GenRad Decade Inductance, Capacitance, and Resistance Boxes

[trobbins]

Absolute accuracy vs. 0 is spec'd in the paragraph above the incremental accuracy.

Ta, I see that now.  Did you do the % calcs on your measurements?  Perhaps the first 100k step has drifted a bit.

[mawyatt]

Imho, the hassle with primary standard caps (I have a GR1404-A 1nF) is the repeatability of the connection interface to the measurement instrument being checked.  My goto LCR is a MCP BR2822 with kelvin clips, which at 1kHz, direct to inner terminals of 1404 measures 1009.6 +/-0.3pF, 0.0002 +/-0.1 D, and rising to 1012pF with LCR meter away from 1404.  To use the 1404 as a periodic check of the LCR meter, I could continue to just do kelvin clips to the inner part of the GR874 sockets, but that seems prone to a variation similar to the 0.3% basic tolerance of the meter.  I'd likely have to set up a custom 4-terminal breakout from the LCR front panel to a set of GR874 locking adaptors.  Given the construction quality of the GR1404, and that it has had one cal indicating +11ppm +/-20ppm back in 1989, but with no initial factory cert, and a spec'd drift of below 20ppm/yr, I'd just be hoping it was still within say 100ppm (0.01%).

Exactly, and this becomes more uncertain as the test frequency increases as the uncontrolled and unguarded parasitics have more influence on the readings, and why we developed the Direct LCR meter BNC connection which has shown excellent repeatability with good reference mica caps even at higher frequencies. The electrical and mechanical environment the DUT "sees" is totally controlled, highly repeatable and guarded, with the only drawback the thermal coupling from the LCR meter 4 BNC connectors to the DUT fixture. We included a thermistor attached to the old Soviet mica caps (these are in hermetic sealed ~30mm square metal tubs) in those fixtures and measured ~2C temp rise with about 15~30 minute time constant when attached to our bench LCR meters (TH2830 and IM3536). Would expect other quality bench LCR meters to show a similar slight temperature rise characteristic with these fixtures.

Another benefit of these Direct BNC Connect fixtures is the ability to create a Short and Open Calibration "Standard" (similar to VNA "Standards"), where the Open and Short fixtures represent a very similar mechanical and electrical environment as the DUT "sees" within the fixture, thus better helping to "compensate" for the parasitics involved in the subsequent measurements.

Anyway, this thread is about resistive decade boxes and not capacitors so we should probably focus on the topic Josh started, and maybe carry on over here if anyone is interested.

https://www.eevblog.com/forum/projects/direct-lcr-meter-connect-test-subjects/

Best

[KungFuJosh]

Today I got a GenRad 1419-K "Silvered Mica" Decade Capacitor.

This bastard is huge (over 14" wide, and nearly 12lbs).

It's also pretty darn accurate. It's rated as +/- 0.5%, but it appears to be much better. Not too shabby.

Thanks,
Josh

[TimFox]

In the good old days, General Radio sold mica and polystyrene standard capacitors and decade boxes.
Their catalog said the mica units had lower tempco, but showed an increase in capacitance at low frequencies.
The polystyrene units were flat at low frequencies, but had a higher tempco.

[KungFuJosh]

Yes, they mention that in the docs I attached. The Mica boxes are considered their higher accuracy versions (1% for poly, 0.5% for mica).

The datasheet specifies the accuracy is the difference between the zero value, and the stated value at 1kHz. So, for this box, accuracy is 0.13%. Still 0.14% when not correcting for the zero value. Not bad at all.

ETA: I'm guessing both of these boxes are around 60 years old. So,  definitely the good ol' days.

[TheDefpom]

Are those my BNC levers ?

[TheDefpom]

Today I got a GenRad 1419-K "Silvered Mica" Decade Capacitor.

This bastard is huge (over 14" wide, and nearly 12lbs).

It's also pretty darn accurate. It's rated as +/- 0.5%, but it appears to be much better. Not too shabby.

Thanks,
Josh

Hey, they look exactly like one of the capacitor standards I use, a 200pF unit, I wonder if they were stolen from something like that and sold on eBay, or if Genrad just made use of them in that unit.

[mawyatt]

Are those my BNC levers ?

No, those are ones we created in Fusion 360 from measurements that Tonghui uses on their LCR fixtures like the TH26048A and Kelvin clips TH26011CS. They are also similar to the ones on the AliX SMD LCR fixtures we have.

