LED reference current source

cryostat)

--
So you say but, as usual, you present no proof, just lip service about
what you say you accomplished in your salad days.

Got a schematic and a PCB layout we can peruse?

[about four-wire resistor with AC remote excitation]

How can that be? If the 100 ohms of the RTD is small, the 1 ohm of the wiring is smaller. I'm thinking of triangle-wave excitation, with the phase-sensitive rectification exactly cancelling the (square wave) capacitive currents. How can that square wave not be phase-locked 90 degrees to the triangle?

That's probably a good idea, it'll take a voltage reference and some other items as well. An RTD is good for temperatures -220C to +550C, though. Op amps, not so much.

I'd like the LM26 for the thermostat parts count winner, though. It'd be a pretty strange quartz crystal that couldn't be trusted without millikelvin absolute temperature accuracy.

What exactly? You haven't mentioned them before. I can't think of any. (And thermocouple offsets go faster than 1/f, typically 1/f**2.)

Getting rid of the Johnson noise of the other arms of the bridge gets you a 6 dB improvement at most, at the cost of a lot of very fiddly magnetics. I'd rather use four sensors instead, and get the same improvement at far lower cost, with spatial averaging thrown in as a fringe benefit.

In the 1940s, the state of the art was _tubes_, and choppers were mechanical. Remember tubes? I still have a few tube-based meters, purely for coolness' sake. My Keithley 405 tube electrometer has a 100 fA full scale range, but takes at least two hours' warm-up before the meter even gets off the peg on that range. My Simpson VTVMs drift all around the shop. For low level measurements, tubes _suck_. Mechanical choppers are slow, jittery, and wear out rapidly.

It's not just size, as you'd know if you'd actually read that OPA378 datasheet I keep mentioning. Modern chop amps are amazing, and if you put them inside the controlled volume, all the other sources of error go away to the same accuracy as the thermal forcing rejection.

You keep ridiculing people for clinging to legacy parts like the 555, and here you are doing it yourself.

Bill, I hate to break it to you, but John has designed instruments a great deal more complicated than your bits of your instruments, and a much greater number of them as well. He's designed a good deal more than I have, too--that's his business, after all, and he's been at it a long time. You might not like that, but it's true.

Not really. Anything that needs really accurate temperature control is doing something hard, or they wouldn't bother.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

Really easy: make a bridge from two 1K thinfilm RTDs and two 1K thinfilm resistors. Power it and an LTC2053 from a 3 volt supply. The diffamp input is 6 mv/K and each RTD dissipates about 2 milliwatts; use less voltage if that's a problem. Set the gain to whatever you like. Epoxy the RTDs in strategic places near the thing to be controlled.

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

ny

f

nd

ith

the

e sensor

istor.

cryostat)

No. They belong to Metals Research. They were a subsidiary of Cambridge Instruments when I did the work, but got floated off and went bust over the next few years. This seems to be the successor company

You could approach them. The circuit replaced the original LVDT driver and demodulator, which had been designed a decade earlier, based on parts which had gone obsolete. I believe a few of my circuits got sold as up-grades to older machines - I'd like to claim that this was based on the superiority of my design, but in fact the real charm of my circuit was that it didn't contain any 741s and didn't - consequently

- inject pop-corn noise into the system. A few uV of pop-corn noise isn't usually noticeable, but in this application it used to turn the

30kW RF induction heaters on and off every minute or so, whereas the OP101 I used was stable enough that the induction heaters ran more or less steadily at about 10kW.

The single crystal GaAs that the crystal puller produced may have had fewer defects in consequence, but if it did, nobody ever told me about it.

-- Bill Sloman, Nijmegen

e sensor

istor.

cryostat)

't

blem.

If the out-of-phase component is big. It wasn't in my application, and I actually delayed the demodulator drive by about half a degree, half way between the optimum for demodulating the oscillator output (for feedback control of the amplitude) and the optimum for demodulating the amplified output from LVDT which lagged the oscillator output by about a degree.

-- Bill Sloman, Nijmegen

e:

are

e in

p, but

at

ter

't

or

hasn't

rd.

d

or

trol

ed.

-

p,

d

one

d

The Blumlein bridge gives better sensitivity and less power dissipation

See the new thread "AC excitation of RTDs and other resistive sensors" for more detail.

But the literature can make up for the defects in your imagination, if you know it well enough.

