-
#50 Reply
Posted by
TiN
on 31 Aug, 2016 15:47
-
I guessed wiring correctly
Measured input current, ~100uA. Hooked input to LTZ1000 (3458A). No adjustments performed yet.
Also excuse the inproper wiring...
-
#51 Reply
Posted by
WN1X
on 31 Aug, 2016 17:28
-
Winner winner chicken dinner!
-
#52 Reply
Posted by
Vgkid
on 31 Aug, 2016 18:12
-
Nice. Assuming that both meters were cal'd identically/ reading the same. That kvd is only out by 1 counts.
.7006129
vs.
.700612797
-
#53 Reply
Posted by
Macbeth
on 31 Aug, 2016 18:48
-
Well done! The adjustments and characterisation will be interesting.
Unfortunately my 6 decade esi RV622A doesn't have adjustments. Some very specific settings are as high as 10ppm out, but most appear to be within 1-2ppm. (Well with my utterly crude setup anyway)
But at least it has metal shafts!
Ok, I know - not anywhere in the same league...
-
#54 Reply
Posted by
TiN
on 08 Oct, 2016 12:59
-
Taking of shafts.
I'm replacing acrylic bad ones with plastic from my 720A parts box. Took few quick measurements to make 3D CAD in case some poor voltnut want to 3D-print some, as I don't have access to this advanced technology.
PDF-drawing3D STEP214 model file.
.
Next step - assembly and full calibration against 3458A.
-
#55 Reply
Posted by
VintageNut
on 08 Oct, 2016 22:53
-
Measuring the output of the KVD with your DMM introduces an error. At a KVD setting of 0.7XXXXXX the output impedance is about 70,000. Thats about 7ppm loading with a 10G DMM. If the DMM is 10M input, the loading is 7000 ppm.
I use my DMM as a null detector and compare the output of a KVD to an identical voltage. Adjust one or the other voltage until the difference is as close to zero as you can get. In my office, the air conditioning and the unshielded cables make the individual microvolts not extremely stable.
-
#56 Reply
Posted by
zlymex
on 09 Oct, 2016 01:07
-
Linear had to select the buffer amp LTC1152 to meet <10pA Ib in AN86(page 9), otherwise the 100pA maximum Ib will produce possible 0.7ppm error at 70k output impedance of the KVD.
I was wondering why Linear did not choose LTC1052 over LTC1152, the former is better in almost every aspects. Datron used many pieces of LTC1052 in 4910 DC reference.
-
#57 Reply
Posted by
Kleinstein
on 09 Oct, 2016 08:47
-
The LTC1052 is an old type, still using external capacitors. This allows using a lower chopper frequency and thus less bias current due to charge injection. But the external caps are also a little more difficult to handle and can pick up "noise" if the layout is not good.
The LTC1152 does not look like especially high performance (except low noise in the high frequency range) - more like a first try to have the caps inside. There should be better suited AZ OPs with internal caps now. Worst case bias current could be corrected: at least it is reasonably constant with AZ OPs. If time is not that critical one could even chose one with more noise to get lower bias.
-
-
@ TiN : I love your box
-
#59 Reply
Posted by
TiN
on 09 Oct, 2016 14:54
-
Lol, I have no idea whatever that says.
I'll be working on F845AB restore to get it working again, perhaps it would be a better choice for KVD output?
3458A would be then reading the input voltage from MFC.
-
#60 Reply
Posted by
VintageNut
on 09 Oct, 2016 15:38
-
The Fluke 845 is a very good choice for the null meter. Any good microvolt meter will work. When you have null, there are at most a few uV across 1M ohm and so the current drawn is a few pA.
If the uV meter is adjusted well, you can look at individual ppm or sub-ppm and know that the current in the KVD is not causing large error.
In my equipment stash, I have qty 2 of the Keithley 147 that were not expensive to buy. The fluke 845 is crazy expensive in the USA on eBay. I paid $20 for one of my KE147. The seller did not realize that you have to install a short to the input to center the meter needle. The meters drift to one side quickly with an open input. When I received it, I installed the short and then turned it on and it works perfectly. The KE147 has 30nV full scale range up to 100mV full scale in 1-3-10-30.... range increments. It is easy to calibrate against any good DMM. The input cable is expensive but Keihtley still produces the cable for this nV meter. Other obsolete digital nV meters use the same cable. The cable is solid copper, low thermal and is good for the null meter application.
In my office, playing with individual ppm measurements with the KVD and the null meter, it is difficult to get perfectly stable null measurements. The air conditioning and the unshielded cables induce microvolts of drift, noise and offset. I am in the process of buying shielded twisted pair cable with solid copper single-wire for each of the two signals in the cable.
Slicing 1uV into 10 parts riding on a 10.00000V signal is a challenge but it can be done with very a old KVD and a very old uV null meter. The tricks are stable temperature, air flow, low thermal cables with good shielding.
