For my Ph. D. on High-Tc-Superconductors I had to build several precision low-temperature thermometers, i.e. where thermoeletric voltages have most influence, compared to current precision. I used a PT100, @1mA, for it's stability, useful downto ~ 20K, and a precision diode, DT 470, @10µA, 1.5..300K, both from Lake Shore, as far as I remember. The Keithley programmable precision current sources were too expensive, as I needed 2-3 fixed currents only, so I built several precision current sources for our institute, 1/100th the price, with the following BOM: REF02 (5V reference) Industrial Chopper OpAmp 7650, low Ibias <10pAe Low-Ileak JFet, e.g. BS170 ww. precision resistors, T.C.= 1..3ppm/K, 0.1%, 4k99, 49k9 and 499k custom specific from German manufacturer Burster Or standard 5k / 500k ww from alpha electronics, sold by Rhopoint Components, U.K. 20 turn trimmers of 20 Ohm, 200Ohm, 2k to trim to 5k/500k exactly, or to trim standard resistors over range of +/- 0.2%, ~1ppm resolution. 20 turn trimmer to trim the REF02 to exactly 5.000V or to a value a little bit above that, for the 5k / 500k + trim pot standard resistors. Several 5V Reed relais (steered by parallel port) to switch between current ranges and to zero current (open) - most important for thermoelectric compensation of PT100, more important than the precision of current. The 7650 regulates exactly a voltage of 5.000V over the precision resistor, connected to ground, by means of an JFet, producing no additional cross current. The load is between the Drain of the JFET and V++ (=15V), giving enough compliance to feed the sensor. I also had an HP3458A to calibrate the current source. Without that, any effort for a precise current is worthless, but you should provide the metrological means for producing it. (I.e. the idea of buying a standard Vishay metal foil is ok, just measure the voltage @ 1mA). Current is by far the most difficult unit to measure or to produce, as it mostly depends on voltage PLUS resistance.. The circuitry delivered about 50ppm stability for 1mA, a little bit less for 10µA. This is very, very good, please compare this to the Keithley current sources datasheets. Much more important for the precision of measuring temperature, using a PT100, is the cancellation of thermoelectric voltages. Therefore, for PT100 you should use 1mA current and a possibility to measure the thermoelectric voltage (without current). The HP34401A has no possibility to make this compensation automatically (a pity) So you have to substract the theremoelectric voltage from the resistance voltage, produced by the 1mA current source. 4W / Kelvin wiring is mandatory. I also used a HP34401A just for temperature measurement, just using the DCV mode. DCV is quite precise/stable, resistance and current is mediocre, but you have the Vishay metal foil for that. By this method, you may get 10mK precision, provided you have a well calibrated sensor. All additional effort in precision and cost for the currents' precision is useless, ie. with those ideas, cost may be below 50$/€ for such a high precision current source. Frank

On Sun, Aug 15, 2010 at 11:05:50PM +0200, Dr. Frank Stellmach wrote: > The 7650 regulates exactly a voltage of 5.000V over the precision > resistor, connected to ground, by means of an JFet, producing no > additional cross current. > The load is between the Drain of the JFET and V++ (=15V), giving enough > compliance to feed the sensor. Your setup is more or less similar of what I have originally thought. -*- I planned to use the MAX6225/6325 reference divided down to 1V, a non-chopper (but very low bias current, very low offset voltage and 0.1uV/C offset voltage drift) OP177 (I was worried of the noise of the chopper), and a common middle power MOSFET I already have, that actually is the weak ring of the chain with his 100nA of gate to body leakage... By the way, I was not able to find a MOS in my component box with gate-body leakage less than 100nA, even small ones. I have not small signal MOS, just (even low) power ones... there are still discrete small signal MOS around? I may buy your BS170 with his 10nA (anyway it's not a JFET, but a MOSFET too). There is a 2SK1656 with 5nA, but I wonder if, at 10uA, a darlington made of superbeta BJTs would not do better (using maybe the IT124) or simply a JFET 2N3819 (max -2nA of reverse current at -15V, but in normal working condition it should stay into the pA range). I already have some of them, I'll try it. -*- Actually, Bruce suggested me that an independent voltage source, however stable, may not be a good choice vs. one tracked to the voltmeter internal one. So my first try would be to replace the MAX6225 with a 1K resistor. This is assuming that the internal circuitry used inside the 34401A to generate the ohm current source (plus my external resistor), is less drifty than his voltage reference. I will try to verify this assumption. The box of components has arrived this morning, so I'll try soon. > I also had an HP3458A to calibrate the current source. Without that, any > effort for a precise current is worthless, but you should provide the > metrological means for producing it. (I.e. the idea of buying a standard > Vishay metal foil is ok, just measure the voltage @ 1mA). I am not so lucky to have a 3458A. I would like to... Anyway I just measured the 10K Vishay resistor with the 34401A using kelvin connection and, after warmup and a math null against a short, it measured 10.0000 Kohm stable. Seems promising. > Current is by far the most difficult unit to measure or to produce, as > it mostly depends on voltage PLUS resistance.. And temperature is harder again :) > Much more important for the precision of measuring temperature, using a > PT100, is the cancellation of thermoelectric voltages. I didn't thought of that, but I see that this matters very much even in measuring my 10K resistor. You disconnected power source with the reed, right? > measure the thermoelectric voltage (without current). The HP34401A has > no possibility to make this compensation automatically (a pity) If you control relays with a computer, you could send the math null command over the interface port. > I also used a HP34401A just for temperature measurement, just using the > DCV mode. DCV is quite precise/stable, resistance and current is > mediocre, but you have the Vishay metal foil for that. Yes. If you say that resistance mode is mediocre, probably the internal current source is driftier than the reference, thus an external reference would be better. Actually my measures with the 10K resistor are anyway good... how do you check the (low) performance of the meters? > By this method, you may get 10mK precision, provided you have a well > calibrated sensor. The sensor is my limit. My goal is that will be the only limit, so I could improve simply by buing a better one. Best regards, Andrea Baldoni