A twisted pair wound onto an RM6 core it "fiddly"? Were you frightened by a transformer as a small child? Many electronic engineers seem to have suffered from this trauma, but it's unexpected in somebody with a physics background.

You can still have two sensors - it's not a Blumlein bridge any more, and it's tricky to trim the inductive arms to take out the tolerance between the sensors, but the configuration does have its charms.

So what. The mathematical analysis of the bridge performance gave exactly the same results then as it does today.

Modern chop amps may be amazing, but they only minimise errors arising inside their package. That they are compact may let you reduce - but not eliminate - other sources of error.

A bizarre misconception. The Blumlein bridge was a good idea when Blumlein invented it, and it will remain a good idea long after he's been forgotten. The 555 was a great idea when Hans Camenzind put it together in 1971 but the combination of a timer and a rather crummy power switch wasn't anything like as universally useful ten years later and technology has now made it a niche part.

He has more designs to his credit than I have, and many more that made it to production. On the other hand I've certainly designed - or more precisely, served as hardware systems architect for - rather more complicated systems that he's ever built, or seems to be able to conceive.

Have you any idea how complicated a shaped beam electron beam microfabricator actually is? Not that the Alvey machine ever go anywhere close to production. We did publish a few thoughts on its data buffer - a couple of Gigawords of error corrected RAM memory - amongst the numerous other sub-systems for which I was responsible.The laser interferometer based stage positioning system was hairy, as was the hardware that translated stage position into coordinates on the integrated circuit we were writing - which couldn't be relied on to be exactly aligned with the axes of the interferometer as it whizzed under the the beam.

The multi-flash stroboscopic electron beam tester did get as far as a fully working prototype, but marketing weren't convinced that we could sell enough of them fast enough to keep our cash flow right, so they canned the project after we'd spent three years and a few million quid on getting it to work.

I wrote the original proposal - based on Gigabit's GaAs logic - primarily as a way of showing up my boss's insistence of 10psec time quantisation as impractical. Unfortunately, it wasn't impracticable, and management was taken with the idea of building a machine that would be hard to copy, without thinking too hard about what was going to make it hard to copy.

The 1996 paper describes the temperature controller we built to go with the IASys biosensor unit

Cush R, Cronin J M, Stewart W J, Maule C H, Molloy J and Goddard N J 1993 Biosensor Bioelectron. 8 347=9653

I didn't have to design the electronics for that - merely correct the

- numerous - inadequacies of the electronics that someone else had designed, after he got a lectureship in Manchester and made himself scarce. The temperature controller was a less substantial piece of work, but rather more than a nit in comparison. And we pretty much got it right first time, which certainly wasn't the case for the IAsys electronics. I had to substantially rework the original circuit to make it work at all - the original analog front end was really very shoddy - and once we had enough working machines to do serious testing we learned enough to design it properly (though I'd been dragged off to the Netherlands by then, and only contributed the layout of the new analog front end, separated off onto a plug-in daughter board.

-- Bill Sloman, Nijmegen

but

hasn't

At DC, there's no issue, and you get the same sensitivity.

I'm aware that mathematics is not time-dependent. Engineering tradeoffs are, however. The amps now are, roughly, 120 dB more stable, 50 dB smaller, and dissipate 40 dB less power than the ones Neubert knew about back in the '40s. That makes the tradeoffs very different now.

You don't seem to understand the point about being able to put the amp and all the associated components and wiring inside the controlled zone. You can't readily do that with tubes, so the problem domain is entirely different.

A billion chips a year, last I heard. I'd love to have a niche that big, myself. ;)

Not so. Being a "hardware architect" or systems engineer is not the same as being an instrument designer. You draw boxes and parcel out the actual hard work to the slobs doing the actual design work. If you do it right, the pain is roughly equalized; if not, they come after you and get the architecture changed. The best designers can do the job of everybody on the team--probably not as fast as they can (assuming they're any good), but at least has a grip on every part of the system.

Teams too big for that to be true are very much more difficult to manage, and require a lot more documents and bureaucracy.

I'm sure it was a pretty whizzy device. How much of that did you actually design yourself, i.e. do the calculations, drawings, code, parts selection, layout, ...?

To paraphrase one B. Obama, "You didn't build that"--not by yourself. Or did you?

Writing a proposal isn't the same as designing a whole complicated instrument, Bill. Otherwise they would only have needed you.

You're proud of it, which is quite reasonable. However, it isn't the sacred scriptures of thermal control--technology moves on.

