Resistance transfer 10k -> 10Meg with Solartron 7071

[e61_phil]

Hi,

the Solartron 7071 is specified with an ohms transfer of 0ppm of reading + 1 ppm of Range in all the ohms ranges (except the 1Gig). I would think that means 11ppm of uncertainty for an 1 to 10 transfer within a range. To verify that I used two resistors with low tolerances. First one is a Vishay VHP101 10k with 10ppm tolerance and the second one is a series connection of two Caddock USF270 5Meg with 100ppm tolerance (10Meg).


To transfer the resistance from 10k to 10Meg I used my Fluke 5450A and the Solartron 7071. I've done the following:

Solartron 7071 in 100k Range:
measure VHP101 10k
measure Fluke 5450A 100k
calculate "real" value for the 100k

Solartron 7071 in 1Meg Range:
measure Fluke 5450A 100k again
measure Fluke 5450A 1Meg
calculate the "real" value of 1Meg with the "real" value of the 100k

Solartron 7071 in 10Meg Range:
measure Fluke 5450A 1Meg again
measure Caddock 10Meg and calculate the "real" value

The value of the Caddock should be within 100ppm (Caddock tolerance) + 10ppm (Vishay tolerance) + 3*11ppm (Solartron transfer) = +/- 143ppm of 10Meg
However, the result (see attachment) is ~300ppm away from 10Meg.


After that I did a short crosscheck with the Fluke 5440B. I hooked the Vishay 10k up on my 34401A and connected the 10Meg in series. I applied 1000V on this divider and I measured the voltage on the 10k. According to the 30days specification of the Fluke 5440B the ratio uncertainty should be better than 11ppm. With this method, I measured the 10Meg within 56ppm. Which is in tolerance.

What is my mistake with the Solartron transfer?

Best regards
Philipp

PS: I measured the cable and the cable insulation is better than 2TOhms

[amspire]

I think you are reading the 7071 specs wrong. When they quote the transfer spec, that is the case where you have say, a 10K reference and you trim a second 10K resistor to exactly the same reading as the reference. When after you go back and forwards a few times and find the readings are the same, you can say they are within 1ppm fs relative accuracy which equates to 2pmm for a 10K resistor (since full scale is 20K).

If you are reading a 10K and a 100K on a 200K range, then the 10K accuracy is 20ppm + 1ppm fs which equates to a maximum error of 40ppm. The 100K reading is accurate to 22ppm, and so the relative accuracy guaranteed can be as high as 62ppm guaranteed. This assumes the temperature is constant. I do not know the 7071 - that was just based on my quick reading of the specifications page.

A better approach would be to build yourself some resistive transfer boxes. These are closed boxes with 10 or 12 resistors of the same value in series. Something like decent wirewound would be good. You want the resistors to match in value to at least 0.1% (which can give you a 1ppm transfer accuracy), but 0.01% match would be be fantastic. You use them like this. Add shorts so 10 of the resistors are in parallel - that gives you a R/10 value. Open up the shorts and you have a 10R value. With 0.1% matching resistors, the R/10 and 10R values will match relatively within 1ppm. Now there are some other patterns you can use. If you use 9 resistors in series you get 9R. If you have 3 series groups of 3 in parallel, you get 1R and these will match relatively within 1ppm. So with the one box, you can accurately calibrate 3 ranges as long as you have an accurate reference (like the 10K resistor) to do the initial 20K range calibration. All you need from the resistors is short term stability and negligible voltage coefficient (like precision wirewound resistors).

Richard.

[e61_phil]

I think you are reading the 7071 specs wrong. When they quote the transfer spec, that is the case where you have say, a 10K reference and you trim a second 10K resistor to exactly the same reading as the reference. When after you go back and forwards a few times and find the readings are the same, you can say they are within 1ppm fs relative accuracy which equates to 2pmm for a 10K resistor (since full scale is 20K).

Thanks! The FS is 200k and not 100k 
However, I thought one should use the transfer spec in the same way as any other specs. At least Agilent will do it this way in "Calculating Measurement Uncertainty using Digital Multimeter Ratio Measurement Techniques" (http://literature.cdn.keysight.com/litweb/pdf/5992-1058EN.pdf?id=2643219).

