LED reference current source

Below around 20K, you get "carrier freeze-out" and the junction drop goes way, way up. Dynamic impedance goes up too, so a diode gets resistive. I recall numbers like 2 volts drop at 2K, 10 uA bias current. Repeatability from part to part goes to hell, too.

I wanted to make a cryo temp sensor from a vacuum-packaged tuning-fork quartz crystal. Even got some parts made. But I couldn't convince my customer (Jefferson Labs) to test it for me, so we stuck with diodes.

In the late 70's, this was a fairly popular method of developing fixed bias, where the tempco was not considered important. The thing was that different colors and chemistries of LEDs had different voltages, usefully larger than Veb of bipolars, were easily available and had a small footprint. It was considered a plus by some users, that there was no actual part number stamped on them.

This use may have been popularized in Elector.

The fact that there was a visible indication of circuit function wasn't lost on most users. Noise reduction was something that could be an issue in some circuits, as with any low voltage two-terminal reference, at the time. There were not a lot of them.

I used it to bias a high current, low rbb', discrete bipolar amplifier stage for low source impedances, in 1980. Remember those? Still have it in a box somewhere. As I recall, the resulting tempco from ambient was negative, using a common green TI 3/4 part. This was also considered as an advantage, in some cases.

RL

National makes an LM35 in a TO220 package. We mount that on a heat sink with a bunch of power fets.

yeah, it's best to treat them as a directed graph, one strategy is to leave a trail of breadcrumbs.

John Lark>

ny

y 5000

t in

"Not very often" wouldn't violate any non-disclosure agreement. Neither would "most of the time", although it might damage your credibility.

-- Bill Sloman, Nijmegen

but

The disadvantage is that it doesn't do the same job.

I got them all onto a 2" diameter printed circuit board in the weighing head of the Cambridge Instruments/Metals Research GaAs crystal puller, using through hole components back in 1986. It included my improved version of the Baxandall Class-D oscillator to generate quite a good clean sine wave with feed-back controlled amplitude (so two demodulators - in one package - and voltage reference but only one RM6 core).

Nobody bitched that I'd used up a lot of layout time either.

How on earth do you drive and demodulate your LVDT's and synchros? Buy in other peoples modules?

And you are recycling half-baked ideas. What on earth is "factual" about "If its *resistance* has a 1/f component"?

You could learn a bit more about electronics. Phase-sensitive/ synchronous detection isn't exactly rocket science, but you don't seem to have got to first base in that area.

-- Bill Sloman, Nijmegen

As opposed to merely advantageous.

It's a pity about that - you've just made a very silly statement in a public forum where you can't unsay it, and if I feel the need to remind you of it, I can use Google to retrieve it for the foreseeable future.

-- Bill Sloman, Nijmegen

Doesn't bother me. Your case for using AC has been, so far, just insults, no substance.

It's a conditional statement. What's wrong with that?

Everything I've said is both logical and true. If you have a substantive objection, say it. If all you have is lame insults, you're acting like you usually do.

The only 1/f error here is in the amplifier, and chopamps don't have

If the RTD resistance has 1/f noise, that modulates an AC system just like it drive a DC system. No difference.

Say something substantive.

More lame insults, no substance.

One problem with phase-sensitive detection is that it's phase sensitive.

an

of

OK I've been picturing something similar, but with the flex soldered to a little metal 'washer' that I can put a little screw through. (I hadn't thought about using indium solder... thanks) I know I'm adding mass, but a thermometer with a hole seems so useful. I think the washer could be pretty thin, though I haven't done any calculations

re

ably

heir

Grin and assume linearity, (always a good first guess)

George H.

in

Lots of people like you are in the unemployment line. No explanation required.

I am sure most of them are calculating the phase margin error of their synchronous spinning thumbs, while waiting in line, in hopes some one at the desk falls for that malarkey.

Jamie

But you demodulate after you've amplified the AC signal.

Larsen N T 1968 Rev. Sci. Instrum. 39 pages 1=9612

went to the trouble of band-pass filtering the the amplified signal before demodulation - IIRR he threw in all-pass elements to avoid temperature dependent phase shifts in the filter.