Today it would be replaced by a chop amp, a sensor, four resistors, and a capacitor, all in less than 1 square centimetre, inside the controlled volume, would work at least as well, and nobody would think it was anything unusual.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

in

but

hasn't

control

The electronics of building a really good oven has become pretty easy. The mechanical/thermal design hasn't.

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

Agreed. I did one last year, which is hopefully going to get built soon, assuming the customer (a start-up) gets some more dough. Managing thermal diffusion and heat leaks were the main issues. Kapton flex circuit really helps with both.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

Lark>

You don't. The voltage drop across the two transformer windings doesn't change at all as the impedance of the sensing resistors changes, so you get all the voltage change across the resistors at the output. rather than a fraction - twice as much as you get with nominally equal impedances around a Wheatstone bridge. That's not the only reason for using a Blumlein bridge, but it's a real advantage.

nt

n

The Blumlein bridge has always been more sensitive, and the inductive arms don't introduce any Johnson noise. That's not something that changes with technology.

u

go

You don't seem to understand that you can put a Blumlein bridge and all the associated components inside the controlled zone. I got them onto a 2" diameter board more than 20 years ago, using through-hole components.

The 2" diameter was set by the bought in slip-ring that had been designed in a decade earlier to carry the power that operated the rotating LVDT-based weight sensor and take out the weight signal - I hadn't gone to any trouble to minimise the board area I used.

The single RM6 core was a bit bulky - you might be able to get away with something more compact - but surface mount components do take up less space.

The components of an AC bridge will take up more space than the front end of a DC bridge, but it's not a dramatic difference. And getting rid of the sensitivity to thermocouple voltages takes away a lot of motivation to put everything inside the controlled zone. John Larkin is frightened by the electromagnetic interference issues involved in exciting an AC-bridge remotely, but shielded twisted pairs work pretty well. The one that carries the - balanced - excitation current won't radiate much - and the one that carries the output won't pick up much noise

Legacy parts. Stuff that got designed in by good engineers from 1971 to about 1980, and stuff that got copied by less good engineers - or good engineers in a hurry - after that.

a

Not the way I did it. I knew what each bit did, and how it did it. For the shaped beam electron-beam microfabricator I had to design one of the chunks myself - one of the engineers working for me got stuck and I spent about half my time for about a fortnight blocking out his board on a sheet of A1 on a drawing board close to my desk. It wasn't a particularly detailed design - it just specified what got buffered where, and what the timing signals looked like - but it was enough to get him going. I could have completed it - I'd done the detail design on more completed stuff before then, and I did ti again later.

For the electron beam tester I was only formally in charge of the hardware design for the first year - until the project manager's habit of lying about the project time-scale caught up with him. I was one of the many people who pointed out that we had a much better project manager around the place - at a loose end at the time - and was very happy when he took over. I was less happy that he insisted that his hardware engineer was put in overall charge of the system design- his design engineer was a good and competent hardware engineer, but not as good as I was, nor as experienced in using really fast logic - but it was a price worth paying.

In practice, I remained the technical authority - I got on well with all the people involved, and was happy to advise when it became necessary and while ostensibly deferring to my "superior's" opinions (when he was brave enough to advance them, which wasn't often - I had mis-estimated the propagation delay through part of the analog processing, which did create some problems, but mostly I'd got it right). I was no longer stuck with the administrative crap, like working out the current flowing through the various ground connection

- which my architecture had deliberately minimised - and spent most of my time on the detailed design of a couple of trickier board, with time off for occasional trouble shooting.

What was odd, was that I'd had to produce a weekly progress report when I was in charge of hardware design, which I'd e-mailed to my boss and the people I was supervising (in part to keep myself honest and in part to keep everybody informed). I kept on doing that after I'd been formally demoted, and ended up polling the software and mechanical engineers for their progress as well. It eventually took half a day out of each week, but when I talked about it to the new project leader, he insisted that I kept it up. Apparently it did good things for team morale.

IBM could support that kind of team. Cambridge Instruments wasn't that big.

Not a lot. I had quite a lot of influence on the parts selection - but the guys who were doing the detail work had the last word, and the time to keep up with what had just become available. Layout was always handled by the drawing office, but the engineers got to review the layout pretty frequently. It didn't stop the drawing office from scrambling the order of the inner layers on a board where this messed up the transmission line impedances on one famous occasion, but mostly we got the board layout we needed. I got to sit in on a lot on the layout reviews - I was actually pretty good on layout.

