current source

Hi Electronic Wiz

Diode forward potential usually drops by 2mV/'C near room temperature, while zener diodes of 5.1V are least temperature dependent. Temperature compensation is best achieved in ICs by using two diodes or two bipolar transistors to compensate for the temperature dependent changes. This is one good reason why ICs are more temperature stable designs now a days.

Most important property of the JFET is that they conduct good current even when Gate and source are at same potential. This was not possible in BJTs which require a bias potential to conduct. JFET current sources without resistors sure are of more fun. However, it is to be noted that current will change with power source voltage change, which is not a very good idea in many applications. JFETs are like variable resistances so they behave as good as resistances. Voltage references are more robust designs and are better than zener diodes and hence are better choice even though may be expensive at times. I have LM199A-883 Mil Grade versions, which are nearly 0.2ppm/'C stable (worst case 1ppm/'C). However these come at great price tag so I use where it really calls for that much stability. It was developed by National Semiconductors but now also manufactured by Linear Technology.

I will be more interested in current sources from 1fA to 1nA range with capability to switch them in 1ns to 1us. I design charge amplifiers in this range. Hence, often have to deal with JFETs. I request the manufacturers of these parts to give me from one production batch, but they insists that I buy few thousands and they also do not come cheap. I like Interfet JFET products, but they are priced real high. I will now try NXP JFETs as they look cheaper. I was using IFN152, 2N4116A, 2N4117A, J309. I will use BF862 (www.nxp.com/acrobat_download/datasheets/BF862_3.pdf) in new designs. I am looking for 1GHz JFETs with gm > 30.

What are your areas of interest?

Dear Dr./Prof. Shyam:

I quite agree that certain low voltage zeners do have low TCs, even very close to zero TC at a certain operating point. Careful matching of operating parameters can achieve low TC over a narrow temperature range. The resistor/diode compensation method I mentioned for the LM334 is specified by National Semi to give a very low overall TC of the current, it appears to match the TC of the LM334 chip closely (of course these components should be mounted close together, even thermally coupled for best results).

Internally, a chip's etched film resistors have lousy absolute TCR, the advantage is that they are all on the same substrate so that only temperature differentials across the chip causes TCR mismatches. To the outside world, the chip can have very low TCs, such as the LM199 you mentioned (I have used the LM399/A often with TCs matching the 0.2 PPM/'C you mentioned. Robert Pease at National and I have talked about this part quite a bit, he said that unless you need such a wide operating temperature range, the LM399/A will often do as good or better than the LM199 and is much cheaper. It is also much cheaper, if you need the tighter TC, to select the parts yourself (Robert's own words) than buying them pre-selected, if you can do that yourself.) Somewhere around here, I think I have some data Robert sent me on the LM399 family curves which was not published in the data sheets, including long term TC drift data.

An example of the mismatching that occurs between the internal chip and the external world components is the input offset adjustment of op amps. The TCR (internally) is approximately 3600PPM/'C, through the use of internal circuitry, the op amp's overall apparent offset TCR using a 0 +/- 100 PPM/'C trimmer is very much smaller. I discussed this with engineers from National Semi at a seminar in Denver, CO once. They were very interested if a 3600 PPM/'C trimmer could be developed to match the chip's.

FETs do make very good current sources, in many ways, better than bipolars. Just depends on the required parameters. Speaking of which, have you tried www.linearsystems.com? Linear Systems makes some of the FETs you mentioned which have been discontinued by others and also some new FETs. I have used them from time to time and they have always been very helpful and willing to work with me. There is a distributer in India as well. I know getting parts in India can be quite a challenge, sometimes here too!

By the way, I have a circuit, using Linear Tech parts and a precision resistor that can produce exceptionally stable constant currents. An op amp LTC1052, the LTC1043 coupled with a stable voltage source and a single precision resistor (and and a film/capacitor filter loop). Granted, this requires a few components but produces a current source that is as stable as the voltage/resistor TC, < 1 PPM/'C possible.

A current source down in the femtoamp range is pretty difficult, lots of leakage issues to deal with. I'll see if I can come up with something for you.

My interests? Kind of a long list, but primarily my work has been in precision wire wound resistors, including the design of equipment which makes them and test them. (Mainly the electronic end of it, I keep my nose out of mechanical unless I need to give input.) The current design, standard part is 0 +/- 3 PPM/'C over the Mil range of -55'C - +125'C, the TCR is essentially flat over that range and linear. Our parts are not like other wire wounds, the stability is unaffected by power or temperature. In fact, you can use them at 5 times rated (125'C) power at 25'C ambient, with linear derating to 125'C. We have essentially a zero failure rate. 0 +/- 1 PPM/'C is not difficult to make for us. Matched TCRs can approach zero. These resistors exceed the specs of working resistor standards but without any power derating. Order turn around can be as quick as five days if the required wire/materials are in stock.

I also do some work in audio, including vacuum tube amplifiers and solid-state, I've worked at two TV stations as a maintenance engineer and designed a wide array of electronics test and measurement instrumentation and controls, a little oil and gas electronics as well and a few other things. Resistors are pretty much my primary work. I'm working on a new resistor standard which can operate at up to five watts (ten watts later) with near zero TCR, a self contained system. I've made some low ohm power resistor standards for General Electric (up to ten watts operating) but those were just a resistor in a case with the 10 watt in oil. Unlike conventional standards, I can make just about any wire wound value just as easily as the common decade values. Something along the range of 1 ohm to 30 Meg. Below 1 ohm, I have to go to a different resistor configuration which I haven't completely finished as yet.

