EEVblog Electronics Community Forum
General => General Technical Chat => Topic started by: quarks on December 06, 2012, 07:22:59 pm
-
Hello all,
probably as most of us, here in the forum, I am very interested in metrology (especially accuracy and calibration know how).
Therefore I studied the Keithley Low Level Measurement Handbook and read a lot of the very interesting posts and articles from Conrad Hoffmann (also member of this forum) about metrology, KVDs, Hamon Divider and bridges.
Inspired by robrenz (and with his permission) I used his low resistance spreadsheet, expanded it, added some of my better gear and The Reference Meter Fluke 8508A.
I was quite surprised to see, how my meters compare and that even my good 6.5 digit (Tek DMM 4050 and Gossen 30M) meters are quite bad (below 1 Ohm and above 1 MOhm).
To be able to get better results, the best seems to be working with Bridges and resistance references. But unfortunately I do not have a Null-Meter. Is there anything newer/better to use (maybe like the Null-Detector for DVMs from Conrad Hoffman)? And has anyone good readings about how resistance transfer standards work?
Another way could be adding a Micro-Ohmmeter (in the range 0.01% to 0.1% down to 1 mOhm ) and for high Ohms an Electrometer.
But so far I have not found a nice meter like robrenz's LOM-510A in Germany (or at least Europe). If you have suggestions, what could be worth to have a closer look at, please let me know.
My almost perfect all in one dream could be the Fluke 8508A, but besides the expected price tag, I was not even able to find a source for it. If there is one (or maybe a Datron/Wavetek 1281 / Fluke 1281A) please let me know.
thx
quarks
Edit: refined the chart and added the Burster 2304 Resitomat, which so far is the best Micro-Ohm Meter I could find
Edit: finally got the Fluke Reference Multimeter all in one solution, perfect for mOhm to GOhm
-
Look at the Ratio mode on you tek4050 it should be the same as my 8846A. That will funtion as a null meter at 1ppm resolution except it will read 1.000000 when nulled instead of zero.
The % error on the chart is % of value.
-
That's a good hint. Have you used this function, is it usefull and with stable reading? When I read the German manual I did not understand how to use it. It seems to be badly translated, so I will check the english version and give it a try.
Thx
-
It is the ratio of the input voltage which is input on the left hand "INPUT" terminals divided by the reference voltage which is input on the right hand "SENSE" terminals. You also need to short the commons together at the meter and have analog filter off. I assume 100 NPLC will give more accurate and stable readings.
-
It is the ratio of the input voltage which is input on the left hand "INPUT" terminals divided by the reference voltage which is input on the right hand "SENSE" terminals. You also need to short the commons together at the meter and have analog filter off. I assume 100 NPLC will give more accurate and stable readings.
Many thanks, I really overlooked and never used this indeed very usefull function, because of the bad translation in my German manual. Now I tested it and checked the Self Calibration procedure on my Fluke 720A and I am very pleased that all steps appear to be just perfect.
PS: I edited and refined the chart and added the Burster 2304 Resitomat, which so far is the best Micro-Ohm Meter I could find
-
Too bad nobody (except robrenz) had a suggestions for me so far.
I kept on testing the suggested ratio method with my KVD 720A ("2-39. Measuring Unknown Resistance" as described in the manual). For best accuracy it is suggested to use a Reference Resistor in the same order of magnitude, low emf leads and a Lead Compensator (the last two I do not have, therefore I used just ordinary unshielded test leads (but at least good gold plated banana with 2.5mm² cable)).
For this test I used my Burster Decade (0.01% and TK<1ppm/°C) as reference (Rstd) and a 1k resistor (as unknown Rx, because I wanted to be able to go up and down in value with Rstd) and dialed in 10/100/1k/10k and 100k Ohm. The calculated result were then 994.73/995.61/995.73/995.65/995.69 Ohm.