Significant chunks of it. see above. I was degraded to spear-carrier halfway through the project and spent a lot of my time cleaning up boards where other people had done the detailed design.

About ten of me. I could write that proposal - and get it right - because I could have designed every bit of that hardware. Writing the software would have taken longer, and I would have had to have learned modern programming techniques - I knew about them from sitting in on numerous software reviews, but that's not the same as having developed the right habits by doing it.

l

l is

Have I ever said anything that implied that it was any kind of sacred scripture? It was worth publishing at the time - the use of a 20-bit sigma delta A/D converter in that application hadn't then been documented in the literature, and the advantages of digital signal processing still aren't always obvious to the people who are publishing today - but it was more a snap-shot of what you could do with 1990's parts than any kind of revolutionary break-through like the Blumlein bridge.

You don't seem to have read the paper - I'll send you a reprint if you can't get hold if it.

It did rely a dead boring resistive bridge, which delivered ten times the performance that we needed for the application. A chopper amp wouldn't have improved the performance in any useful way - I'd have used one if it could have. Paul Buggs subsequently edited the circuit to turn it into a temperature controller for single mode laser diode when we finally realised that we had to go that way. I've no idea what he took out - most likely some of the filtering.

The unique selling points of the design are listed in the introduction to the paper, and the bridge configuration isn't one of them.

-- Bill Sloman, Nijmegen

:

Lark>

But they can include features - like an AC-excited bridge - that you clearly haven't mastered

Nor getting the electromagnetic compatibility right. That you are scared out of your mind by the prospects of coping with this aspect of the design an AC-excited bridge doesn't say anything good about your competence in this area.

-- Bill Sloman, Nijmegen

usual

e

e sensor,

But the thermocouple voltages generated where you go from platinum to copper to lead frame do generate noise that is more or less 1/f. Compact design can let you see less of the thermal gradient producing the thermocouple voltages, and careful layout can minimise - but not eliminate - the gradient that the junctions are looking at. AC excitation gets rid of it completely, and lets you use the slightly more sensitive Blumlein bridge, whose inductive arms don't generate any Johnson noise either.

So what? I can't remember a design where we were space-constrained in a way that would have allowed a TSSOP (Thin Shrink Small Outline Package) - packaged chop amp to do something different. I once designed bunch of FET chips onto on a thick film hybrid, which is the old-fashioned way of getting a quart into a pint pot. They worked, but it didn't do as much for us as we'd hoped.

-- Bill Sloman, Nijmegen

Lark> > > > >>>>>>>>>>>>> On Wed, 12 Sep 2012 19:09:28 -0400, Spehro Pefhany

Lark> >

I understand perfectly. Probably pretty good going for the time, but no longer highly relevant.

Any transformer whatsoever dwarfs the circuitry needed for the DC approach, besides being a fiddly custom part.

Okay, assuming that the other members of the team would agree to that description, what happened to you? The way you put it, it sounds like you walked on water back in those days. Not so easy to believe, based on recent performance.

Okay, so it sounds like you can't take credit for much of the actual design.

Hmm. And you walked on water before that, it seems. Any idea why?

Okay, so you can take maybe 10% of the credit.

Yes, you have, by blowing off arguments that show that it's no longer the way to go in general. In particular, you've repeatedly ignored the point about how putting all the circuitry inside the controlled zone changes the problem completely, by eliminating the thermocouple offsets and bridge-resistor drift. Anyone reading your posts can see that yours is the One True Way.

And the Blumlein SNR advantage, while real, is only about 6 dB, and you can get that back by using two RTDs in diagonally opposite arms of the bridge. Give me that over some fiddly ultraprecise transformer that there's no way to test.

A temperature control loop isn't exactly news nowadays.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

Lark>> >> > >>>>>>>>>>>>> On Wed, 12 Sep 2012 19:09:28 -0400, Spehro Pefhany

Lark>>

You are nothing but an insult machine any more.

The first AC bridge design that I did was about 1971. It was to excite and signal condition some Talyvel inclinometers, which behave like a cross between a pendulum and an LVDT. This was part of the Boresight Alignment Kit for the C5A military transport. I also did the HeNe laser power supplies and some quadrant photodetector stuff, all to mil spec standards and testing.