I don't expect the new standard to be all that expensive, less than $5,000 USD, but the price hasn't been fixed yet, too early in the design/prototyping.

I'll get back to you as soon as I have anything on the switchable current sources.

Ed

Dear Electronic Wiz

I think, your work in precision instrumentation is very valuable.

You reminded me of Robert Pease (Bob Pease), the man of great ideas, who has been with NS for so long. I think he was also with Teledyne Philbrick, the company which used to make modules for DAC, and Amplifiers using BJT on ceramic Hybrid. That was some time near 1970. I have used lots of NS ICs over the years.

As you have suggested that LM339 will be good enough and at low price, I will try this one for new applications. Lots of people are asking 1ppm/'C stable variable voltage sources and also control and measurement ICs such as DAC and ADC better than 20-bit is becoming a challenging area for designs. These go with high voltage power supplies for me and getting 1ppm/'C potential divider resistors for 20kV is very challenging. Some times people want to use 10-turn pots for adjusting power supplies, which again is difficult as <50ppm/'C 10 turn pots are to be specially designed and bound to be very expensive. Current sources, voltage sources and their measurement will remain highly challenging field and it is almost like an art to achieve a bit of extra edge each time. People started using thermo electric coollers inside the amplifiers for lower electron noise. For portable designs this is again a difficult thing to implement due to lots of power loss.

I will be very much interested in variable resistances and resistance dividers that have low tc. 1ppm/'C resistances for amplifier gain stability sure are of great use.

Dear Dr. /Prof. Shyam:

Yes, Bob has been around for quite awhile, a little longer than me in fact. He worked for Teledyne Philbrick when they were using vacuum tubes. That company was responsible for some really great tube circuits. I think Bob was there in the early 60s or late 50s (I'll have to ask him).

I'll look around for the long term drift data Bob sent me and I'll give you a summary. I think the LM399A should do the trick for you. One other thing, Bob said that running the LM399 for one or two hundred hours or so before putting them in a circuit will stabilize them better. Basically, that's what they're doing for you with the MIL grade and charging a lot of money, plus a little screening of TC on the side just to make sure. If you don't need the temperature extremes, you are wasting money.

I specialize in tight tolerance resistors and low TC. The main limitation with wire wound resistors is the voltage limit of the wire's enamel, this has generally been accepted at 150 volts per section (or pi as it is called). This limits the upper voltage to 900 volts in a six pi bobbin. Now that I mention it, I have not verified that our parts have this limit as we specify a particular enamel for our wire. The enamel coating must be very thin and this limits how much voltage can be applied without causing a puncture (holes can form in the enamel after application to the wire). Our supplier is the best in the business.

Film resistors made for higher voltages have a higher limit of TCR inherently, they're trying to bring it down but so far, 25 PPM is the best they can do within a limited range and those aren't too easy to find. Unfortunately, wire wounds do carry a size penalty, the smallest wire diameter is 0.0004" (~ 0.01mm) and that small size costs somewhat more. Of course, that is offset by its characteristics, but we are limited in how high a value we can put in a given bobbin size.

I've got a wire wound trimmer design that we haven't gotten to as yet, it is fairly small, screw driver adjusted (at least this version), has the same TCR as our resistors and essentially infinite resolution. It is only in a prototype phase.

May I inquire why do you need to switch the current source so fast and is this a rise and fall spec with a length of time on and off or is this a pulse width spec? How accurate does the current need to be? If you are wanting to provide a pulse width of 1μS at 1fA, that is only 0.01 electrons and cannot be done. Most FETs have a charge error of more than 10 electrons and then there is the question of leakage currents from everything touching the circuit. If you could supply some more details I could perhaps give you some guidelines as to what can be done.

Thank you very much,

Ed

Dear Electronic Wiz

Lowest DC current I come across in the design is 0.1fA (1-10Hz) and 1pA (100kHz to 1Mhz) and 1fC and above up to 1pC for charge pulses 100ps to 10ns range. For different time domain, there is need for different levels of current and charge measurement. At least 10 electrons are required in the charge pulse at the lowest limit to be properly accounted for.

All this is for measurement of charge particles in time of flight and also small charge produced by ionization.

Hi cnpower

I was discussing with Electronic Wiz who works with world's most experienced man in electronics Mr. Robert Bob Pease alive today. I knew Bob's abilities in 1974 and he never faded in so many years. All Linear ICs from National Semiconductor you may have used must have come through his vision or cooperation with team mates. Bob when was about 40+ thought that he may also die early as one of his friend died at that time. I feel good just to hear his name and can come all the way to say hello to him. Electronics Design reserved special page for him. I used his designs of 16-bit DAC when he was at Teledyne Philbrick and these were actually used for Radar Screen displays for USA army. He designed current mirrors and current switches of high precision (16-bit dynamic range using PNP transistors) for making DAC and we are still talking about these single current sources.

There are many ways. You can measure very small current by counting the electrons or collecting them in a capacitor bucket and then emptying them fast to get more current in short time or just measuring the collected charge. For me current is not always in wires, it is some time flow of charge particles in space at speed, random distributed or in bunches. Some time it is drifted charge in semiconductor.

If you can get rid of tribo-electricity in wires, then you can measure current down to about 10aA with cooled FET with electron noise of that level (~+/-10 electron). Use discrete Cool-FET and not any IC for such measurement. You may also need lots of proper shielding. If you have more time then you can measure more accurately.

also see www-mtl.mit.edu/research/annual_reports/2005/ar2005cs_all.pdf