LCR Quadtech/Genrad/IET Digibridge 1689 - use different calibration resistors

[rodpp]

Hi,

IET sells a kit of 4 resistors plus a open and a short (https://www.ietlabs.com/1689-9604-digibridge-calibration-kit.html) to calibrate the Digibridge 1689 (https://www.ietlabs.com/1689.html).

The resistor values are 95.3K, 5.97K, 374R and 24.9R.

If I could find a kit for a reasonable price it would be great, but I'm afraid it will not happen.

Standard resistors can be find, not cheap but not like this kit price. The problem is the resistor values of the kit are not common, and it is easier to find like 1R, 10R, 100R, 1K, etc.

According with the 1689 cal procedure, it asks the resistor value to adjust it's internal constants. How bad it will be if I use 100K, 10K, 300R, 10R instead of the kit values 95.3K, 5.97K, 374R and 24.9R?

Regards,
Rodrigo.

[Vgkid]

I would only use the required values , though the 100k(possibly 100ohm) might be ok to use. The rest are running close to the min range values.  Since it runs ratiometrically as long as you are within a few percent , you should be good.

[TimFox]

Over the years, I have purchased several General Radio decade resistor boxes that cover this range.
Typical accuracy is 0.01% plus small constant.

[rodpp]

Over the years, I have purchased several General Radio decade resistor boxes that cover this range.
Typical accuracy is 0.01% plus small constant.

I have a Yokogawa 279301 good from 0.1R to 1K1. And a IET Labs 1433-Q from 0.1K to 1M1. So these two boxes cover the kit range.

I can measure the resistances using my Keithley 2002, not with a accuracy of 10ppm from the standard as the Kit, but good enough.

The problem is to measure the Q values. I have a HP 4342A Q Meter, but it frequency range is from 22KHz to 70MHz and Q range from 5 to 1000, with accuracy around 10%. The kit has its Q value measured at 1KHz with accuracy of 25ppm.

Having standard resistors intead of decades, I could sent it to calibrate and use the measured values.

[manganin]

IET sells a kit of 4 resistors plus a open and a short (https://www.ietlabs.com/1689-9604-digibridge-calibration-kit.html) to calibrate the Digibridge 1689 (https://www.ietlabs.com/1689.html).

The resistor values are 95.3K, 5.97K, 374R and 24.9R.

If I could find a kit for a reasonable price it would be great, but I'm afraid it will not happen.

I guess the odd value resistors are the same as the reference resistors in the inverting amplifier feedback loop (unity gain).

The high price "calibration standards" are most likely the exact same type as the feedback resistors.

So copy those and the approximate wire lenghts on the main PCB.


These instruments seem to keep the calibration forever. Very unlikely that any adjustment of the constants is required.

[rodpp]

IET sells a kit of 4 resistors plus a open and a short (https://www.ietlabs.com/1689-9604-digibridge-calibration-kit.html) to calibrate the Digibridge 1689 (https://www.ietlabs.com/1689.html).

The resistor values are 95.3K, 5.97K, 374R and 24.9R.

If I could find a kit for a reasonable price it would be great, but I'm afraid it will not happen.

I guess the odd value resistors are the same as the reference resistors in the inverting amplifier feedback loop (unity gain).

The high price "calibration standards" are most likely the exact same type as the feedback resistors.

So copy those and the approximate wire lenghts on the main PCB.


These instruments seem to keep the calibration forever. Very unlikely that any adjustment of the constants is required.

Your guess is correct, these are the values of the reference resistors of each range (R96-R99).

About keeping the calibration forever, the problem is that the calibration constants are stored in RAM. The battery was dead and the cal constants lost.

[timeandfrequency]

[...]
About keeping the calibration forever, the problem is that the calibration constants are stored in RAM. The battery was dead and the cal constants lost.

... and after the resistances and Q values are known, there are also the K factors for each range 
(see attached pictures)

[rodpp]

[...]
About keeping the calibration forever, the problem is that the calibration constants are stored in RAM. The battery was dead and the cal constants lost.