It all worked fine. The Talyvels didn't meet the required linearity spec, so I had to include some breakpoint trimpots. I had a barrel full of sand as a base, with a huge steel plate in top, and a micrometer/lever thing to make calibrated inclines. Nobody was allowed to walk around for 30 feet or so in all directions when I was working on this; they would flex the concrete slab of the building.

I also did the Winch Control Intercommunication Subsystem for the C5A.

As noted, I've done a lot of LVDT and synchro stuff recently, heavily FPGA/DAP.

Why are you so nasty, and so wrong?

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

ote:

ohn Lark>

y p

is

o

On the contrary, you are carrying on as if smaller area occupied by the components needed to activate a DC bridge is magically smaller than that occupied by the components required to activate an AC- excited bridge.

It will make a crucial difference in some very specific applications, but we aren't talking about a gross difference.

This isn't exhibiting "perfect understanding" on anything that looks vaguely like it.

The board to fit right next to the resistance sensor is already a fiddly custom part. You are straining at a gnat after swallowing a camel.

I was doing what I'm really good at - designing rather complicated innovative systems and getting tehm to work - in a company that was in the business of selling complicated and expensive systems.

These aren't skills that I can showcase here. John Larkin talks about spending two weeks on a new product - which doesn't give much time for innovation or complexity.

I wasn't in the business of taking credit. There's a saying that you can get pretty much anything done if you don't need to take credit for it. If you can persuade people that they though of something for themselves, they get very enthusiastic about turning it into reality. Earlier in my career I'd noticed people trying to do that to me - which didn't worry me, since I'd felt more flattered than manipulated

- and I've kept it in mind since.

a

uld

y

uid

Sure. I'd warned everybody that the project was going to be a bitch - practicable rather than practical - but we'd just spent quite a lot of money and effort on buying CAD tools and adjusting our design practices to take advantage of the investment, and I had qualified my warning with the observation that a couple of boards designed under the new system had required very little debugging.

What happened was the the project leader and his boss decided that they were going to try to get a working model of the machine to to some exhibition in the US in less than a year. According to the project planning it wasn't impossible, if everything worked out. We farmed out work to sub-contractors which meant that my feet didn't touch the floor for the next six months, and I didn't get the time to notice board designs were going out to layout without having gone through design review - I actually got into a state where I was relieved that I didn't have to find the time to sit in on the design reviews.

Then the shit hit the fan. The project leader left. His boss got left with egg all over his face. I should have had enough sense to blow the whistle before the shit hit the fan, but I'd been too busy to think about what was going on, so some of the egg ended up on my face as well.

e ,

ng

Rather more. Nobody else in the company could have written that proposal, and much of the detail in it came from my odd background. I was the only guy around who'd used ECL for anything except line drivers/receivers, and the only Gigabit Logic databook in the place was sitting on my bookshelf before we bought up what was left of Lintech - the first electron-beam testing company - and hired it's ex- boss as our new technical director

he

It doesn't eliminate thermocouple off-sets, just reduces them.

Dratler did it a long time ago - he published in 1974

Dratler J 1974 Rev. Sci. Instrum. 45 1435=9644

and his approach wasn't widely adopted, cited, or imitated.

Good grief. Are you turning into John Larkin?

If you knew much about transformers, you'd realise that a bifilar wound centre-tapped winding isn't in the least fiddly, and gets to be ultra-precise essentially because physics is on its side, rather than because you've sacrificed the right number of virgin graduate students to the right deity.

And such transformers can be tested against one another without too much difficultly. National Standards labs love them, and that's a group of obsessive-compulsives who test everything that does anything.

Whoever said it was?

Rev.Sci. Instrum. allowed Libbrecht and Libbrecht to publish - in 2009

- what was essentially a copy of the 1991 paper by Bradley C C, Chen J and Huet R G 1990 Rev. Sci. Instrum. 61

2097=96101, with one added extra mistake, so the physics community doesn't seem to share your opinion.

-- Bill Sloman, Nijmegen

His dad was Don Rickles, and his uncle was Rodney Dangerfield.

He didn't inherit their talent or their humor. Or their wealth. He's just sour and nasty.

--

John Larkin Highland Technology Inc

jlarkin at highlandtechnology dot com

Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom timing and laser controllers Photonics and fiberoptic TTL data links VME analog, thermocouple, LVDT, synchro, tachometer Multichannel arbitrary waveform generators

He got what was leftover from both. He's nasty, and he doesn't deserve any respect.