My first 5.5 digits meter Solartron 7150 and LT1021 ref

[Grapsus]

Hi there,

I've been reading the measurement section of this forum for a while and I've never felt the need to have a meter with more than 3.5 digits but now I think I was contaminated and you are all responsible for this 

Last week I got my first bench meter. It's a Solartron 7150, it is marked as "Enertec/Schlumberger" and has all inscriptions in French but it looks identical in every aspect to the Solartron for which I found the manual on this forum. I got the meter from a craigslist-like ads site for 60 euros.

It was a good surprise when I saw that it got its latest cal in 2010. From the stickers it looks like it has been used at an Alcatel lab since 1994. I don't know if it has been manufactured in 1994 or it was moved to that lab in 1994 since the 7150 has been around since the 80s.

I have read here that 7150 have a nasty mains filter with capacitors that can leak and spill their contents inside the meter. Do you think I should replace them now ? Will opening the case and fiddling with the board have any impact on the calibration ?

Unfortunately I don't have any precise references to tell if the device is really in cal. I've tried comparing its readings with my other less precise meters and at least they all agree within their respective specs. It might still be off, but less than the two last digits.

The closest thing to a voltage reference I've found in my parts bin was two LT1021CCN8-5. I've made two identical very simple boards powered by 9V batteries and with a trimming circuit exactly as described in the datasheet page 9. They talk about using a "cermet" type multi-turn trimmer but I only had old 10 turn trimmers from a mysterious brand (that google doesn't know about) so I hope that they are stable enough. I let the meter warm up for several hours and I trimmed the two references as close to 5.00000 Volts as I could. Then I've been re-measuring them every day since Friday to see if anything drifts.

After building the references I realized that 5V is a bad value, since the first digit goes only up to 2, but I don't have a stable enough divider to get 1 Volt. I use the meter in FILTER mode, in this way it takes many measures and averages them to get better reading, it also displays one extra digit so I have a 10 microvolts resolution with my 5 Volts source.

All in all everything seems pretty stable. When the meter has just been turned on, it is about 20 microvolts off for both references. After an hour it gets to 5.00000 Volts and starts slowly wandering between 4.99998 and 5.00002 Volts, which is 4 ppm. At this stage both the source and the meter are within spec but I really wonder if the slow variations come from the LT1021 or from the meter.

Here are the pictures of my reference board, feel free to tell me if I have done anything bad in terms of precision here. I have really no experience with precision electronics.

What would you advise me if I want to build dividers to get 1 and 0.1 Volts ? Do I need a chopper stabilized op amp to buffer the output of the divider ? Also, maybe the next step is to get a few LM399 and see how stable they read ?

[wiss]

It is a 6 digit meter, at longer integration times you will get one more digit. Try pressing the 6x9 button a few times!

[Grapsus]

It is a 6 digit meter, at longer integration times you will get one more digit. Try pressing the 6x9 button a few times!

Yeah on the pictures I already pressed the 6x9 button and the meter is in 6.5 digits mode. Sadly the first digit only goes up to 2 which is less than 5, which is why the five goes on the next position and we get 6 digits : 5.00000. With a 1 Volt reference I would get 7 digits: 1.000000

[wiss]

Yeah on the pictures I already pressed the 6x9 button and the meter is in 6.5 digits mode. Sadly the first digit only goes up to 2 which is less than 5, which is why the five goes on the next position and we get 6 digits : 5.00000. With a 1 Volt reference I would get 7 digits: 1.000000

Yes, of-course, didn't notice 
It is still a full 6 1/2 digit
It is 1 GOhm+ at low ranges so you can measure a "normal" divider directly. Maybe not with resistance values in 1 Meg range but it will work with kohms.
It will need hours of warmup for top performance.

[Grapsus]

It will need hours of warmup for top performance.

I confirm that it takes at least an hour for the last digit to stabilize after powering up the meter in 6.5 digit mode.

Also I noticed that the LT1021 datasheet suggests that the IC should be totally isolated from ambient air movement. In that conditions they say 0.1 Hz noise goes below 0.5 ppm, and otherwise it might go as high as 20 ppm ! Since my LT1021 are in free air, I think it's really something I should consider.

For my first measurements I used those hideous cables that I made myself a few years ago. Now I understand that if the materials are poorly chosen and that some junctions are at different temperatures pretty significant voltages can arise at those junctions. I tried searching the forum on this topic, but didn't find anything. Do you have any suggestions about the test leads and banana plugs ? Can decent leads be bought with an amateur budget ? Where can I find practical info about the temperature coefficients of common junctions like copper/lead ?

It is 1 GOhm+ at low ranges so you can measure a "normal" divider directly. Maybe not with resistance values in 1 Meg range but it will work with kohms.
It will need hours of warmup for top performance.

Ok, thanks, I will look up the exact input impedance and try to calculate the maximum deviation caused by the loading of the divider.

I still don't know what to use to make those dividers. Foil resistors cost a fortune and can't be trimmed. So far I am thinking about buying lots of cheap resistors and searching for matched TC or buying suitable low-TC wire and trying to make my own wirewound resistors. Any suggestions on this ?