We had to print about 5 versions to get the proper snug fit dialed in to the BNCs, and found the panel male BNCs are not all the same outside diameter, not to mention the variation in shrinkage of the different types and brands of 3D filament!!

Best

[mawyatt]

Yes, they mention that in the docs I attached. The Mica boxes are considered their higher accuracy versions (1% for poly, 0.5% for mica).

The datasheet specifies the accuracy is the difference between the zero value, and the stated value at 1kHz. So, for this box, accuracy is 0.13%. Still 0.14% when not correcting for the zero value. Not bad at all.

ETA: I'm guessing both of these boxes are around 60 years old. So,  definitely the good ol' days.

Nice capacitor array! Looks like they are in Bakelite molded cases?

Those micas are huge much bigger than the old Soviet ones that a hermetically sealed in the ~1" square metal tubs!!

Wonder how stable they are? The old Soviet micas we have are quite stable and repeatable, usually out to ~6 digits in the direct BNC connect fixtures, which questions the LCR meter repeatability!!

Best

[KungFuJosh]

Hey, they look exactly like one of the capacitor standards I use, a 200pF unit, I wonder if they were stolen from something like that and sold on eBay, or if Genrad just made use of them in that unit.

Got any photos?

[KungFuJosh]

Nice capacitor array! Looks like they are in Bakelite molded cases?

Those micas are huge much bigger than the old Soviet ones that a hermetically sealed in the ~1" square metal tubs!!

Wonder how stable they are? The old Soviet micas we have are quite stable and repeatable, usually out to ~6 digits in the direct BNC connect fixtures, which questions the LCR meter repeatability!!

Thanks! It's even better than I expected. I don't know if the mica units are hermetically sealed, but the sheet says the polystyrene ones are.

The 1419-K model I have is claimed to be the better stability version. What kind of stability tests are you interested in? Tell me what you'd like to see and I'll setup a test. You know this stuff entertains me.

[mawyatt]

Nice capacitor array! Looks like they are in Bakelite molded cases?

Those micas are huge much bigger than the old Soviet ones that a hermetically sealed in the ~1" square metal tubs!!

Wonder how stable they are? The old Soviet micas we have are quite stable and repeatable, usually out to ~6 digits in the direct BNC connect fixtures, which questions the LCR meter repeatability!!

Thanks! It's even better than I expected. I don't know if the mica units are hermetically sealed, but the sheet says the polystyrene ones are.

The 1419-K model I have is claimed to be the better stability version. What kind of stability tests are you interested in? Tell me what you'd like to see and I'll setup a test. You know this stuff entertains me.

The Micas we have range from 2pF to 10nF, one the best is the 10nF Soviet can mica, which varies less than 0.1pF from day to day on both the TH2830 and IM3536. The other Soviet 4.4nF also varies less than 0.1pF (typically less than 0.05pF) and the Soviet 312pF varies less than 2fF. These are the old hermetic sealed micas, some of the old molded micas are almost as good. We found the molded types more stable than the dipped type, so that's what we selected for the various cap ranges (2pF, 10pF, 100pF, 1nF).

Would think these old GR Micas should be as good, since they are well aged!! Only issue is the banana connections and cables to the LCR meter as mentioned earlier, this creates another level of uncertainty between measurements.

Best

[KungFuJosh]

The Micas we have range from 2pF to 10nF, one the best is the 10nF Soviet can mica, which varies less than 0.1pF from day to day on both the TH2830 and IM3536. The other Soviet 4.4nF also varies less than 0.1pF (typically less than 0.05pF) and the Soviet 312pF varies less than 2fF. These are the old hermetic sealed micas, some of the old molded micas are almost as good. We found the molded types more stable than the dipped type, so that's what we selected for the various cap ranges (2pF, 10pF, 100pF, 1nF).

Would think these old GR Micas should be as good, since they are well aged!! Only issue is the banana connections and cables to the LCR meter as mentioned earlier, this creates another level of uncertainty between measurements.

Day to day stability testing? Temperature changes should be enough to simulate that along with removing & reattaching the clips. As you mentioned, these are well aged, so I don't expect to see any significant drift.

I didn't find any issue with connecting to the terminals, but I will test them more...aggressively. In their manual, they suggest using kelvin probes connected to the wire holes at the bottom of the terminals, and a guard wire connected to the case ground for 3 (4) wire measurements.

