacitve coupled Hcur and Hpot and won't pass the DC static calibration test. Don't think this DC Short test can be bypassed on the DE-5000 like available on the TH2830 and IM3536 Lab LCR meters.

Best,

[mawyatt]

We've done enough testing using this Bias Adapter with our two LCR meters (IM3536 & TH2830) to consider offering a couple PCBs. The Bias Adapter works quite well indeed, and why we've decided on giving up a couple PCBs

Although a little slow as the series film caps charge and discharge but gives very accurate results almost entirely dependent upon the LCR meter performance, which is what you want and not limited by the adapter.

The Al extruded case with Al end plates provides a very good stable physical and electrical environment and using the "finger test" around the case while measuring a Venkel 1.0pF chip C0G cap at up to 8MHz doesn't affect even at the 1.0009pF level least digit (0.1fF), of course getting a finger within ~75mm of the SMD test fixture where the DUT is held does effect this 0.1fF digit!! So the end DUT test fixture is the limiting factor regarding external influences, not the Bias Adapter, and results don't seem any different than if the DUT fixture was directly attached to the LCR meter terminals. 

Even works down to 4Hz, and has provisions for adding much larger valued electrolytic capacitors (back to back for bipolar bias use) for situations where large value capacitors need to be measured at very low frequencies.

Someone more industrious could "close the loop" with DC bias voltage measurements and reduce the charge time using the LCR and DMM USB or LAN interfaces with some custom developed code. This would be especially beneficial for measuring large capacitors at very low frequency, since the charge time will be long.

We have only a couple PCBs to offer those that would like to build up a DC Bias Adapter for their LCR meter (need to keep a couple for ourselves).

This is a small gesture in return for all the great help some folks here have provided and a small return to the "community" 

Please, don't inquire if you don't intend to build the Bias Adapter, as with only a couple PCBs available this takes away from those that do intend to use the PCBs and benefit from such.

Anyway, send a PM and SASE if interested in the PCBs.

Edit: Added Gerber, Schematic and BOM. Be sure and change Gerber to .zip and BOM to .csv file types. PM for details on building.

Best,

[nmak]

Thank you for sharing your very interesting work.
Have you tried it with HP4284A or Keysight's E4980A?
Which is the maximum bias voltage of this network?

[mawyatt]

We only have access to the Tonghui TH2830, Hioki IM3536 Lab type LCR Meters, so this is all we've tested with. However, would think this should work with other Lab type LCR meters unless they have some limitation during Short Calibration (DC cal must be bypassed).

The maximum bias voltage would be limited by the coupling capacitors utilized and the protection diodes used. Of course the LCR meters also have a maximum voltage they can accept at the Inputs/Outputs.

Best,

[Smokey]

Not sure how I missed this.  I'm totally going to build one of these. 
Thanks!

[mawyatt]

This works very well with our TH2830 and IM3536 Lab Grade LCR meters, however it does not with the DE-5000, note below.

Edit: One potential issue with the DE-5000 might be during Short calibration, if this requires a DC type galvanic connection between the Hcur and Hpot Inputs then the Bias Adapter won't work because the adapter has a capacitve coupled Hcur and Hpot and won't pass the DC static calibration test. Don't think this DC Short test can be bypassed on the DE-5000 like available on the TH2830 and IM3536 Lab LCR meters.


Be sure to enclose the PCB in a rigid metal case (as shown), for mechanical and electrical stability. We have no problem with sub picofarad repeatability using the Bias adapter with SMD fixtures.

Best,

[Smokey]

Interesting.  I have a MS5308, which from what I understand is the same chipset as the DE5000. 

Is that issue with calibration something that someone confirmed on an actual unit with your bias fixture or is that still speculation? 

Assuming the meter does in fact fail calibration because of the short test, what sort of bad things would happen if that cap was just held shorted for that test then opened back up afterwards?

[mawyatt]

It was tested on our DE-5000, so not speculating.

If one could figure out how to bypass the DC short calibration, then suspect it should work fine.

Would think you could use the bias fixture without a short calibration, but not sure what effect it would have on the results tho. Also not sure what level of DC bias a DE-5000 can accept, the lab instruments we use have a relative high rating which is what the protection diodes in the DC Bias Fixture are rated for. Since the DE-5000 has effective +-4.5 volt internal rails derived from a 9V battery, would think anything above +-4.5 on the inputs might be a problem and protection diodes should be adjusted accordingly.

Don't know anything about the MS5308, so might be worth a try to see if you can bypass the short calibration without a DC short, or if the software has the ability to bypass this like the lab grade instruments do.

Best,