If you are reading a 10K and a 100K on a 200K range, then the 10K accuracy is 20ppm + 1ppm fs which equates to a maximum error of 40ppm. The 100K reading is accurate to 22ppm, and so the relative accuracy guaranteed can be as high as 62ppm guaranteed. This assumes the temperature is constant. I do not know the 7071 - that was just based on my quick reading of the specifications page.

Why 62pmm?

10k   * 0ppm of reading  + 200k * 1ppm = 20ppm of 10k
100k * 0 ppm of reading + 200k * 1ppm = 2ppm of 100k

That should lead to 22ppm uncertainty for the 10k to 100k transfer in the 200k range according to the Agilent AppNote.

A better approach would be to build yourself some resistive transfer boxes. These are closed boxes with 10 or 12 resistors of the same value in series. Something like decent wirewound would be good. You want the resistors to match in value to at least 0.1% (which can give you a 1ppm transfer accuracy), but 0.01% match would be be fantastic. You use them like this. Add shorts so 10 of the resistors are in parallel - that gives you a R/10 value. Open up the shorts and you have a 10R value. With 0.1% matching resistors, the R/10 and 10R values will match relatively within 1ppm. Now there are some other patterns you can use. If you use 9 resistors in series you get 9R. If you have 3 series groups of 3 in parallel, you get 1R and these will match relatively within 1ppm. So with the one box, you can accurately calibrate 3 ranges as long as you have an accurate reference (like the 10K resistor) to do the initial 20K range calibration. All you need from the resistors is short term stability and negligible voltage coefficient (like precision wirewound resistors).

I'm searching for a fully automated setup to calibrate my Fluke 5450A. Using a DMM would be a convient way, but the higher range (above 1Meg) are difficult due to the changing input current of the meter.

Philipp

[amspire]

Why 62pmm?

10k   * 0ppm of reading  + 200k * 1ppm = 20ppm of 10k
100k * 0 ppm of reading + 200k * 1ppm = 2ppm of 100k

That should lead to 22ppm uncertainty for the 10k to 100k transfer in the 200k range according to the Agilent AppNote.

You using the transfer specifications that doesn't really apply to relating a value at 5% of full scale to one of 50% of full scale. If there is a specification for linearity, then you could use this along with the transfer specification. If there isn't, then all you can rely on is the accuracy specifications. The accuracy specifications add up to 62ppm (22ppm + 40ppm).

The thing with these specifications is in practice the results are usually way better then the specs, but worst case, you could get the 10K resistor measuring, say, 40ppm high, and the 100K measuring 22ppm low. That is where the 62ppm worst case error can come from.

[Dr. Frank]

Hello Philipp,
the relevant specification is the linearity, which is given for the DCV only, at about 0.2ppm.
Using a DCV ratio technique with your 5440B, 10:1 transfers of around <2ppm per Ohm range were possible.
At high Ohm values, you may use the full 20V range, (i.e. 20V => 2V transfer), which reduces the errors..The current source of the Solartron may not be as linear.. and Ohm linearity is not specified, as far as I can see.

Please compare that to the calibration of my 5450A:
https://www.eevblog.com/forum/blog/eevblog-544-fluke-5450a-resistance-calibrator-teardown/msg1113471/#msg1113471

There,  I even compensated for the bias current of my 3458A, but the real problem were leakage currents inside the 5450A, which were really hard to discover.

That I also assume in your setup... Check your cables and measure their isolation resistance / leakage currents.
This may easily give such gross errors.

Frank

[e61_phil]

You using the transfer specifications that doesn't really apply to relating a value at 5% of full scale to one of 50% of full scale. If there is a specification for linearity, then you could use this along with the transfer specification. If there isn't, then all you can rely on is the accuracy specifications. The accuracy specifications add up to 62ppm (22ppm + 40ppm).

The thing with these specifications is in practice the results are usually way better then the specs, but worst case, you could get the 10K resistor measuring, say, 40ppm high, and the 100K measuring 22ppm low. That is where the 62ppm worst case error can come from.

Yes, 5% is maybe to low for the transfer. i have to look into the manual again. However, could you explain to me how your 40ppm is calculated, please? I end up with 20ppm for the single 10k measurement and another 2ppm for 100k.