-
-
@ TiN : Something like "My jewellery box . The things I love bring back many memories."
-
#62 Reply
Posted by
Macbeth
on 09 Oct, 2016 20:19
-
-
#63 Reply
Posted by
VintageNut
on 11 Oct, 2016 19:48
-
Hello TiN
If you want to "read" the KVD , you should use a buffer with low pA input bias current. The Jim William appnote is a perfect example.
Normally, the KVD creates a voltage to compare to an unknown voltage.
I insert a picture below to show using the KVD. I have a Fluke 731B driving the 721A Lead Compensator which in turn drives my KVD and it drives my home-brew bridge of two 10k resistors.
I balance the bridge with a piece of wire. I tap the wire to find the balance point. The polarity of the bridge is flipped to look at offset. Moving the wire tap and adjusting the KVD for a "perfect" balance with no offset lets you know that the bridge is balanced and when you are finished the reading on the KVD is exactly 1/2.
My KVD is low by 5 out of 5,000,000 so it is low 1ppm. I paid $300 for this KVD and I am very pleased with it.
The DMM7510 is the null meter.
My plan is to use the bridge inside a DIY cascade divider to adjust every 10V in the bottom 10 resistors to be exactly the same as my 731B 10V.
-
#64 Reply
Posted by
TiN
on 12 Oct, 2016 04:38
-
VintageNut
Yes, I plan to use buffer amp to try first.
Perhaps you can you plot a schematic of your setup?
I don't have 721A Lead Compensator, so thinking ways how I can get away without using it. I have multiple battery powered LTZ sources instead and 845AB (which need to get fixed first), so perhaps if I can get away without 721A if keep isolation high?
-
#65 Reply
Posted by
VintageNut
on 12 Oct, 2016 16:54
-
Lead compensation is mandatory if you want to compare the KVD output to another voltage.
The KVD input draws current which induces voltages into all of the cables/wires between the voltage source that drives the KVD and the KVD input.
You have to balance the voltages of the wiring so that there is no voltage difference between the KVD HI and the HI of your divider under test.
The same for LO. You have to balance the voltages in wiring so that there is no voltage difference between the KVD LO and the LO of the divider under test.
I will post the Fluke diagram for the test setup. It is in the KVD manual and the 721A manual.
-
#66 Reply
Posted by
VintageNut
on 12 Oct, 2016 17:52
-
Here you go. This is the basic idea. Variations of the basic idea are used to compare an unknown resistance against a known standard resistor.
Current is always going to flow in the KVD input. You just do not want current flowing out of the KVD output. If the two sides are balanced and equal, the null meter has 0V across it and there is no current flowing out of the KVD output.
-
#67 Reply
Posted by
TiN
on 16 Oct, 2016 06:48
-
I get that, however my current task is not exactly comparison to unknown divider but calibration of 720 itself, using 3458A as linearity reference. I'll follow initial procedure from manual first.
Alternative idea I have instead of getting 721 might be use of SMU to inject compensating current perhaps to get null..
-
#68 Reply
Posted by
VintageNut
on 16 Oct, 2016 12:26
-
What will you be using to drive the KVD? You have to account for the voltage drop in the cables between the driving voltage and the KVD and the voltage drop inside of your voltage source because of the loading of the KVD on your voltage source.
For example, I drive my KVD with a Fluke 731B. The 731B manual discusses how much error is introduced by loading the 731B. Driving a 100K KVD introduces an error of 0.5ppm. This probably assumes that there is no voltage drop in the cables between the voltage source and the KVD. Any voltage drop in the cables is more error.
You may be able to use your 3458A measure the voltage drop in the cables between the voltage source and the KVD.
If you build a buffer amp to isolate the KVD output, you have to characterize all of the uV of error in the wiring of your buffer amp and in the cabling between the KVD and the amp and between the amp and your 3458A.
Its not a simple plug and play situation.
A null meter with a 1 uV full-scale range or even lower is a good instrument to show you where uV are induced. An analog nanovoltmeter is even more sensitive.
If you have the Keihtley 1801 clone working, that is the best digital nanovoltmeter on the planet.
-
#69 Reply
Posted by
TiN
on 16 Oct, 2016 12:43
-
HP 3245A 10V will be used to drive KVD. I already confirmed it to be stable within 0.5ppm over 24 hours after pimp-up with my LTZ module. Current compliance of this source 100mA, 0.5 ohm output impedance.
Measurement of input voltage will be done by KVD via 3458A, output by buffered 2002.
I'm not looking for plug and pray solution, so it's alright to go round way for the goal.
My A10-nV setup not ready yet, so this 720A project is another checkpoint to get nV-setup going. I need to build LNA and nV-divider to complete the cycle.
Fluke 845AB and Keithley 182M will act as extra meters for sanity checks.
