Favourite Class D amp for instrumentation purposes

Suggestion: be really, really paranoid about putting switchers anywhere near the low-level stuff. There will be lots of coupling modes, including ground loops in the copper planes, conduction, capacitive coupling, magnetic coupling, and sneaky capacitive coupling from parts to the tube and back, essentially wiggling the entire ground plane.

Consider using some integrated switchers, like the LTM8023 sort of things lots of people are making lately. They are very tight and clean. One of them could drive your TEC maybe. Does it have to heat and cool?

John

It needs to do both, but probably not at the same time. I'm planning to put it in its own steel can, separate from the rest of the innards.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

ent

al

d

ck

nt

f

SMPS.

t

With the disclaimer that I never drove a thermoelectric cooler, why couldn't you just roll together some PWM scheme. A class D chip should have lots of engineering toil regarding THD and stuff, but your application doesn't sound like a high fidelity requirement. If you switch your PWM fast enough, the thermal mass should smooth out the pulses. You wouldn't even need an inductor.

If you want to avoid analog current limiting scheme, you could build what amounts to the guts of a step up boost converter with discontinuous conduction. Just charge up a coil to the imax limit then dump it. Use pulse skipping to regulate the temperature. It might even be possible to hack a step up chip to do this. The disadvantage to this scheme is you would never reach 100% of the cooling capability. With discontinuous conduction, you could limit both V and Ipeak.

If you wanted maximum cooling and efficiency, maybe you could adapt a hysteretic battery charger chip. The older NiMH chips were designed to deliver a set current with an overvoltage limit.

Do you really require an AC waveform?

TI got a bunch of single IC Class D amplifiers in your power class. Pick the one you like. An ICs of filterless Class D look better candidates for your application. This is about as simple as it gets.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

P.S.

Q: Why it is impossible to have sex in the middle of the Red Square in Moscow? A: Because every idiot bystander will be offering his invaluable advice.

I

rrent

as

ical

it

and

of

back

vent

off

he SMPS.

g

net

Wrong. Peltier junctions have a resistivce loss, which rises with the square of the current. You want to keep any ripple down to 10% or less.

My 1996 paper - and I think Phil already has a copy - has a detailed write up of our class-D drive scheme, and an equally detailed description of the filtering arrangements, which were put together from exactly the point of view that John Larkin presents, and proved to be effective enough that swapping back to a linear drive (as used on the proof of principle prototype) didn't make any difference to the performance (+/-0.001K).

-- Bill Sloman, Nijmegen

Ideally, I want an efficient bipolar current source of 2A or so, with minimum board space and very very low ripple--the ripple will get into everything otherwise. The laser wavelength has to be stable to about

100 parts per trillion over times of many hours, and any switching crud will be a big problem--even if it's only on the TE cooler.

Everything has to be calculated, or at least estimated to 1 significant digit--sensitivity to pressure, temperature, and thermal gradients; stress birefringence in the glass; fibre misbehaviour, you name it.

Also of course the schedule is _nuts_, so paying an extra $5 and saving a square inch and three days' work is a huge win. I'm leaning towards the higher power chips that can work off a higher voltage supply--that means I can use smaller conductors in the flex, which helps with the temperature gradients.

So any actual experience and hopefully part numbers would be a big help.

Thanks

Phil Hobbs

(On an airplane to Vancouver)

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

Eight inches? That's huge.

That's at odds with your 4V TEC voltage. Can be done but I think you'd be better off using a small switcher up front to bring that down.

If you can bring down the rail voltage:

Have used it, works. The datasheet is not so great and I found an error in an equation, not sure if they corrected it by now. One downside is the price tag. There used to be an eval board, maybe they still have it. Oh, and as John pointed out you may have to filter the output really well.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

That's including the optical system and the laser, plus two layers of thermal control. That has to house the actual laser controller, between three and nine temperature control loops (I'm not sure how many yet, that depends on the calculation results), monitoring and self-test stuff, a two-output SMPS, and the TEC controller. I could use a lot more space than that!

Price is not a big problem--downhole instruments have a short and very difficult life, and there's at least $1k worth of optics and another $500 worth of laser inside. Schedule, now that's a problem.

Thanks

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

ent

al

d n

ck

nt

f .

ll.

if you can make a 5V supply, and do with +/-1.5A I don't think it gets much smaller than:

-Lasse

d I

current

has

ptical

fit

, and

s of

e back

revent

t off

the SMPS.

ing

to

l.net

Then easily solved with a hysteretic converter. Just more complicated than simple PWM,

I

rrent

as

ical

it

and

een

e

of

back

vent

off

he

d

or

it.

well.

That looks pretty slick. Bidirectional too, i.e. heat and cool.

I'd probably be more worried about noise leaking over via the power rails and suboptimal ground connections than via the TEC lines. If you want to go totally linear I can dig out a source but afraid it's too big (like one of those power modules in a stereo) and too much heat.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

I

rrent

as

ical

it

and

een

e

of

back

vent

off

he

d

or

it.

well.

The problem with putting the switching regulator in the same chip as the temperature sensing electronics is that the switching noise inside the chip seems to limit you to +/-0.01K temperature regulation. The equivalent Intel adn Linear Technology chips seem to have exactly the same problem.

My 1996 paper describes a better - if bulkier - solution

-- Bill Sloman, Nijmegen

SMPS.

Careful, driving a TEC with a hysteretic stage can cause unwanted "signals". Depending on how sensitive the laser is you may be able to "see" the temp cycling. And the one/off transistions can couple in electrically. I have seen laser setups react to the air movement when someone walked out the door on the other side of the lab.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

and I

m current

ne has

optical

to fit

ng, and

rts of

the back

prevent

it off

ff the SMPS.

pling

f

at

g to

cal.net

d

But what is the hysteretic converter actually buying you? All you need to do to get PWM to work is to filter most of the hf content out of the current you put through the Peltier junction. You can mess around with actual frequency spectrum of the hf content - as we did back in

1996

- but you still need much the same sort of filter hardware.

-- Bill Sloman, Nijmegen

d I

current

has

ptical

fit

, and

s of

e back

revent

t off

the SMPS.

ing

to

al.net

d

I'd just use that Maxim chip. The vast majority of the time, the chip designer has worked out bugs the board designer hasn't thought of.

I have an old P23C CCD camera. Might be fun to cool the sensor and see how low the noise can go. My understanding is there is some trick to doing this so you don't get condensation.

ent

al

d n

ck

nt

f .

ll.

t- Hide quoted text -

How about a SMPS servoed a volt or so over linear control? (I'm not sure quite how to do this and certainly not on a schedule)

George H.

SMPS.

The laser tunes about 11 GHz per kelvin, and the stability spec is +-20 kHz, i.e. if I were actually getting the accuracy by temperature tuning, it would have to be within a couple of microkelvin. Fortunately it has an external reference etalon, so feedback can take that out pretty well.

But any switching junk on the TEC will be hard to get rid of.

Thanks

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

Thanks. I did it that way last time, but I don't have the space or the dissipation to spare. I'm planning on exiling the TEC driver and the DC-DC stuff to a separate steel can with their own board inside.

I like the AD chip--I may well use that one. Pity about needing so much current from the +5V rail.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net