Hello Philipp,
the relevant specification is the linearity, which is given for the DCV only, at about 0.2ppm.
Using a DCV ratio technique with your 5440B, 10:1 transfers of around <2ppm per Ohm range were possible.
At high Ohm values, you may use the full 20V range, (i.e. 20V => 2V transfer), which reduces the errors..The current source of the Solartron may not be as linear.. and Ohm linearity is not specified, as far as I can see.

Please compare that to the calibration of my 5450A:
https://www.eevblog.com/forum/blog/eevblog-544-fluke-5450a-resistance-calibrator-teardown/msg1113471/#msg1113471

There,  I even compensated for the bias current of my 3458A, but the real problem were leakage currents inside the 5450A, which were really hard to discover.

That I also assume in your setup... Check your cables and measure their isolation resistance / leakage currents.
This may easily give such gross errors.

Frank

My understanding of a transfer specification is exactly what I have done. This is also described in the linked document for the 3458A. See Example 3 for "manual ratio mesasurement using DMM's transfer accuracy specification". They transfer 1V to 10V using the transfer specifications of the 10V Range (0.05ppm of reading + 0.05ppm of range). This leads to 0.05ppm + 0.5ppm for 1V and 0.05ppm + 0.05ppm for 10V = 0.65ppm. The same way as I calculated the transfer with the Solartron.
Maybe, I'm wrong or HP/Agilent/Keysight is talking about another thing than Solartron.

Thanks, for the interesting link to your 5450A investigations.

I already measured the cable and they are above 2TR. That shouldn't matter.


Edit: It was only a crosscheck with the 5440B. This 1:1000 Transfer should be quite good (10,4ppm using the 30days specifications). For 1:10 voltage transfers, I would the think the 34401A is better than the 5440B.

[amspire]

Yes, 5% is maybe to low for the transfer. i have to look into the manual again. However, could you explain to me how your 40ppm is calculated, please? I end up with 20ppm for the single 10k measurement and another 2ppm for 100k.

The accuracy is 20ppm + 1ppm of full scale. +/- 1ppm of full scale is on the 200K range is +/- 0.2 ohms.  For a 10K resistor, this equates to +/- 20ppm. So the total error for the 10K resistor is +/- 40 ppm. Without a linearity specification for resistance, you have to add the error for each measurement because you cannot assume the error for each reading is in the same direction.

The thing about the resistor transfer process I suggested is it is orders of magnitude better and its accuracy comes from physics and mathematics instead of the trust in how an instrument agrees with a spec sheet. As long as you use resistors with a negligible voltage coefficient, you will see if there is a stability problem with the resistors simple by going back and forward between the two resistances about 3 times. If the resistor values read the same each cycle to less then 1ppm, you know you are down near the 1ppm accuracy relative accuracy between the ranges even with resistors that match to 0.1%.

With the method you are suggesting, you just will not know the accuracy for anything but the 10K range - you just know each transfer could have an error somewhere between 0ppm and 62ppm (or more if you look at temp variations, emf, range reference resistor changes, etc).

[e61_phil]

Yes, 5% is maybe to low for the transfer. i have to look into the manual again. However, could you explain to me how your 40ppm is calculated, please? I end up with 20ppm for the single 10k measurement and another 2ppm for 100k.

The accuracy is 20ppm + 1ppm of full scale. +/- 1ppm of full scale is on the 200K range is +/- 0.2 ohms.  For a 10K resistor, this equates to +/- 20ppm. So the total error for the 10K resistor is +/- 40 ppm. Without a linearity specification for resistance, you have to add the error for each measurement because you cannot assume the error for each reading is in the same direction.

I used the transfer specification which is 0ppm + 1ppm

I absolutely agree with your solution of using proper resistance transfer standards. However, it is a lot of handwork needed to calibrate the Fluke 5450A this way. If the transfer specification of the Solartron would really allow for a 22ppm transfer in the ranges above 1Meg it would be accurate enough for most of the meters.

[VintageNut]

Yes, 5% is maybe to low for the transfer. i have to look into the manual again. However, could you explain to me how your 40ppm is calculated, please? I end up with 20ppm for the single 10k measurement and another 2ppm for 100k.

The accuracy is 20ppm + 1ppm of full scale. +/- 1ppm of full scale is on the 200K range is +/- 0.2 ohms.  For a 10K resistor, this equates to +/- 20ppm. So the total error for the 10K resistor is +/- 40 ppm. Without a linearity specification for resistance, you have to add the error for each measurement because you cannot assume the error for each reading is in the same direction.