Any 1/f noise from the demodulator really isn't going to be a problem. Duty cycle sensitivity would be a problem if you weren't careful to get the duty cycles right, but that isn't rocket science. If you want to go nuts you can lock a 4046 to some multiple of your modulation frequency and divide down from that, but the last seriously accurate system I put together worked fine with comparator-derived demodulator drives.

But you can't actually avoid thermal gradients, though you can hope to make them tolerably stable and more or less orthogonal to your measurements - as the paper that Gerhard Hoffmann dug out makes clear

Jim Thompson may know about this sort of stuff - analog integrated circuit designers started talking about symmetrical layouts about thirty years ago.

-- Bill Sloman, Nijmegen

You mean, you use goolge most of the time and understand what you find only some of the time. A very small part of the sum, that is.

You don't do very well at covering your tracks. You can brush away your foot trails all you wish however, your stink gets you found every time.

Jamie

but

't

It's certainly more complicated - I wouldn't have said much more complicated, but I'm more comfortable with complex systems than most - but minimising thermal gradients isn't a walk in the park either.

I got a whole oscillator/amplifier/demodulator set-up onto a 2" diameter board back in 1988, using through-hole components and without making any great effort, so "complicated" didn't mean big.

-- Bill Sloman, Nijmegen

in

but

n't

It should.

My case - as far as you can understand it - is just insults, no substance? You are good at feeling insulted when people don't agree with your opinions, less good at appreciating technical content.

My case is that the output from an AC excited bridge is an AC signal at the excitation frequency.

Phil Hobb's case is that the output from a DC excited resistance bridge is close enough to perfect that it's good enough to use a chopper amplifier to amplify both the DC imbalance signal and everything else that happens to be there.

You don't actually put it quite the same way, and in fact seem to think that the chopper in the chopper amplifier is doing exactly the same job as the AC excitation of the bridge. Saying that probably is insulting, but I can't understand the stuff that you've been posting any other way.

-- Bill Sloman, Nijmegen

ut

't

No. I didn't mean that, and I most certainly didn't say that, or anything like it. Your reading difficulties are leading you astray again.

-- Bill Sloman, Nijmegen

but

Don't be silly. I'm 69 and retired (against my will). I'm not allowed to join any unemployment line.

As usual, your vivid imagination - uninhibited by the facts that you could read, if you could read reliably - is leading you astray, not for the first time.

-- Bill Sloman, Nijmegen

The AC measurement of (for instance) an RTD doesn't require any fancy sources, just an oscillator and some synchronous rectification. The sensor is as small as you would care to make it, a four-wire (Kelvin) resistor. By not putting amplifiers inside the 'oven' (it could as easily be a cryostat) you can make the regulated zone smaller, which is a cost advantage.

Most important, though, the effective noise bandwidth is nil. The usual ten minute warmup time for ovenized devices implies a few minutes time constant; call it two minutes. A 1 kHz exciting signal piped to the sensor, goes with a synchronous rectifier and one-minute integrator to make a noise bandwidth (-3dB) of a few millihertz, centered at 1 kHz, away from DC offsets, popcorn noise, thermocouple effects .

If you care about self-heating in the sensor, that's important, because the exciting signal has to be kept small.

Hmm, more like testicle content from you.

An excited bridge in any form has no output until it becomes unbalanced. I don't see where AC has anything to do with it? Unless we are talking about inductive linkage, but then we'd be talking about things like LVDT's and such and I don't think you would be going there, it's beyond you.

And what would that everything else be? DC current usually works well in eliminating Hi-Z issues. All you have left over is low frequency noise, noise you get at 0 Hz and that really has nothing to do with the bridge. Inserted AC with in the circuit after the bridge works better than putting AC in the bridge, since AC in the bridge brings in a whole new set of problems.

Yes, you finally made a correct admission, you don't understand.

Thank you for sharing that with us.

Yes, I do know something, it may not agree with you, however, what I know and understand seems to work when I apply it.

Have a rotten day and may it rain hard on your bald head and give you waterdrop syndrome.

Jamie