Even I did not have an ideal setup, the results seem to be quite accurate. I wonder if anyone has experience with this kind of measurement and likes to share it. If so, please let us know what you think. Is it worth keep on going this way? What improvement can be expected with low emf leads (maybe POMONA shielded cable with Tellurium/Copper gold plated bananas) and a Lead Compensator?
thx
quarks
-
I should be demoing one of these over Christmas: http://www.home.agilent.com/en/pd-1000001296%3Aepsg%3Apro-pn-34420A/micro-ohm-meter?&cc=GB&lc=eng (http://www.home.agilent.com/en/pd-1000001296%3Aepsg%3Apro-pn-34420A/micro-ohm-meter?&cc=GB&lc=eng)
I'll let you know how I get on.
-
That was originaly on my chart. @ quarks, did you remove it on purpose? Very nice specs
-
Hello lewis,
thanks for the reply. Please let us know, when you got it.
That was originaly on my chart. @ quarks, did you remove it on purpose? Very nice specs
Hello robrenz,
I have not removed it from the sheet, but when I expanded your low resistance sheet up to a wider range (from 1µ to 200G Ohm) I did not update the specs of the 34420A above the 200 Ohm your data was, so it looked funny for me not to have the rest of the data in the chart. I now have included it again with your original data. All other graphs show the whole range from every meter (except the two that cross the y axis, they go further down, see screenshot). So where the individual line starts it is the max. resolution (least significant digit) and where it ends it is the max. of its range.
Right now I have 15 meters in the sheet. Only for the Burster 2303 I have not found a datasheet with the needed specs. If anyone has this, it would be great to have a copy. If there is another meter worth having a closer look if it performs even better than the ones I already checked, please let me know.
regards
quarks
edit: during my lunch break I edited the sheet, now all graphs go to the max. range
-
I do not know, if anyone else is interested, but while experimenting with resistors in the range from 100mOhm - 1mOhm (see attached picture), I found out that my Hameg 8118 seems to have missleading accuracy information in the documentation (see attached picture).
Instead of the claimed 10µOhm to 100MOhm, it goes down to 1µOhm and up to 1GOhm. I wonder if any one else has this meter and this could be Firmware related (my Version is 1.31 and the latest is 1.52). Any feedback would be helpfull.
Actually I do not know why this is, but this meter is measuring 1mOhm with resolution of 1µOhm quite well (last digtit is not very stable, but I repeated the measuremet many times, because I could not believe it).
Here are my findings (see picture), with meter set to SLOW, average 20 and auto. Before every measurement I shorted the Kelvin leads for zero adjustment.
Accuracy seems to be at least in the range of the test resistors (0.1% to 0.5%). So I am really surprised, because I was prepared to see errors in the range of 100% for 1mOhm and 10% for 10mOhm. .
-
I am interested but I dont have any input on this question ;D
-
Because of the promissing results, I did measure a few more 100mOhm Resistors (Dale LVR-10 see picture) to select a possible candidate for a DIY Reference Resistors project. And I achieved a very stable 100.00 mOhm reading. What ever the real value might be, this looks really good.
-
I've been thinking about a resistor reference to, something like the one Dave made.
I was thinking on using a couple of Vishay Z-Foil resistors (Z201T and Z201T),
RS stock# 763-7309 and 763-7303.
These are 0.005% tolerance and 2 ppm/C.
The plan was to solder these directly to a gold plated banana jack, perhaps something that fits directly into the DMM ports.
Still thinking tho', these resistors is expensive. :)
-
Do not solder direct, rather use a set of standoffs and solder the resistor to that ( so that leads have no strain) or make a small PCB to hold the resistor, then from the pads make a 4 wire connection to have the sense and force wiring separated as close to the resistor as possible. Then short thick wires to the 4 panel mounted sockets, or another 2 wires to the pins on a small box, so that ant flexing os the case on insertion will not be transferred to the resistor. With a large enough resolution you can actually see the resistor changing value as you apply stress to it by pressing the body, and with these high accuracy ones it will show up thermal stress as well.