I used the transfer specification which is 0ppm + 1ppm

I absolutely agree with your solution of using proper resistance transfer standards. However, it is a lot of handwork needed to calibrate the Fluke 5450A this way. If the transfer specification of the Solartron would really allow for a 22ppm transfer in the ranges above 1Meg it would be accurate enough for most of the meters.

I posted in a separate thread that my 5450A was calibrated at a primary standards lab as a favor to me. The tech/engineer who performed the calibration told me that it required 30 minutes for every individual resistance of the 5450A. It was an all-day effort to calibrate my 5450A by a trained professional with a top-of-the line MI bridge.  They have some standard resistors in a bath right next to the MI bridge.

[The Soulman]

Definition of transfer accuracy: (from a Keysight manual).

Transfer Accuracy
Transfer accuracy refers to the error introduced by the multimeter
due to noise and short-term drift. This error becomes apparent when
comparing two nearly-equal signals for the purpose of “transferring”
the known accuracy of one device to the other.

I.E. the short term accuracy, not transferring/scaling between ranges.

[e61_phil]

I posted in a separate thread that my 5450A was calibrated at a primary standards lab as a favor to me. The tech/engineer who performed the calibration told me that it required 30 minutes for every individual resistance of the 5450A. It was an all-day effort to calibrate my 5450A by a trained professional with a top-of-the line MI bridge.  They have some standard resistors in a bath right next to the MI bridge.

That's why I'm trying to do it automated

Definition of transfer accuracy: (from a Keysight manual).

Transfer Accuracy
Transfer accuracy refers to the error introduced by the multimeter
due to noise and short-term drift. This error becomes apparent when
comparing two nearly-equal signals for the purpose of “transferring”
the known accuracy of one device to the other.

I.E. the short term accuracy, not transferring/scaling between ranges.

I havn't scaled between ranges, but I found a text in the solartron manual which also says "between similar values". Therefore, what is allowed for the 3458A in voltage mode isn't allowed to the solartron ohms mode.

Let's forget the transfer accuracy and take the 24h specifications. That should allow transfers within a range.


Solartron 7071/7081 FS is 1.4

100k range: 3ppm reading + 1ppm of range
(3ppm + 140k/10k * 1ppm) + (3ppm + 140k/100k * 1ppm) = 21.4ppm (first reading + second reading)

1Meg range: 4ppm of reading + 1ppm of range
(4ppm + 1.4/0.1 * 1ppm) + (4ppm + 1.4/1 * 1ppm) = 23.4ppm

10Meg range: 10ppm of reading + 1ppm of range
(10ppm + 14/1 * 1ppm) + (10ppm + 14/10 * 1ppm) = 17.7ppm

This sums up to 80.2ppm for the transfer + 10ppm for VHP101

Therefore even with the 24h specifications the reading should be 10Meg +/- 200ppm and not +300ppm

[amspire]

Solartron 7071/7081 FS is 1.4

I didn't realise the full range was 1.4 - I assumed it was just under 2

100k range: 3ppm reading + 1ppm of range
(3ppm + 140k/10k * 1ppm) + (3ppm + 140k/100k * 1ppm) = 21.4ppm (first reading + second reading)

1Meg range: 4ppm of reading + 1ppm of range
(4ppm + 1.4/0.1 * 1ppm) + (4ppm + 1.4/1 * 1ppm) = 23.4ppm

10Meg range: 10ppm of reading + 1ppm of range
(10ppm + 14/1 * 1ppm) + (10ppm + 14/10 * 1ppm) = 17.7ppm

This sums up to 80.2ppm for the transfer + 10ppm for VHP101

Therefore even with the 24h specifications the reading should be 10Meg +/- 200ppm and not +300ppm

You keep assuming that the 7071 is perfectly linear on ohms mode, even though there is no specification that says that it is.

Without a linearity spec, you have to add the error for a 10K resistor on the 100k range to the error for a 100K resistor. You cannot just say "I am only looking at a relative reading so I will chuck out the absolute accuracy term". The thing is the 100K reading may be out in one direction, and the 10K out in the opposite direction and the meter is still within specs.

