1689 LCR Meter Specifications

[ElizatronicWarfare]

Hi all,

I have just come into possession of an old QuadTech 1689 RLC Digibridge (i.e. LCR meter), but the specs don't seem to line up with my expectations. Is it really twice as good as a HP/Keysight 4284A/E4980A (±0.02% vs ±0.05%)? Is there some specific operating condition required I am not aware of to reach these specs? Are HP/Keysight LCR meters really not that good? I do not have as much experience comparing LCR meters as I do operating them, so I appreciate any context you all can provide.

Regards,
ElizatronicWarfare

[rodpp]

Yes, the Genrad 1689 digibridge is a precision instrument, even used in calibration labs. IET Labs still sell it new for around USD 8-9K.

They have an online tool to calculate its accuracy:
https://www.ietlabs.com/notes/digibridge_accuracy_calculator

Read some available tech notes available here:
https://www.ietlabs.com/notes/Digibridge

This one is about metrology applications of the Digibridge, using the 1689:

https://www.ietlabs.com/pdf/application_notes/035005%20Digibridge%20for%20Metrology%20Applications.pdf

[rodpp]

This Keysight marketing material compares the E4980AL vs the Digibridge 1693:

https://www.keysight.com/us/en/assets/7018-03933/flyers/5991-2314.pdf

The 1693 is very similar to the 1689, but goes to 200KHz while the 1689 stops at 100KHz.

The accuracy graphics shows that above 10KHz the E4980AL is much better than the digibridge. But for low frequency (10Hz-10Kz) the digibridge is better.

Of course, as a Keysight marketing material it shows the specs in the more favourable way to Keysight equipment.

[rodpp]

One interesting thing about the 1689 is the calibration is performed using only a set of four known resistors of specific values: 95.3 kΩ, 5.97 kΩ, 374 Ω and 24.9 Ω. So, if you have a resistance decade box and means to measure these values with high precision (10ppm), you can "easily" calibrate the 1689.


IET Labs sells this set of resistors as a kit, NIST traceable, but expensive:

https://www.ietlabs.com/1689-9604-digibridge-calibration-kit.html

You probably will get more information regarding the 1689 digibridge asking in the Metrology section of the forum.

[zrq]

I adapted the Javascript code of the IET's 1689 and 1693 online accuracy calculator (https://www.ietlabs.com/notes/1693_digibridge_accuracy_calculator) for programmatic evaluation in python, and I plotted the calculated accuracy of the main parameter as a function of frequency and impedance to allow easy comparison with competing models. The SecondaryUnitValue (D/Q) is set to 0.

The accuracy specification of A/K E4980A is attached for reference. It seems 1689/1693 significantly outperforms competitors in the low |Z| region (<100 Ohm) but many other LCR bridges are better for higher |Z| (>10 KOhm). Of course, due to the ancientness of the ADC (actually DAC due to the weird architecture) and DSP in 1689/1693, the frequency range is limited to <100kHz and 200kHz (only a single point) and the noise is likely worse than more modern LCR meters.

[croma641]

It seems that the 1692 is exactly the 1689 with different firmware. Simply replace it and redo the calibration


https://www.eserviceinfo.com/downloadsm/34470/Genrad%20Quadtech_1659%201689%201692%201689M.html

[zrq]

It seems that the 1692 is exactly the 1689 with different firmware. Simply replace it and redo the calibration


https://www.eserviceinfo.com/downloadsm/34470/Genrad%20Quadtech_1659%201689%201692%201689M.html

I think so. My Genrad 1693 mother board have part number of 1689-470201E. If someone is brave enough for trying an upgrade, I can dump the firmware.

[croma641]

Ah, ok, 1692 is the desktop version, 1693  the shelf version.

The only problem for the owner of a 1689 is the calibration.

I believe that the ROM dump (200 KHz) would be very useful to anyone wanting to repair/upgrade a surplus or already compromised 1689.

[Vgkid]

Going from the manual. While 1689 snd 1693 share the same main board. The 1693 has another add on board. That one has a bunch of ALU chips on it. That manual is from the early 90's I think.

[zrq]

I'm not sure about the add on board. I don't seems to find it in my 1693 although I haven't took it fully apart yet (waiting for the repalcement battery). Can you mistaken for the interface board and the "high speed" option?

[zrq]

Here I attached some photos of the main board of my GR 1693 I took during battery replacement. I didn't find any difference with 1689 except the sticker on the ROM.
The board also looks very similar to GR 1659's: https://www.eevblog.com/forum/testgear/quadtech-1659-rlc-digibridge-project/msg5534243/#msg5534243

I didn't dare to take off the ROM chip to make a copy, as it's too close to the calibration RAM and I'm afraid of losing the calibration constants if I bent the board too much and break the contact. Hope in the future someone with a newer make using Dallas BBRAM can make such a copy.