... and after the resistances and Q values are known, there are also the K factors for each range 
(see attached pictures)

... and before that all, there is a Frequency Calibration easily done using a universal counter.

So the procedure is:

1- Frequency Calibration adjustment (one measurement of the 1KHz signal using a universal counter);
2- Reset Frequency Correction K Factor;
3- Calibrate (adjustment) all ranges with the standard resistors (open, short, plus 4 measurements, one for each range);
4- Calibrate (adjustment) K Factor (8 measurements, two for each range at 1KHz/20Khz on range 1 and at 1KHz/100KHz for the others ranges);
5- Calibrate (adjustment) all ranges again with the standard resistors (open, short, plus 4 measurements, one for each range).

It is indeed a relatively simple and fast procedure.

It seems complicated, but having the Calibration Kit and a universal counter it can be done fast.

[Electrole]

The actual value of each calibration resistor is entered into the 1689 during calibration, so there's no need to target the exact values that have been specified.

For the calibration of a 1689 I made a kit with Vishay S102C and Vishay HZ resistors, both with 0.01 % tolerance, as follows:
24.9 Ω replacement:  50 Ω // 50 Ω = 25 Ω
374 Ω replacement: 750 Ω // 750 Ω = 375 Ω
5.97 kΩ replacement: 5 kΩ + 1 kΩ = 6 kΩ
95.3 kΩ replacement: 100 kΩ

[rodpp]

The actual value of each calibration resistor is entered into the 1689 during calibration, so there's no need to target the exact values that have been specified.

For the calibration of a 1689 I made a kit with Vishay S102C and Vishay HZ resistors, both with 0.01 % tolerance, as follows:
24.9 Ω replacement:  50 Ω // 50 Ω = 25 Ω
374 Ω replacement: 750 Ω // 750 Ω = 375 Ω
5.97 kΩ replacement: 5 kΩ + 1 kΩ = 6 kΩ
95.3 kΩ replacement: 100 kΩ

Nice, I think it is better than using a decade box. Because one can send the resistors to calibrate and have the values of R and Q at 1KHz measured with good accuracy.

Did you measured the R and Q at 1KHz "in house" or used an external Cal Lab?

For the R I can use a precision multimeter (DC, not at 1KHz). But I can't measure Q.


The manual specifies R with accuracy of 20ppm and Q 25ppm.

[Electrole]

It's a while back, but I measured the response with a network analyzer, and converted the results to obtain the Q.
I no longer have the 1689m, and I have sort of forgotten the details.
The resistive part was measured with a 3458A multimeter under calibration.

[Conrad Hoffman]

The Q of a resistor at 1 kHz is going to be pretty low. How do you measure it with any great accuracy? The old GR 1608 bridge will do it, with a claimed accuracy of 2% and another term, but I remember it as being a bit of a fussy thing.

[rodpp]

In the Digibridge 1689 service manual, the examples of values of Q of the resistors are:

24R9 -> Q= 15ppm
374R -> Q= 5ppm
5K97 -> Q= 0.22ppm
95K3 -> Q= -280ppm

Really pretty low!

[timeandfrequency]

In the Digibridge 1689 service manual, the examples of values of Q of the resistors are:

24R9 -> Q= 15ppm
374R -> Q= 5ppm
5K97 -> Q= 0.22ppm
95K3 -> Q= -280ppm

Really pretty low!

Isn't that somehow strange that the Q factor is expressed in ppm (or percentage)?
Ppm or percentage of what, actually ? Of the nominal resistor value ?

For a dimensionless parameter, I would expect the 'quality factor' Q values given as positive numbers or power of 10, like 5 E-3.
Are the frequencies provided for which these Q values are expected ?
Why do we have negative values like '-280ppm' ? Does a negative Q value mean there's a capacitive burden ?


And asking, as stated above in the manual, for a precision of 25 ppm for a Q value seems really demanding.
Anyone knows a VNA or bridge or TE that mesures Q faithfully with a 0.0025 % worst case precision ?

[rodpp]

Isn't that somehow strange that the Q factor is expressed in ppm (or percentage)?
Ppm or percentage of what, actually ? Of the nominal resistor value ?