So on the 100K range, here are the errors:

10K resistor - 20ppm + 1ppm of fs = 20ppm + 1*(140K/10K) = 34ppm
100K resistor: 20ppm + 1ppm fs = 20ppm + 1*(140K/100K)  21.4ppm

Add the two together and you have 55.4 ppm. That is a very different number to your 21.4ppm error. 55.4 ppm is according to the specification sheets the worst case amount the 100K resistor reading could be out when compared to the 10K reading on the 100K range. It is not 21.4ppm

I haven't included a 24 hour specification. It is not really needed if you do the transfer in one go and you repeat the two readings several times to ensure there is no extra drift in the short term. That is one specification you can get rid of.

If the 7071 was linear to within 1 count - as you seem to want to assume - why wouldn't the manufacture want to proudly publish this fabulous performance in their specifications? I gather they do specify linearity for volts, but don't for ohms. There is a message in that.

It is possible to get greater accuracy then the specifications if you test the meter and find out what its true linearity is. But to do that you need something that has a known absolute relative error - such as $30 of resistors matched to 0.1% in a resistor transfer box. I haven't suggested using the 5450A since it would add a whole lot of extra errors into this process. The 5450A would be over an order of magnitude worse then $30 of resistors wired as a resistive transfer.

The point I am trying to make is 1ppm accurate resistive transfer is not actually that hard or expensive to achieve with the right methodology. With relays, you probably could automate the resistive transfer method. Below 10K, the contact resistance becomes a big issue, and at high resistances, the relay leakage current is an issue. You always have to take error sources into account.

If you can live with the relatively big errors your method will introduce, then your method is completely fine. The possible errors will just be much higher then you have been hoping.

[e61_phil]

You keep assuming that the 7071 is perfectly linear on ohms mode, even though there is no specification that says that it is.

No, I apply the 24h specifications and these should include non-linearity and so on.

Without a linearity spec, you have to add the error for a 10K resistor on the 100k range to the error for a 100K resistor.

That is what I've done.

You cannot just say "I am only looking at a relative reading so I will chuck out the absolute accuracy term". The thing is the 100K reading may be out in one direction, and the 10K out in the opposite direction and the meter is still within specs.

So on the 100K range, here are the errors:

10K resistor - 20ppm + 1ppm of fs = 20ppm + 1*(140K/10K) = 34ppm
100K resistor: 20ppm + 1ppm fs = 20ppm + 1*(140K/100K)  21.4ppm

Add the two together and you have 55.4 ppm. That is a very different number to your 21.4ppm error. 55.4 ppm is according to the specification sheets the worst case amount the 100K resistor reading could be out when compared to the 10K reading on the 100K range. It is not 21.4ppm

The difference between my 21.4ppm and your 55.4ppm is the used specification. I used 24h and you use 1year.

I think one can apply the 24h specification if one will do all transfers within 24h. To use a known value and measure it IS a calibration. The one year spec will include drift over one year which doesn't matter for the transfer. All short term issues like, noise, drift and so on should be included in the 24h specifiation.

I haven't included a 24 hour specification. It is not really needed if you do the transfer in one go and you repeat the two readings several times to ensure there is no extra drift in the short term. That is one specification you can get rid of.

There are other problems than drift. An input voltage dependend input current will generate a systematic error which will be the same even after 100 measurements. Another thing is the current source which may be dependend on the load resistance. And of course the linearity of the ADC.
Therefore, I think it isn't that easy to use your own specifications.

If the 7071 was linear to within 1 count - as you seem to want to assume - why wouldn't the manufacture want to proudly publish this fabulous performance in their specifications? I gather they do specify linearity for volts, but don't for ohms. There is a message in that.

I don't want to assume it is linear to one count. Therefore, I applied the 24h specification instead of the transfer specification. My assumption with the transfer specification as used in the HP 3458A application note was wrong.

It is possible to get greater accuracy then the specifications if you test the meter and find out what its true linearity is. But to do that you need something that has a known absolute relative error - such as $30 of resistors matched to 0.1% in a resistor transfer box. I haven't suggested using the 5450A since it would add a whole lot of extra errors into this process. The 5450A would be over an order of magnitude worse then $30 of resistors wired as a resistive transfer.