-
Thanks for the ideas, SeanB.
-
Yes, the Vishay Z-Foil are the best candidates for these DIY references I know of.
Besides the Z201 I also had a closer look at the HZ Series with Zero TCR (these go even down to 0.001% and 0.2 ppm/°C).
But I have not yet tried to buy some, so I don't know if I really can get them.
Meanwhile I have some very nice BURSTER Type 115 with 0.004% and around 1ppm/°C (this data is individually printed on each resistor, very nice; see picture).
But these are no longer available and the successor Types 1140/1150/1160/1178 are not quite as good (spec wise).
-
If it's ok to ask in your thread. Does anyone know of any good candidates for salvaging these kinds of resistors from old unwanted/faulty equipment?
-
If it's ok to ask in your thread. Does anyone know of any good candidates for salvaging these kinds of resistors from old unwanted/faulty equipment?
That is exactly what I did. I bought a defective Burster Digistant DC Standard 6406 as for spare parts for my working one.
But if you also want to do that, you should know what is inside and what you are willing to pay for it.
The downsite is, at least for the reference purpose I am looking for, if cannot meassure/compare the parts to known/good values, you may end up not to be confident enough to rely on these salvaged parts as your only reference.
-
Besides the Z201 I also had a closer look at the HZ Series with Zero TCR (these go even down to 0.001% and 0.2 ppm/°C).
But I have not yet tried to buy some, so I don't know if I really can get them.
That's really interesting, wonder what the prices of those would be. ;)
I'm in no rush with this as I don't have a need for that precision just yet.
I don't even have a mOhm meter yet, but that is certainly on the wishlist. ;)
-
Do not solder direct, rather use a set of standoffs and solder the resistor to that ( so that leads have no strain) or make a small PCB to hold the resistor, then from the pads make a 4 wire connection to have the sense and force wiring separated as close to the resistor as possible. Then short thick wires to the 4 panel mounted sockets, or another 2 wires to the pins on a small box, so that ant flexing os the case on insertion will not be transferred to the resistor. With a large enough resolution you can actually see the resistor changing value as you apply stress to it by pressing the body, and with these high accuracy ones it will show up thermal stress as well.
That is good input and I would do it the same way, you suggest. But what Du you think about the Banana posts and the solder? Do you know if it is worth thinking about Tellur Copper (like Pomona 3770) and if there is low emf solder (like the one with Cadmium)? I still cannot not really understand why this could make a big difference compared to very good gold plated posts (like from Hirschmann or maybe MC). The thermo couple / contact material would be Gold to Gold anyway. Maybe you or someone else can help me understand.
-
IMO for only this application of 4 wire measurement of resistance, all concerns about material combination EMF's is mute. You have to use some form of thermal emf cancellation anyway at low ohm levels. It also only applies to the sense terminals not the source. So use offset compensated ohms or current reversal to eliminate the thermal EMF. The type of testing we are talking about here is static so we don't need to worry about rapidly changing thermal emf voltages.
-
I had them ad constants into a computer, where you could see them drift if you flexed the board, and where they were used as gain setting resistors around the input amplifiers. Solder on one and it changes value for hours afterwards.
-
That's really interesting, wonder what the prices of those would be. ;)
If I remember correctly, I saw a price of 50$ or 60$ for one "HZ Series with Zero TCR". But because I do not even remember where that was, please do not rely on this.
Maybe someone else has a source and likes to share with us.
-
Checked RS prices, they are about 54 USD + tax and shipping for the Z201T.
If someone finds better prices, preferably the HZ series, I'd be very interested.
-
thx robrenz,
that is what I thought and with AC measurement like with HAMEG 8118 it is anyways not relevant. So I will keep on going with the Hirschmann posts/plugs I already have.