For a dimensionless parameter, I would expect the 'quality factor' Q values given as positive numbers or power of 10, like 5 E-3.
Are the frequencies provided for which these Q values are expected ?
Why do we have negative values like '-280ppm' ? Does a negative Q value mean there's a capacitive burden ?


And asking, as stated above in the manual, for a precision of 25 ppm for a Q value seems really demanding.
Anyone knows a VNA or bridge or TE that mesures Q faithfully with a 0.0025 % worst case precision ?

No, it isn't strange. The resistor Q factor is a relation of the reactance (inductive - X_L, capacitive - X_C or both) and resistance (Q = X/R). As the reatance of a resistor at 1KHz is near zero, the Q value is very low. From the above values for example, the resistor with R = 375R and Q = 5ppm means a X = 1.875mohm. It could be expressed as 0.000005 also, but 5 ppm is more convenient.

The negative value of Q means capacitive reactance.

[manganin]

1.  Find resistors of the same type and dimensions as the reference resistors inside the instrument. GenRad used those in the calibration resistors too.

2.  Copy the original 4-wire calibration resistor PCB     or     use bare components with the wire length similar to PCB mounting.

3.  Measure the actual resistances.

4.  Use the Service Manual example Q values, which are very close.

5.  Profit!

[zrq]

I think the best way to get the Q values is measure the parasitics at a higher frequency (with a network analyzer or another LCR bridge), then extrapolate back to 1 kHz.
And in many cases, if you use small SMD resistors, the parasitics may be small enough to be neglected. One can refer to this paper: https://arxiv.org/abs/2407.17805

[timeandfrequency]

Thanks @rodpp, 5 ppm stands here just for 5E-6 , and yes this is a very low Q value.


a) It is still interesting to notice that the required Q precision is asked at 25 ppm (as stated by the manual) and that at the same time, Genrad provides examples of values that are 100 X lower (5K97 -> Q= 0.22ppm).



b) I had a look to what kind of TE could be used to calibrate your reference resistors.
The precision E4980A LCR meter (successor of the HP4284a) seems to be unable to measure such low Q values.
The datasheet (see page 17, tables 23 to 25) tells that the TEs 'phase angle of impedance θ' is calibrated with a precision of 1E-4 radians at 1kHz for impedance ranges from 0.1 to 1 k Ohms. However, the displayed mesurement range starts much lower at ±1 arad (page 7).
If we take the previous example (R =375 Ohms, X = 1.875 mOhms, Q = 5 ppm) the angle of impedance θ is 5 µrad, so quite 2 orders or magnitude lower than the precision of calibration of the E4980A.
I don't know if there's a way to make a kind of relative measurement to take advantage of the very small display resolution provided by the E4980A.



c) I would also agree with @zrq when he suggests to raise up the measurement frequency. This is also mentioned in the article he linked to (see page 6).

1.  Find resistors of the same type and dimensions as the reference resistors inside the instrument. GenRad used those in the calibration resistors too.

d) Of course, gathering the genuine feedback resistors of a defective unit would be the best choice. As the 1689 seems to accept different cal resistors values than those in the genuine cal kit, a less than perfect way could be as follows :

- Design the PCB that fit into the standard Digibridge connector (4 wires + guard ?). Connexion areas have to be gold plated for best long term corrosion resitance
- Buy the best precision resistors you can find for an acceptable price
- Meet all of the soldering requirements (reflow curve for SMD); Use leaded solder with 4% Silver for through hole parts
- Assemble your own cal kit (4 standard resistors + O + S). Additionnal or duplicates values can serve as 'verification kit' for extended confidence
- Optionnaly build a 3D printed bracket + brass hand screws that holds the boards vertically
- Optionnaly add some potting around the resistors and PCB to avoid any moisture ingress (before potting, 'bake' the devices in your kitchen kiln for 24 hours at 50°C) (*)

@Electrole gave you very nice resistor combinations to build your cal kit. Here are some others (not so close to the genuine values).