[TimFox]

One interesting thing about the 1689 is the calibration is performed using only a set of four known resistors of specific values: 95.3 kΩ, 5.97 kΩ, 374 Ω and 24.9 Ω. So, if you have a resistance decade box and means to measure these values with high precision (10ppm), you can "easily" calibrate the 1689.


IET Labs sells this set of resistors as a kit, NIST traceable, but expensive:

https://www.ietlabs.com/1689-9604-digibridge-calibration-kit.html

The inherent calibration of the GR Digibridges and later similar LCR meters relies on precision calibrated resistors and an accurate 90 deg phase difference between two AC waveforms at the test frequency.  This includes later low-cost units such as the DE-5000.  Earlier bridges relied on precision reference capacitors.
The electronics measure the voltage across the DUT and the current through it with phase-sensitive detectors driven by the two quadrature reference waveforms.  The precision resistor is used in the current measurement.
The “calibrate” function in these meters (open and short termination) compensates for the parasitic components in the test fixture, while the impedance calibration itself uses the reference resistor.

[zrq]

I noted (likely) several bugs in the firmware of my GR1693 (or some very subtle malfunctions, bit rot?), let alone the general annoyance when using the terribly designed UI and GPIB interface.
The first minor one is sometimes the measurement have a higher noise, although still in spec, the noise can be much higher than the "2 ppm" variance as bragged in the application notes when using the slow mode, it can be in the range of 4 ppm to 7 ppm if it happens. When the instrument boots up, it goes by default to this noisy state, and interestingly when I switch the frequency from the default 1 kHz to 1.22 kHz (or many arbitrary values) and back to 1kHz, the noise gets much closer to the "2 ppm" like 2.8 or 3 ppm. Rebooting the digibridge can consistently reproduce this.
The second major bug is when you set the frequency to 0.234375 kHz (=60 kHz/256), the reading starts to jump around all over the place, while setting the frequency to any nearby values can give perfectly correct and stable values.

[zrq]

Here I attached the IH test voltage waveform for the "0.234375" kHz and the nearby 0.23622 kHz. There is something very wrong with the primitive DDS generation of the sine test signal.

[zrq]

Given how close the 11.719 Hz is precisely 1/20 of the set frequency and 3kHz/256, I suspect that the pre-divider U19 of the test frequency is set incorrectly in this case to 0.96 MHz instead of 3.84 MHz, and the correct division factor of 256 is applied...

Update: It gets a bit mysterious, IET support replied that they cannot reproduce the 0.23437 kHz problem with production 1689 and 1693, and they explictly said they cannot fix my unit given the age..

[zrq]

Testing a 10k VHP102Z versus frequency, resistor mounted on a Tonghui test fixture and resistance expressed as derivation from 10.000kOhm to get the most resolution. The consistency at low frequencies (<10k) are great, while I'm not sure if the derivation at higher frequencies is from the resistor (and its leads) or error of the Digibridge itself.
Also attached the allan variance of the resistor reading (expressed as ppm of 10.000kOhm), before and after sequence of switching to 1.22 kHz and back to 1kHz to avoid the noisy state.

[zrq]

I did some more elaborate characterization of the noise of my 1693 LCR meter measuring a 363 Ohm calibration resistor made of Vishay S102 at different frequencies. Every point is the standard deviation of 100 consequent measurements at slow speed (~ 1Hz). The % and DQ in ppm mode is used to get the best resolution. The accuracy is great, 1kHz reading is within 20 ppm (resolution limited, improve to noise limited <10 ppm if using the % readout) compared to the DC value by my Keithley 2001M and L4411A (which differs from each other by 3.6 ppm). However, the noise is not so good and the minimum is about 2 ppm of reading in the comfort zone of 300 Hz to 3 kHz. The noise greatly increases to 0.02% at 200 kHz. Of course, one should not expect such 80s' technology to compete with new Tonghuis' that can use modern Msps 16bits ADC to get 6 digits of stable reading at a wide range of frequencies.

In addition, I did some characterization of a 50 Ohm load at different frequencies with a improvised four terminal to BNC adapter. The slight difference in the blue is showing the sign flip. This should give a upper bound of the frequency flatness of the reading.

I also attached a crude python driver for the 1659/89/93 with the older GPIB card using pyVISA. It's a bit tricky to initialize the communication as one need to issue control_atn to make the GPIB card a listener. The go to local function is still with some minor catches so use with care (1689/93 have no capability of going back to local though the front panel ).

[zrq]

GR 1693 ROM dump is now available on KO4BB, feel free to try on GR 1689. It contains the 0.234375 kHz bug but I don't know 6502 assembly to dig it out.