[TimFox]

For my first measurements I used those hideous cables that I made myself a few years ago. Now I understand that if the materials are poorly chosen and that some junctions are at different temperatures pretty significant voltages can arise at those junctions. I tried searching the forum on this topic, but didn't find anything. Do you have any suggestions about the test leads and banana plugs ? Can decent leads be bought with an amateur budget ? Where can I find practical info about the temperature coefficients of common junctions like copper/lead ?

I have had good luck making my own banana-plug test leads FOR NON-SAFETY PURPOSES using "audiophile" gold-plated banana plugs (branded Nakamichi, not sure of actual manufacturer).  These have gold-plated insides with two setscrews pushing the wire against the inner diameter of the barrel.  They are available with cool-looking UNSAFE metal outsides or with somewhat better plastic shells.  They are not stackable, however.
At the other end, I use real Mueller copper (not nickel-plated) alligator clips.  The medium size is available with a screw to clamp the wire against the body.  They are available with colored vinyl covers for a modicum of safety.
Also, I use stranded bare copper (not tinned) test lead wire, which is a bit expensive.  The silicone-insulated wire from http://caltestelectronics.com/ctitem/75-coaxial-test-lead-wire/CT2799
is very flexible.  I think it best to avoid solder in these cables, but have not tested that theory.
If you are worried about thermal emf, avoid nickel-plated connectors, since nickel has about 10 uV/K emf against copper, but gold is < 0.5 uV/K.  Copper oxide has a bad emf against copper, so clean the wire before clamping into the connector.  Silver is similar to gold, but it tarnishes (sulfide).

[wiss]

It will need hours of warmup for top performance.

I confirm that it takes at least an hour for the last digit to stabilize after powering up the meter in 6.5 digit mode.

Using similar meters, I take 4 hours for anything serious: get home from work, turn the stuff on, and then by 2200 you can start comparing refs

Also I noticed that the LT1021 datasheet suggests that the IC should be totally isolated from ambient air movement. In that conditions they say 0.1 Hz noise goes below 0.5 ppm, and otherwise it might go as high as 20 ppm ! Since my LT1021 are in free air, I think it's really something I should consider.

A plastic bag will give, by far, the largest bang for the bucks here

For my first measurements I used those hideous cables that I made myself a few years ago. Now I understand that if the materials are poorly chosen and that some junctions are at different temperatures pretty significant voltages can arise at those junctions. I tried searching the forum on this topic, but didn't find anything. Do you have any suggestions about the test leads and banana plugs ? Can decent leads be bought with an amateur budget ? Where can I find practical info about the temperature coefficients of common junctions like copper/lead ?

Copper to copper is very good, copper to copper-oxide is much worse than any nickel plating!
Unless you use solder-wire as the lead you should not worry too much about solder joints, just as long as the two copper-parts are in very good thermal contact.

It is 1 GOhm+ at low ranges so you can measure a "normal" divider directly. Maybe not with resistance values in 1 Meg range but it will work with kohms.
It will need hours of warmup for top performance.

Ok, thanks, I will look up the exact input impedance and try to calculate the maximum deviation caused by the loading of the divider.

I still don't know what to use to make those dividers. Foil resistors cost a fortune and can't be trimmed. So far I am thinking about buying lots of cheap resistors and searching for matched TC or buying suitable low-TC wire and trying to make my own wirewound resistors. Any suggestions on this ?

You can make a divider and then see if it is stable over time and temperature, try blowing or breathing over it and see how much it changes. If you happen to find two good resistors you should try to measure the ratio as good as possible, your meter is probably linear to within +-1 ppm, so you only loose 10 ppm at a ratio of ~1:10

This is why you now desperately need a 7.5 digit meter! 

You should also try the same temp-test with cables, plug in the cables and short the far end, measure how much the voltage changes when you hold it in your hand. Some are better and some are worse, BNC seems to perform pretty good! I guess that since one contact point is enclosed in the other there will be no or small thermal variations and therefore no/low thermal EMF. I use a better quality DIN5 for my sol - 7075, it performs well within spec.

[Grapsus]

I have had good luck making my own banana-plug test leads FOR NON-SAFETY PURPOSES using "audiophile" gold-plated banana plugs (branded Nakamichi, not sure of actual manufacturer).

Thank you for the advice, Nakamichi look very nice, especially the ones with plastic shells. The thing I fear with audio plugs is that they're not meant to be plugged/unplugged all day long, usually people leave their speaker leads alone once they're connected. What is your experience with audio-grade gold plating ? Doesn't it wear after a few cycles ?

A plastic bag will give, by far, the largest bang for the bucks here

Good idea

I wonder what kind of material are the 7150 contacts made of. Maybe it's worth cleaning them ?

[TimFox]

I've had good luck so far with the Nakamichi's.  No noticeable degradation of the plugs.
If you're worried, you can use normal banana plugs for everyday use and only use the gold for calibration and other special equations.
If your full-scale range setting is high enough that you don't care about 10 uV errors, then you probably want the safety-style banana plugs anyway.

[wiss]

The audiophile plugs have usually a thin layer of gold on top of TiN, they will still look golden even when the gold has been rubbed off...

[TimFox]

To see if there is any difference, I just ordered some expensive Pomona 5406 plugs (gold-plated BeCu springs) to compare against the cheap ones.