I selected the best precision resistors from Mouser (Tolerance <= 0,01 % ; Tempco <= 2 ppm/°C) so that you gain simultaneously a reasonable long term stability feature (ideally not worse than ± 0.01 %/year). The manufacturers provide reliable long term stability data only for (some) high precision resistor series.

https://www.mouser.fr/c/passive-components/resistors/?resistance=20%20Ohms%7C~350%20Ohms~~400%20Ohms%7C~6.8%20kOhms%7C~9%20kOhms%7C~20%20kOhms%7C~25%20kOhms%7C~68%20kOhms~~100%20kOhms%7C~1%20MOhms&temperature%20coefficient=0.2%20PPM%20%2F%20C~~2%20PPM%20%2F%20C&tolerance=0.005%20%25~~0.01%20%25&instock=y&rp=passive-components%2Fresistors%7C~Temperature%20Coefficient%7C~Resistance%7C~Tolerance

Having 0.01 % resistors is nice but it's still far from the 20 ppm precision asked by the SM. So a measurement of your DIY cal kit with an 'in specs' 8.5 digits Ohmmeter is probably needed.

To replace the 95.3 kOhms standard, use 68 k + 25 k = 93 k  (100 k // 1 M = 90.91 k might also be interesting if 100 k is 0.005 %)  https://www.resistance-calculator.com/advanced-resistor-calculator
To replace the 5,97 kOhms standard, use 6.8 k (or 9 k // 20 k = 6.21 k)
To replace the 374 Ohms standard, use 400 Ohms 0.005 %  (350 Ohms 0.01 % is a bit cheaper)
To replace the 24.9 Ohms standard, use 20 Ohms

Have also a look at other distributors (Digikey, Newark, RS) if they source other values or better tolerances.

If the parasitic inductance value is provided in the datasheet, a ballpark of the impedance at 1 kHz could be calculated.

Keep your cal kit in a ziplock bag. Add some silica gel to avoid any moisture.


For about 300 bucks, you should be in business with a DIY cal kit that is an acceptable replacement of the genuine IET Labs/Genrad item.
Maybe other forum members could be interested in having/buying/building such a cal kit if you publish the Gerber and STL files.

(*) at PTB, they dried the parts with deep vacuum. Personnaly, I would have been a bit reluctant in vacuuming the resistors, as the low pressure could alter their internal structure. But as they did it that way, it's probably safe. At least for SMD parts.

Edit by gnif: contents of attachment inserted.

[rodpp]

Thank you all for the contributions.

As @manganin said, using the Q values from manual examples would be a good choice IF using the very similar resistors as the Digibridge 1689 range resistors.

The article @zrq linked is very interesting, but the graph in Fig. 5 shows a 3000 ppm variation of D between 1KHz and 10KHz. And the article analysed only one resistance value of 12K9, while the reactance normally is more inductive in low resistance values and more capacitive in high resistance values. As the inductive portion of Q increases when the frequency increase and the capacitive portion decreases, the Q x f relation is different for each sample. In some specific ranges of frequency the Q may remains constant, or have very little variation, as the "inductive" Q variation cancels the "capacitive" Q variation.

And thank you @timeandfrequency for the detailed answer, including the options of commercial R values to construct a set of standards similar to the 1689 calibration kit.

As I understand the service manual specification, it says that the Q value should be know with 25ppm of the real value. So that a Q value of 5ppm could be measured from -20ppm to 30ppm, and not 25ppm of the 5ppm (meaning a range from 4.999875ppm to 5.000125ppm).

I have a Wayne Kerr B642 bridge (https://www.casa.co.nz/equipment/GEC/Wayne-Kerr-B642-OpMan-52p.pdf) with a especified accuracy of 0.1%. It measure directly the G (conductance) and C (capacitance) of the DUT, where the C being negative indicates an inductive reactance. And the Q value is very simple to obtain: Q = C reading / G reading (see page 19 of the linked manual).

And when doing this relation considering 0.1% accuracy of R and "C", the Q value can be measured with the service manual specification 0.001 x 0.001 = 0.000001 = 1ppm.