I tried as they described on the decade resistor, and also lazily attaching them to the top of the banana port openings, and didn't see any notable difference. The decade capacitor was measured the lazy way. Partly because it was easier, but also because it seems more consistent for a good connection than trying to get the probe clips through the smaller hole at the base.

I'll do some "screwing around with the connections" tests, and some frequency sweeps.

Thanks,
Josh

[mawyatt]

Day to day stability testing? Temperature changes should be enough to simulate that along with removing & reattaching the clips. As you mentioned, these are well aged, so I don't expect to see any significant drift.

I didn't find any issue with connecting to the terminals, but I will test them more...aggressively. In their manual, they suggest using kelvin probes connected to the wire holes at the bottom of the terminals, and a guard wire connected to the case ground for 3 (4) wire measurements.

I tried as they described on the decade resistor, and also lazily attaching them to the top of the banana port openings, and didn't see any notable difference. The decade capacitor was measured the lazy way. Partly because it was easier, but also because it seems more consistent for a good connection than trying to get the probe clips through the smaller hole at the base.

I'll do some "screwing around with the connections" tests, and some frequency sweeps.

Thanks,
Josh

Using the Kelvin connections is probably the best with those banana terminals.

BTW the Kelvin Connections should not be 4-wire but actually 5-wire as the Guard terminal from the LCR meter should be attached to the Guard (case) terminal on the DUT box. Don't use the shorting bar on the decade box, use the case terminal connected separately from the Kelvin Clips to Guard on your LCR meter. Also keep this Guard cable length short, as the parasitic displacement currents (capacitive) around the interior and exterior of the box flow thru this connection and this isn't DC but AC!!

Best

[zepto]

Day to day stability testing? Temperature changes should be enough to simulate that along with removing & reattaching the clips. As you mentioned, these are well aged, so I don't expect to see any significant drift.

I didn't find any issue with connecting to the terminals, but I will test them more...aggressively. In their manual, they suggest using kelvin probes connected to the wire holes at the bottom of the terminals, and a guard wire connected to the case ground for 3 (4) wire measurements.

I tried as they described on the decade resistor, and also lazily attaching them to the top of the banana port openings, and didn't see any notable difference. The decade capacitor was measured the lazy way. Partly because it was easier, but also because it seems more consistent for a good connection than trying to get the probe clips through the smaller hole at the base.

I'll do some "screwing around with the connections" tests, and some frequency sweeps.

Thanks,
Josh

Using the Kelvin connections is probably the best with those banana terminals.

BTW the Kelvin Connections should not be 4-wire but actually 5-wire as the Guard terminal from the LCR meter should be attached to the Guard (case) terminal on the DUT box. Don't use the shorting bar on the decade box, use the case terminal connected separately from the Kelvin Clips to Guard on your LCR meter. Also keep this Guard cable length short, as the parasitic displacement currents (capacitive) around the interior and exterior of the box flow thru this connection and this isn't DC but AC!!

Best

I learned about this when trying to measure a capacitor standard with kelvin clips. The difference with/without guard was 7% (1070 pF -> 1000 pF)! I don't think it will make as big of a difference on decade capacitor boxes.

[KungFuJosh]

Using the Kelvin connections is probably the best with those banana terminals.

BTW the Kelvin Connections should not be 4-wire but actually 5-wire as the Guard terminal from the LCR meter should be attached to the Guard (case) terminal on the DUT box. Don't use the shorting bar on the decade box, use the case terminal connected separately from the Kelvin Clips to Guard on your LCR meter. Also keep this Guard cable length short, as the parasitic displacement currents (capacitive) around the interior and exterior of the box flow thru this connection and this isn't DC but AC!!

Exactly. The two connection methods I was referring to are attached. I meant 4-wire + guard, but we can also call it 5-wire. They're trying to confuse as anyway, calling kelvin + guard 3-wire.

Thanks,
Josh

[mawyatt]

Exactly. The two connection methods I was referring to are attached. I meant 4-wire + guard, but we can also call it 5-wire. They're trying to confuse as anyway, calling kelvin + guard 3-wire.

Thanks,
Josh

That's seems very high for Rs (0.127Ω) for a Reference Cap of 1uF @ 1KHz!!

Best

[zepto]

Yes, they mention that in the docs I attached. The Mica boxes are considered their higher accuracy versions (1% for poly, 0.5% for mica).