But I still wonder, what the impact of gold plated Tellur Copper could possibly be, compared to other gold plated materials. I would think only the contact metal (here Gold to Gold) itself should be responsible for the resulting Seebeck-Effekt (if I do not have an error in the way I think of it).
For what I know so far, Tellur seems to be used mainly because it makes copper easier to machine work with.
Also in the LT app note AN86, Tellur seems to be not even mentioned. Nevertheless the connection type comparison on page 49 (see screenshot) is quite interesting and I think and hope if the contact materials are the same (I use only Gold plated Hirschmann banana plugs and posts, with identical base material), everything should be fine.
Edit: Just found a good app note from Fluke http://support.fluke.com/find-sales/download/asset/2548277_6001_eng_a_w.pdf (http://support.fluke.com/find-sales/download/asset/2548277_6001_eng_a_w.pdf)
-
Checked RS prices, they are about 54 USD + tax and shipping for the Z201T.
If someone finds better prices, preferably the HZ series, I'd be very interested.
that seems a bit to high. I looked at mouser and it was about 20€ in Germany
-
My secret trick for low thermal emf wiring is a length of multi-conductor phone wire I salvaged from a junk pile somewhere. The stuff is solid unplated copper wire and has proved excellent for interconnecting voltage sources, 6 1/2 digit meters and KVDs.
Plated bananas, lugs and similar are a disaster. Even if you have to cut your own lugs from copper sheet you'll be better off. I've yet to have to resort to low emf cadmimum solder, but if you see solder joints marked with (I think) green paint in L&N equipment, those use the special solder.
Any place you have dissimilar metals that create thermal emfs, put the outgoing and return joints near each other so they track in temperature and cancel out.
-
Hello Conrad,
thanks a lot for your reply. As stated in my initial post, you are also one who inspired me with your articles and knowledge to do all this interesting stuff.
So my gold to gold contact to be the best for no trouble is busted.
With your input I do even remember that I have read this before, on a very informative site (see link, unfortunately mainly in german, so anyone intersted here should look for translation)
http://www.amplifier.cd/Test_Equipment/other/other_Test_Equipment.htm (http://www.amplifier.cd/Test_Equipment/other/other_Test_Equipment.htm)
Also there was a suggestion like yours, to use telephone (massive) copper wire. But unfortunately I never tried it and even forgot about it.
About using pure copper I have in mind that oxidation is to be the worst for emf. How do you take care of this?
Also all of my meters only have female banana plugs and therefore no post to screw the massive copper wire or a lug on. Can you suggest what to do on the meter site?
About the gold plating on the tellur copper posts (like the Pomona 3770 and IET BP-1000, I think these are also on my ESI resistors), what do you think about them? And would you still suggest to use pure copper instead of a gold plated lug or male banana with them?
Thx
quarks
-
Checked RS prices, they are about 54 USD + tax and shipping for the Z201T.
If someone finds better prices, preferably the HZ series, I'd be very interested.
that seems a bit to high. I looked at mouser and it was about 20€ in Germany
Yes, that's more like it. Thanks. :)
-
with the input of Conrad, I did some testing around cables/connectors/adapters.
Please be aware this is not scientific at all and therefore maybe even useless, but here is what I did and found:
- I let my Valhalla 2701C (set to 100mV) and TEK DMM 4050 (set to 100mV DC, D-Filter, 100 PLC, High Input Z) warm up for > 3h
- I switched beetween different cable and connector types (see picture) several times
- I noted the readings every time after a settling time of > 30s to 60s
- after that I calculated the average of each cable set and the average of all cables and calculated the individual deviation compared to this average
- ambient temp was stable 20.4 °C
There are several uncertainties during the measurement readings, like drift of 2701C and DM4050 during the test, individual contact contamination of each connector, me sitting in front of it and so on.
And out of all of that, the max. calculated difference was 5µV (and this was between the BNC RG58 and the HCK 1m cables).