-
#70 Reply
Posted by
VintageNut
on 16 Oct, 2016 15:47
-
Sounds like you have everything well planned. I look forward to seeing results.
For nV divider, the Keithey 260 claims to use all copper components. I have not performed a teardown of the divider since it is enclosed and it works well. What are you planning for the divider?
-
#71 Reply
Posted by
TiN
on 16 Oct, 2016 17:26
-
For divider something DIY will be used. Like
Keithley 262, but without switches.
-
#72 Reply
Posted by
Echo88
on 09 Sep, 2017 13:06
-
I recently aquired a Keithley 720A, sold as used but working. Turns out its was defective.
Failures were as following:
The input resistance in the 1.0 and 1.1 range were 110 and 120kR instead of 100 and 110kR as stated in the manual, the output resistance didnt follow Figure 2-3 (Dial-Setting determines output resistance) and the function switch S8 was very stubborn and would only switch smoothly between about 3-functions instead of all functions. The optical inspection revealed nothing, everything was nice and without obvious errors.
After making a drawing of the function switch S8 and comparing it to the schematic, measuring the whole stuff and so on i concluded that the function switch S8 doesnt work, because it doesnt shunt the decade resistors (for example R301 and R302 shunted by Shunt S2). Thats why the input resistance was 10kR high on both input-ranges. After bypassing the S8-switch between R1051 and S1C (so that it will shunt at least the A-decade correctly) i could get the A-decade to work halfway properly and the input resistance was according to the manual again.
Now it was time to disassemble the S8-switch and see if it was really the culprit: Indeed, the plastic-clutch was broken and the remaining plastic crumbs led to the stubbornness of the switch, while the wipers of the switch didnt move. I think it maybe repairable, but wanted to ask here first what would be the best stuff to glue it back together? I thought of cleaning the metal-shaft on the switch-cover and the remaining plastic shaft in the switch and then gluing it together with epoxy. Since theres also a thread on the metal-shaft and in the plastic-shaft it might have enough surface to properly transmit the torque. Or maybe someone here has such a switch lying around and wants to sell it to me?
Also: After drawing the connections of the S8-switch and measuring the connections in the A-decade and the S2-shunt it appears that the schematic is not fully correct i think (but i might be very well wrong). For example: (look at the lower left side on the schematic) between the Resistor R1051 and Pin 26 of the R314 in the oil tank i measure 0R instead of about 250R (R1048 + R1049 || R1050), so it seems R1051 and the trimmer-combo are interchanged in position on the schematic, while working fine of course.
-
-
Comparing KVDs, I've used a "poor man's compensator", a balancer actually. It's just a length of bare copper wire between the two units, with a copper alligator clip attached at the right point between them. Or soldered. Allows a few ppm of adjustment with minimal thermals or added resistance.
-
#74 Reply
Posted by
plesa
on 10 Sep, 2017 20:36
-
I recently aquired a Keithley 720A, sold as used but working. Turns out its was defective.
Failures were as following:
The input resistance in the 1.0 and 1.1 range were 110 and 120kR instead of 100 and 110kR as stated in the manual, the output resistance didnt follow Figure 2-3 (Dial-Setting determines output resistance) and the function switch S8 was very stubborn and would only switch smoothly between about 3-functions instead of all functions. The optical inspection revealed nothing, everything was nice and without obvious errors.
After making a drawing of the function switch S8 and comparing it to the schematic, measuring the whole stuff and so on i concluded that the function switch S8 doesnt work, because it doesnt shunt the decade resistors (for example R301 and R302 shunted by Shunt S2). Thats why the input resistance was 10kR high on both input-ranges. After bypassing the S8-switch between R1051 and S1C (so that it will shunt at least the A-decade correctly) i could get the A-decade to work halfway properly and the input resistance was according to the manual again.
Now it was time to disassemble the S8-switch and see if it was really the culprit: Indeed, the plastic-clutch was broken and the remaining plastic crumbs led to the stubbornness of the switch, while the wipers of the switch didnt move. I think it maybe repairable, but wanted to ask here first what would be the best stuff to glue it back together? I thought of cleaning the metal-shaft on the switch-cover and the remaining plastic shaft in the switch and then gluing it together with epoxy. Since theres also a thread on the metal-shaft and in the plastic-shaft it might have enough surface to properly transmit the torque. Or maybe someone here has such a switch lying around and wants to sell it to me?
Also: After drawing the connections of the S8-switch and measuring the connections in the A-decade and the S2-shunt it appears that the schematic is not fully correct i think (but i might be very well wrong). For example: (look at the lower left side on the schematic) between the Resistor R1051 and Pin 26 of the R314 in the oil tank i measure 0R instead of about 250R (R1048 + R1049 || R1050), so it seems R1051 and the trimmer-combo are interchanged in position on the schematic, while working fine of course.
I have unit with same symptoms. But I was unable to find the spare part switch. Is this switch still available on market?