The datasheet specifies the accuracy is the difference between the zero value, and the stated value at 1kHz. So, for this box, accuracy is 0.13%. Still 0.14% when not correcting for the zero value. Not bad at all.

ETA: I'm guessing both of these boxes are around 60 years old. So,  definitely the good ol' days.

The mica had better accuracy, but the polystyrene ones had lower DF and DA.

[KungFuJosh]

That's seems very high for Rs (0.127Ω) for a Reference Cap of 1uF @ 1KHz!!

Maybe? But keep in mind it's a decade switch with 4 capacitors together to get that value.

You've got 4 capacitors + switch + connectors in that measurement. D factor is probably better to look at, but still need to consider the network being measured.

[mawyatt]

That's seems very high for Rs (0.127Ω) for a Reference Cap of 1uF @ 1KHz!!

Maybe? But keep in mind it's a decade switch with 4 capacitors together to get that value.

You've got 4 capacitors + switch + connectors in that measurement. D factor is probably better to look at, but still need to consider the network being measured.

Yeah, that's a lot of stuff to get to the actual DUT capacitor, and why we prefer the direct connect method rather than a decade box with banana connectors, switches, wires and such. It's very difficult to "null" all this out, whereas the direct method allows a very good and repeatable Short and Open Calibration which closely represents the DUT conditions.

We thought about getting a Decade Cap box awhile back, but decided against such for the reasons mentioned. Altho using it for a "switchable" cap too fine tune circuits is quite useful.

Anyway, you might get a better result (lower Rs) by directly Kelvin Connecting to the DUT cap inside the box.

Best

[zepto]

You could try cycling all the switches a few times and see if the value changes at all.

[mawyatt]

BTW a good quality Polypropylene 1uF can get down to 10mΩ or less Rs at 1KHz. These of course aren't nearly as stable as quality micas and have a higher TC.

Best

[KungFuJosh]

That's seems very high for Rs (0.127Ω) for a Reference Cap of 1uF @ 1KHz!!

Maybe? But keep in mind it's a decade switch with 4 capacitors together to get that value.

You've got 4 capacitors + switch + connectors in that measurement. D factor is probably better to look at, but still need to consider the network being measured.

Yeah, that's a lot of stuff to get to the actual DUT capacitor, and why we prefer the direct connect method rather than a decade box with banana connectors, switches, wires and such. It's very difficult to "null" all this out, whereas the direct method allows a very good and repeatable Short and Open Calibration which closely represents the DUT conditions.

We thought about getting a Decade Cap box awhile back, but decided against such for the reasons mentioned. Altho using it for a "switchable" cap too fine tune circuits is quite useful.

Anyway, you might get a better result (lower Rs) by directly Kelvin Connecting to the DUT cap inside the box.

I think it's pretty awesome for what it is...and the price I paid. I would not have invested significant money into either of these decade boxes.

That said, I've already found the resistance box to be exceptionally useful exactly as you mentioned for fine tuning circuits.

For example, repairing my K2010: https://www.eevblog.com/forum/repair/keithley-2010-repair/msg6279934/#msg6279934

I replaced the TC=0 Zener diode (VR122) which directly impacts measurement stability. R359 is the resistor that supplies the current to VR122.

Before making the adjustments, with a VREF10-001 as the voltage source, the K2010 had a 37.5uVpp over 10 minutes. After using the decade box to adjust, I replaced R359 with the value from the decade box.

Now the K2010 has a 10.26uVpp over 10 minutes (only rises to 10.46uV at 30 minutes, and to 16.58uVpp after 12 hours). And that's still with another error I need to correct.

Thanks,
Josh

[mawyatt]

That's seems very high for Rs (0.127Ω) for a Reference Cap of 1uF @ 1KHz!!

Maybe? But keep in mind it's a decade switch with 4 capacitors together to get that value.

You've got 4 capacitors + switch + connectors in that measurement. D factor is probably better to look at, but still need to consider the network being measured.

The paralleling to get to 1uF will make it difficult for any internal measurements. D, Q and Rs all use the same information for computations, so each will reflect results based upon the other two (assuming a known C).

Best

[KungFuJosh]

You could try cycling all the switches a few times and see if the value changes at all.

I did that in the cleaning process.

There's no error in the reading or anything wrong with the decade capacitor; it's a network measurement of capacitance, not a singular capacitor measurement.