Is this reasonable or does any one have a suggestion to do it better?
thx
quarks
-
My voltage sources and KVDs have the usual tall banana jacks, as does my HP 3455, so using the wire directly is no problem. Even if you have to use bad banana plugs it should be no problem if they're at the same temperature- the thermal emf of each one will cancel out. It's usually the other end I have trouble with because that's the end I handle with my hot fingers. Everybody gold plates their tellurium copper parts so I have to believe it's no big deal. Copper oxide is supposed to be really bad, so I fresh strip my wires and or clean them by pulling through a piece of Scotchbrite. If you're going to work down at the PPM level, thermal emfs are just a way of life and you learn to identify them and either fix them or compensate by waiting for things to stabilize.
Did you know that long ago somebody made a power source by connecting eight zillion thermocouples together so you could put the assembly over a fire, stove or heat source?
-
Hello Conrad,
with the telephone cable, I think of using male bananas, where I can screw it directly in, instead of soldering it.
About the gold plating, I still imagine it is best, especially when both contact materials are gold. At least one can be sure it is the best against corrosion.
With the Tellur Copper parts, have you any clue how Tellur can do anything for low emf?
About the thermocouple powersource, I have never heard of that.
Bye
quarks
-
I came across this one:
http://www.edn.com/design/systems-design/4402904/Complete--low-cost--software-programmable-ohmmeter-measures-micro-ohms (http://www.edn.com/design/systems-design/4402904/Complete--low-cost--software-programmable-ohmmeter-measures-micro-ohms)
-
thanks for the link qno, I will have a look at it.
-
I found an interesting article at a German university (unfortunately it is written in German)
http://www.uni-magdeburg.de/exph/messtechnik1/Parameter_Thermoelemente.pdf (http://www.uni-magdeburg.de/exph/messtechnik1/Parameter_Thermoelemente.pdf)
In there is a nice table with "Seebeck-Koeffizienten" (see picture). As I read/understand it, Gold should not do any harm as it has the same coefficient as Copper (6.5 µV/K).
But Tellur has 500µV/K, in my understanding this should make Tellur Copper connectors bad for low EMF.
I hope some here knows more than I do and can either confirm what I think or can explain why Copper/Tellur connectors are still the best for low emf.
-
Another source of great info and knowledge on high precision stuff is Leslie Green's book : Analog SEEKrets, and its free and you can get it here -> http://eevblog.com/files/seekPDF.pdf (http://eevblog.com/files/seekPDF.pdf) , this book is highly praised around here in this forum, personally I think its even better than Art of Electronic. :-+
Here a snippet to spark your interest ;)
(https://www.eevblog.com/forum/chat/free-pdf-500-page-book-on-analogue-design-secrets-worth-a-look/?action=dlattach;attach=19706;image)
-
wow 588 sites to read, now I am busy for quite some time
thx a lot BravoV, that's a nice present for Christmas
and merry Christmas to all
bye quarks
Edit: just a short update, I am impressed, this is really worth reading and sure is a great source to learn from!!!
-
I hope some here knows more than I do and can either confirm what I think or can explain why Copper/Tellur connectors are still the best for low emf.
I don't know more than you do but here are my thoughts (not necessarily facts) on gold plated tellurium copper connections relative to thermal EMF.
Tellurium copper C145 is only used because of it's machinablility and mechanical properties, not because of its electrical properties. It's conductivity is only 93% of IACS. It only has 0.5% tellurium so I don't think its Seebeck coefficient is going to be radically different than straight copper. The Seebeck coefficient of Copper and Gold only differ by 20µV at 100 degC relative to Platinum with a reference of 0 degC. I know It's not linear but that would be approximately 0.2µV/degC so pretty small. IMO Gold plated Tellurium copper has gotten a unjustified super low thermal EMF status that is totally unjustified. Gold plated C101 oxygen free copper would be a better material but not practical for the typical machining required for connectors.