The point I am trying to make is 1ppm accurate resistive transfer is not actually that hard or expensive to achieve with the right methodology. With relays, you probably could automate the resistive transfer method. Below 10K, the contact resistance becomes a big issue, and at high resistances, the relay leakage current is an issue. You always have to take error sources into account.

If you can live with the relatively big errors your method will introduce, then your method is completely fine. The possible errors will just be much higher then you have been hoping.

In the first step I would like to have reliable uncertainty caluclations. And even if a I add the 24h uncertainty of the used Fluke 5450A resistors to the transfer calculations the result is nearly the same (6ppm for 100k and 7.5ppm for 1Meg).

What do you mean with "$30 of resistors"? $30 for each resistor?

Thanks for the very interesting discussion!

[amspire]

You keep assuming that the 7071 is perfectly linear on ohms mode, even though there is no specification that says that it is.

No, I apply the 24h specifications and these should include non-linearity and so on.

Without a linearity spec, you have to add the error for a 10K resistor on the 100k range to the error for a 100K resistor.

That is what I've done.

You cannot just say "I am only looking at a relative reading so I will chuck out the absolute accuracy term". The thing is the 100K reading may be out in one direction, and the 10K out in the opposite direction and the meter is still within specs.

So on the 100K range, here are the errors:

10K resistor - 20ppm + 1ppm of fs = 20ppm + 1*(140K/10K) = 34ppm
100K resistor: 20ppm + 1ppm fs = 20ppm + 1*(140K/100K)  21.4ppm

Add the two together and you have 55.4 ppm. That is a very different number to your 21.4ppm error. 55.4 ppm is according to the specification sheets the worst case amount the 100K resistor reading could be out when compared to the 10K reading on the 100K range. It is not 21.4ppm

The difference between my 21.4ppm and your 55.4ppm is the used specification. I used 24h and you use 1year.

I think one can apply the 24h specification if one will do all transfers within 24h. To use a known value and measure it IS a calibration. The one year spec will include drift over one year which doesn't matter for the transfer. All short term issues like, noise, drift and so on should be included in the 24h specifiation.

You are using the 24 hour stability specifications - they are nothing to do with error. A meter could have a 1000ppm nonlinearity over a range and have 1ppm stability.

If you compare the 7071 spec to the 3458A, you will notice the 3458A has a spec for 24 hour accuracy. The 7071 does not have this spec at all. Just stability.

The limits of error (1 year) for the pair of resistors is 55.4ppm.

If the 7071 was linear to within 1 count - as you seem to want to assume - why wouldn't the manufacture want to proudly publish this fabulous performance in their specifications? I gather they do specify linearity for volts, but don't for ohms. There is a message in that.

I don't want to assume it is linear to one count. Therefore, I applied the 24h specification instead of the transfer specification. My assumption with the transfer specification as used in the HP 3458A application note was wrong.

It is possible to get greater accuracy then the specifications if you test the meter and find out what its true linearity is. But to do that you need something that has a known absolute relative error - such as $30 of resistors matched to 0.1% in a resistor transfer box. I haven't suggested using the 5450A since it would add a whole lot of extra errors into this process. The 5450A would be over an order of magnitude worse then $30 of resistors wired as a resistive transfer.

The point I am trying to make is 1ppm accurate resistive transfer is not actually that hard or expensive to achieve with the right methodology. With relays, you probably could automate the resistive transfer method. Below 10K, the contact resistance becomes a big issue, and at high resistances, the relay leakage current is an issue. You always have to take error sources into account.

If you can live with the relatively big errors your method will introduce, then your method is completely fine. The possible errors will just be much higher then you have been hoping.

In the first step I would like to have reliable uncertainty caluclations. And even if a I add the 24h uncertainty of the used Fluke 5450A resistors to the transfer calculations the result is nearly the same (6ppm for 100k and 7.5ppm for 1Meg).

What do you mean with "$30 of resistors"? $30 for each resistor?

Thanks for the very interesting discussion!

$30 for 10 resistors. You would need 4 boxes - one for 10R/100R/1K, one for 1K/10K/100K one for 100K/1M/10M, one for 10M/100M/1G.