One problem is that frequency of the internal oscilator of this bridge is f=1591.5Hz (w=10e4), and not 1KHz as needed. But it is possible to use an external signal generator from 200Hz to 20KHz.

So my plan now is to use my two resistors decade box, measure the resistance with a Keithley 2002 and the Q with the Wayne Kerr B642.

To test this setup I measured the needed resistance and it's Q values (at f=1.591,5Hz) from the decade boxes twice, to see the measurements variations after changing the decade dials and changing the DUT from K2002 and B642. Here are the results:

FIRST MEASUREMENTS
R1 - decade1 set to 95k3
R = 95K3019
Q = 4,59716 ppm

R2 - decade1 set to 6K0
R = 5K99964
Q = 0,38864 ppm

R3 - decade2 set to 374R
R = 374R077
Q = 5,01873 ppm

R4 - decade2 set to 24R9
R = 24R9679
Q = 0,98609 ppm


SECOND MEASUREMENTS
R1 - decade1 set to 95k3
R = 95K3015
Q = 4,40758 ppm

R2 - decade1 set to 6K0
R = 5K99966
Q = 0,37668 ppm

R3 - decade2 set to 374R
R = 374R077
Q = 5,020419 ppm

R4 - decade2 set to 24R9
R = 24R9693
Q = 1,01269 ppm

It seems good enough. Any comments are very welcome!

[manganin]

Some resistor manufacturers have inductance/capacitance or impedance ratio charts available (or SPICE models in the modern world). Then you are not stucked with the original type.


It doesn't need to be precision, an ordinary good quality metal film is short term stable enough. The resistance is easy to measure before calibration.

Multiple resistors (serial/parallel) makes it very difficult to define the reactive part.

Using a decade box is not a good idea either, much larger and more complex residuals plus noise/shielding/leakage issues.

[rodpp]

Using a decade box is not a good idea either, much larger and more complex residuals plus noise/shielding/leakage issues.

But if I can measure the resistance and Q of the decade box as I mentioned, just before the calibration of the Digibridge 1689, isn't it good enough? 

[manganin]

But if I can measure the resistance and Q of the decade box as I mentioned, just before the calibration of the Digibridge 1689, isn't it good enough?

Using a decade box is not a good idea either, much larger and more complex residuals plus noise/shielding/leakage issues.

[rodpp]

In the measurements I did this issues (noise/shielding/leakage) aren't relevant quantitatively, as the changes from measurement to measurement after changing the dials and cables are very low.

[manganin]

In the measurements I did this issues (noise/shielding/leakage) aren't relevant quantitatively, as the changes from measurement to measurement after changing the dials and cables are very low.

But you can't assume the instruments internal circuitry to be ideal and non-dependent on the outside world. For example the amplifier drive capability, RF susceptibility, common mode rejection have their limits.

It may work just fine but it is always a good idea to keep the calibration resistor and the surroundings as small, simple and well shielded as possible. You need to think about the whole system when measuring and even more when calibrating.

[rodpp]

In the measurements I did this issues (noise/shielding/leakage) aren't relevant quantitatively, as the changes from measurement to measurement after changing the dials and cables are very low.

But you can't assume the instruments internal circuitry to be ideal and non-dependent on the outside world. For example the amplifier drive capability, RF susceptibility, common mode rejection have their limits.

It may work just fine but it is always a good idea to keep the calibration resistor and the surroundings as small, simple and well shielded as possible. You need to think about the whole system when measuring and even more when calibrating.

Yes, I agree.

As my Digibridge is way off, because it lost it's cal constants, I'll calibrate it with the decade box now and look for construct a proper set of standard resistors.

Thank you very much!

[timeandfrequency]

As I understand the service manual specification, it says that the Q value should be know with 25ppm of the real value. So that a Q value of 5ppm could be measured from -20ppm to 30ppm, and not 25ppm of the 5ppm (meaning a range from 4.999875ppm to 5.000125ppm).