The seebeck charts I have seen do not correlate the metals (Cu) with semiconductors (Te) so I am not sure the 500 for TE and the 6.5 for Cu are on the same scale
BUT thermal EMF's are generated by a thermal gradient in a material. The actual junction of the two dissimilar materials has nothing to do with the voltage generated other than, 1 establishing an electrical connection and, 2 having the two materials at the same temperature at the junction.
The gold plating is extremely thin (typ 1/4 to 1/2 micron) and should have an almost insignificant thermal gradient relative to the tellurium copper it is plated on. In my mind this is no different than a isothermal block situation where secondary junctions cancel. I think that in a thermally static situation any plating material will not generate a significant thermal EMF. Lets say this statement is blatantly wrong and it does generate a thermal EMF. Lets assume unplated copper wire leads that are perfectly deoxidized and swaged (not soldered) to the gold plated Tellurium copper Spade connector which is connected to a gold plated tellurium copper binding post which has a internal swaged (not soldered) unplated deoxidized copper wire connection. We now have a Cu>Au>Cu>Au>Au>Cu>Au>Cu connection and it is its own isothermal block so all the Cu-Au thermocouples cancel each other. It still cancels if one is not gold plated Cu>Au>Cu>Au>Cu>Au>Cu. Notice I stated swaged not soldered connections because the solder will be another thermocouple junction allthough these can also cancel if they are symmetrical and close to isothermal conditions with the main connection.
Banana plugs that are gold plated are almost always brass that is gold plated and the actual spring contact is gold plated beryllium copper. This adds several more thermocouple junctions in the mix. and the banana spring connection is not a thermally efficient situation so we cant consider it a isothermal situation. So this is obviously inferior to spade lug connections.
IMO using short duration or single pulse offset compensated 4 wire ohms measurements on a thermally stable setup eliminates all these uncertainties.
-
Generally agree but I don't think the thickness enters into it. Nickel (or whatever) plated alligator clips are so bad as to be worthless, and they're probably a copper or brass underneath.
-
I don't know more than you do but here are my thoughts (not necessarily facts) on gold plated tellurium copper connections relative to thermal EMF.
Thanks a lot that you share your "thoughts". I really appreciate it and wish I would only have a little bit of your knowledge and skills. I already learned a lot from you, Conrad and others here. But I realize that I have to learn far more to get where I want to. I am still at the Low Ohms only, but really enjoyed the learning path.
Here are some more details I hope are helpful to go on with.
(data is from various sources I already referred to and where possible double checked against diff. Application Notes)
(https://www.eevblog.com/forum/chat/help-wanted-low-and-high-ohm-measuremet/?action=dlattach;attach=35827)
If there are any errors, please let me know.
One thing I still hope to get deeper knowledge of, is how Resistance Transfer Standards work. Because I am desperate to find out about it, I will go ahead with Kelvin Double Bridge and Lead Compensation topics (and just ordered some Fluke and ESI gear). I hope this way I will learn a lot and If I am able to understand it, I should also get very good accuracy.
Bye
quarks
-
There is no need to avoid bare copper to copper connections. The Agilent 34420A nanovolt meter/micro ohm meter uses "almost pure copper connections" (translated Tellurium copper) in its connector. They recomend Caig Deoxit G100L on the connector and even supply it with the meter. So use Deoxit G100L and if you have heavily oxidized copper just use a metal polish to clean the copper and then use the Deoxit.
-
here is a short update.
Today I received my long awaited kelvin ratio bridge and I just used my lunch brake for a first quick/dirty test. Attached is the setup. I know it looks very messy and not ideal at all, but I am happy with the very first result. I compared my Burster 1424 to my SR104, used the Valhalla 2701C as generator and the TEK DMM4050 in Ratio Mode (100PLC) as Nulldetector (thx again to robrenz for this great tip). The Deviation on the bridge shows exactly 0.01% compared to SR104 which happens to be exactly within the tolerance of the Burster. Now I still do not know the absolute exact value but I am confident that all my involved gear is within its own specs. When I have more time I will also do comparisons with my SR1010s.