Although spending $$$ for super stable, low drift resistors makes the use of the boxes much simpler, you can get 1ppm accuracy with fairly cheap resistors. If the resistor is good enough to give a stable reading within 1ppm for a few minutes (assuming it is in a box and not subject to air currents), and if it has negligble voltage coefficient (like a wirewound), it is good enough. No calibration (other then the 0.1% matching) is needed to get 1ppm. I noticed you can get wirewound power resistors in the case that can be bolted to a heatsink with a 25ppm/C coefficient for just a few dollars. Could be perfect. Could help minimize temperature change during the measurement.

Using 12 resistors just gives more serial/parallel patterns you can use, and it lets you pick a few different group of 9 or 10 resistors just for a confirmation. You can only use symetrical patterns to end up with error terms that are equal to the square of the resistor matching error. You cannot for example have a pattern that has a group of 3 resistors in parallel in series with a group of 4 resistors in parallel.

If you are rich, you can buy boxes like this: http://www.ietlabs.com/esi-sr1010-resistance-transfer-standard.html

but you are paying a lot for matching and stability that is convenient, but not essential. The simple maths is that if the resistors match within 0.1%, then the achieved 1:100 ratio matches within 1ppm. You will be able to see on the 7071 if the resistors are stable or drifting during the period of the measurement. The resistors you use can be 1%, but with lower precision resistors, you would want to check regularly to make sure they all match to 0.1%. 1% resistors can often have spec with a lot of long term drift.

[amspire]

If you want to play with the resistive ratio transfer boxes, I have attached a simple spreadsheet. You enter the matching error (in percent) and it uses the random number generator that makes 10 resistors within the error percentage and works out the ratio. With totally random 0.1% resistors, the 100:1 ratio accuracy is typically within 0.4ppm.

With 0.01% matching or better, the ratio errors get so small that other errors will dominate.

The F9 key in Excel or Libre Calc will force a recalc so you get a new set of random resistors.

Just appears the 100:1 actual ratio can never be below 100:1. So the error is always positive.

Richard

[e61_phil]

You are using the 24 hour stability specifications - they are nothing to do with error.

 
Yes, I mixed it up with 24h accuracy. I have never seen such a 24h stability specification before. Thanks for your patience with me

$30 for 10 resistors. You would need 4 boxes - one for 10R/100R/1K, one for 1K/10K/100K one for 100K/1M/10M, one for 10M/100M/1G.

Although spending $$$ for super stable, low drift resistors makes the use of the boxes much simpler, you can get 1ppm accuracy with fairly cheap resistors. If the resistor is good enough to give a stable reading within 1ppm for a few minutes (assuming it is in a box and not subject to air currents), and if it has negligble voltage coefficient (like a wirewound), it is good enough. No calibration (other then the 0.1% matching) is needed to get 1ppm. I noticed you can get wirewound power resistors in the case that can be bolted to a heatsink with a 25ppm/C coefficient for just a few dollars. Could be perfect. Could help minimize temperature change during the measurement.

Using 12 resistors just gives more serial/parallel patterns you can use, and it lets you pick a few different group of 9 or 10 resistors just for a confirmation. You can only use symetrical patterns to end up with error terms that are equal to the square of the resistor matching error. You cannot for example have a pattern that has a group of 3 resistors in parallel in series with a group of 4 resistors in parallel.

If you are rich, you can buy boxes like this: http://www.ietlabs.com/esi-sr1010-resistance-transfer-standard.html

but you are paying a lot for matching and stability that is convenient, but not essential. The simple maths is that if the resistors match within 0.1%, then the achieved 1:100 ratio matches within 1ppm. You will be able to see on the 7071 if the resistors are stable or drifting during the period of the measurement. The resistors you use can be 1%, but with lower precision resistors, you would want to check regularly to make sure they all match to 0.1%. 1% resistors can often have spec with a lot of long term drift.

Thanks, for these infos! I already own a SR1010-10Ohm/step and I'm buying another 1k/step one. Furthermore, I already own a couple of Vishay wire wound power resistors. Marcus here in the forum has already some experiences with these resistors and they seem to be better than specified.

I the past I build a 10kV measurement system and I also included a 1:16 hamon divider for calibration. I attached my ratio error "simulation" with 100ppm resistors. In this high voltage application I end up with high voltage shorting cables. Any relais had ruined the accuracy of the box due to bad insulation. For lower voltages than a couple kV it could be possible with reed relais.