It is quite sure that your assumption is correct : 25 ppm is actually the allowed uncertainity on the Q value.
Because asking for a 25 ppm precision on a Q value of 5 ppm compares to a moonshot

Using resistor decades
At first glance, I would also favor @manganin 's point of view.
But on the other side, 1 kHz is still a very low frequency. And the Q values you measured tend to confirm that there's not much reactance.

Generally speaking, your two sets of measured values are really close.
Using the test equipement you mentionned, as far it is calibrated, plainly fullfils the Genrad SM requirement (20 ppm precision for R and 25 ppm max uncertainity for Q)


And many thanks to @gnif (one of the forum administrators) for his quick help, because I could not post a section of my previous message (the forum app raised an error).
He found the reason why my message was rejected and is currently working on a fix.

[rodpp]

I did the calibration of the Digibridge 1689.

First the frequency calibration using a signal generator and an universal counter. The Digibridge stores the constant (6ppm in my case) but the test frequency doesn't change after this "adjustment". It seems that it uses the constant in some compensation calculation, not affecting the electronic circuit.

The frequency correction K factor adjustment doens't work in my unit, probably because it uses an old ROM version. I have two service manuals, and only the newer one has this adjustment (item 5.9.. Maybe in the future I'll update the ROM to a newer version to check this.

And finally I adjusted the ranges using the decade boxes, multimeter and bridge.

Everything seems to be working good.

Thenk you all!

[RolandK]

For calibration, read the 1693 instruction, too. The procedure is ah ... very interesting.

If your 1689 still has the calibration only check your calibration resistors with mean value and highest averaging (measure Q in ppm, write down both values) . If you realize one of the calibration resistors in the meter has drifted out of tolerance, only recalibrate this range with the R-Value from the Ohmmeter and the Q-Value from the check measurement. The Q value does not drift significant, as it is based on the construction of the internal reference, the pcb layout and the like.
Do NOT try to calibrate if avoidable. It will not get better and to reach the same level is very hard work.

Calibration:

FIRST: Start with the frequency calibration.

I used the Q factors from a genrad calibration certificate, until i can measure it better (see Q values below).

i just use 4 bare Vishay Precision resistors, there is no need to have exact the manual values, in fact in the 1689 the reference resistors are used in series, so the sum values are better. I just measured them with my best ohmmeter. Therefore the parts tolerance doesn`t matter. But I use the Vishay because of their very low temperature coefficient. Standard resistors with 100 ppm/°C are not my liking there. So cheaper vishays with e.g 0.1% tolerance are fine.

R= 25 Ohm Q = 11 ppm. Y169025R0000T9L is a four wire shunt resistor upto 8W. For the desktop variant i put a piece of paper between the legs, so each of the 4 legs makes only contact to one electrode, so a perfect Kelvin connection. The inner contacs are bend outwards, so they clip the paper with the 2 outer ones (see picture)
R= 400 Ohm Q = 0 ppm Y0007400R000V0L
R= 6K Q = 2 ppm  in use: 1k + 5k soldered together, so that i can use all 3 values (if i need a 1k or a 5k) Y14531K00000T9L and Y14535K00000V9L
R= 100k Q = -258 ppm Check this value with your lowest D Foil capacitor, eg. a X2 type, should be arround 100 ppm. D of capacitors is never negative.

Of course make all measurements with correct open short calibration. For checking the open short must be done for each frequency extra. (eg the hp 4274a / 4275a does it for all frequencies at once and no special code, just 2 buttons)

I checked it against a genrad 1404 100 pF vacuum capacitor, which has D < 10 ppm according to the article in genrad experimenter Aug 1963 page 8.
The procedure for it was really annoying, as i use different cable sets and had to make open and zero calibration when changing between the 100k and the 100pF. I was tweeking the Q of the 100k resistor until the D of the capacitor was at 10 ppm.
This is useful to measure low D foil capacitors.

On picture: pins of 6 k resistor in 60° positions: 1k - 5k - 6k, all 3 values are usable
25 Ohm kelvin resistor pins bent to paperclip for kelvin connection in desktop 1689
short consists of a piece of 1,5 mm² wire flatened with a vice with flat surface. This is ok for 4 wire connection.