-
Wow some nice gear there and interesting test. :-+
-
here is a short update of the comparison chart.
Now with Wavetek 1271 and the ESI 240C Kelvin Ratio Bridge (which is suppose to be 10ppm accurate in direct readout and has an amazing 0.1ppm resolution)
edit: error correction in chart
-
Nice! This chart makes you realize why you can't find any .001% .001 ohm resistors. My LOM-510A is pretty good but I want a Burster :'(
-
To bad I missed this topic until now. Good to know I'm not alone in the quest for the holy grale of precision ( but sometimes it feels like chasing windmills)
I had the same troubles about what cables. I tried massive copper ( VD wire, you use for mains in the walls), coax, triax, shieded telephone wire. With normal bannas, goldplated ones, bnc-banana adapters, home made copper forks to screw under terminals ect,
Problem is you get different results but what is the right one ( the man with to many clocks story...)
I am very interested do some tests too when my lab is finnished. I sold a bunch of Racals that were in a 19 " rack. Now all my calibrators and KV deviders are in there and the Guildline standard cell cabinet is on top.
I have also the 1010 resistors and a GR decade. I can meaure G for high Ohms
I can measure using a GR1608, Boonton 63H, Prema and solartron 7,5 digit, Keithley 2000 and 196.
Also a HP sourcemeter that goes into the PetaOhm region but it is analog.
And the two KV deviders ( Fluke 720 and ESI decavider) but I have to figure out how to do this. There is a brief description in the Fluke manual as I recall.
Maybe there is a way to measure the one KV with the other using a calibration source and a null detector and so calculating some parmeters ?
Problems with seebeck I only experienced while calibrating, but if you wait long enough the contacts have the same temperature so no seebeck ( I assume, Roberts seems to tell the same if I understand it correct)
I had problems with dirt ( for very high Ohm) , pressure ( the grip of the kelvin clips, but even tiurning a banana a bit can give different reading in the mOhm or lower end) and current that heats up the resistor.
I found out that cleaning a 10 M resistor with ipa and steelwool for the legs made the difference between unstable and stable readings, but the writer of my Bible ( analog SEEKrets) seems to have other experiences. I measured leakage of my Fluke 720. I am very bad in remembering numbers but it was a factor 100 to much leakage. Putting the KV on 2 cleaned teflon blocks made leakage less, so the table conducted to much. i then cleaned the whole inside and outside with TRI ( at that time it looked like a good idea, and IPA after that) that improved leakage about 150X. Still not what the factory described but the best I could do. ( my 720 was total loss when I got it, had to make all swtich isolation myself, because all the studs and wiper adapters were broken in many peaces. Photos are on my site)
Now I stop, I can talk about this stuff for hours ;-)
Should i use some format for testing. You talk about a spreadsheet but I have to rest now so I'm typing this on my ipad and I can not open it right now, but should i use that ?
The Keithley book is for me too the source of this madness :-)
-
To bad I missed this topic until now. Good to know I'm not alone in the quest for the holy grale of precision ( but sometimes it feels like chasing windmills)
As we both already found out, we are chasing for ppm/precision. But you have so much great gear I feel like a beginner/hobbiest. Nevertheless I do have fun with it and today I received a Fluke 721A Lead Compensator (see picture) for my next test together with my 720A to see if Lead Compensation makes a noticable difference. But right now unfortunately I do not have enough spare time to do that very soon. BTW the KVD method is a quite easy setup.
With my actual gear I have good results in the range 1Ohm to about 10MOhm. Values below that and at the really High Ohms I have not yet touched. But I am looking forward to get there.
About the cables I am still experimenting (see my latest collection: Coax Aircell 5 and 7, special ESI and Fluke cables) and Conrad Hoffmann's secret tip with the twisted pair/telefon cable is really good, when you have binding posts on your meter. Triax cables come to my list, when I go >100MOhm. It would be nice I you could provide cable types and connectors you use.
I will post an update when I have results.
-
Nice! This chart makes you realize why you can't find any .001% .001 ohm resistors. My LOM-510A is pretty good but I want a Burster :'(
luckily the chart only is showing the datasheet specs.
As I know from your posts, especially your LOM-510A is almost perfect accurate. To bad I cannot get one here in Germany/Europe.
Also my TEK DMM4050 seems to be far more accurate than it should be. According to my Wavetek 1271 and 10k Standard Resistor SR104, it seems to be around 10x better than the specs. ;D
-
Just in case there is still some interest, I finalized my tests/research with good repeatable results in the ranges I was most interested in (from mOhm to GOhm)
-
am interested in this, missed it originally unfortunatelly.
what were your findings here? what I am specifically interested in is:
-what accuracies did you achieve with the 720A, and how does that compare to the 3458A being used for voltage divider measurements?
-did the 721 lead compensator turn out to be required? the 720 manual says beyond 100 ohms total chain resistance it should (not shall/must) be used, but it is not shown in the schematics part.
My error propagation calculations show that between 10 ohms and 1megohm, the 3458a can be used reasonably well, with only a few ppm of accuracy error added, beyond, the 720a needs to be used.
-
it is a long time ago, but from my memory the Fluke 720 method was about as accurate as my TEK DMM 4050 (which turned out to be very accurate). But in my tested setup the 4050 was the Null Detector and I do not know if a Fluke 845 would lead to even better accuracy.
About Lead compensation, I have only played around with this setup and therefore have not really solid data. But as described in the 721 Manual "It is particularly useful in those situations demanding the utmost in accuracy".
Although with the Fluke 720/721 it was very interesting to see how they where used for unknown resitance measurements in the old days , I would (and did) go for an easier DMM solution.
-
thanks for answering this old thread.
if you say you would rather go for an easier DMM solution, I do not really see which one would be more precise than a 720A.
when working from a 10k standard accurate to better than 0.5ppm, the error propagation calc shows derived values accuracy to be 5ppm and better from 10ohms to 1Mohm (using voltage divider method and a 3458a).
however at 1 ohm and 10Mohm, the inaccuracy increases to arround 50ppm, one reason being low voltage range inaccuracies. so for these values, a 720a seems to be the only solution to get something better.
do you have any suggestions (other than a measurements internatinal bridge). Also, I am using these resistors to vaildate the accuracy of a 3458a artefact cal, so using my 3458a is not a solution
-
Overall I found the Fluke 8508A to be the best way to go for. Accuracy combined with convenience seems to be very hard to beat with other gear.
-
sure, the 8508 is a very good instrument, however its transfer accuracy is worse by about a factor of 2 than the 3458A, hence it does not solve this problem better unfortunatelly, it makes it worse.
-
if a 8508A is still not good enough, maybe you should look at ESI 242D Resistance Measuring System (very old but a great piece of art) or have a look at Guildline Model 6625A. According to Guildline "the 6625A System is the industry standard replacement
system for the historical ESI 242".
-
yes, thanks, if money was no issue the 6622A xr would be a fine piece of gear. my guess is this is arround 50k though. that does not make sense.
the 242d was actually what I was trying to emulate by using the 720a and the 721 lead compensator, and given the fact that the 720a is self-cal (the 242d resistor decade is more complicated to calibrate, also it has quite a high temp drift of several ppm/K, and I see fluctuations of 1-2 K during a measurement cycle), I would think it should not be worse in resistance comparison. at least on the paper the 720a solution looks actually better.
the nullmeter can be adjusted to better than 1uV I would think, and at 10v or 100v total (for the 10megs measurement) with 0,1ppm referenced to input should give some reasonable results. i have not done the math yet, will do soon, and am waiting for the 721 to arrive, so doing tests to see how the noise is doing will be important too. I thought that you might have had some results here overall.