Crystal oven theory

Clearly not. Most people understand over-compensated to mean very long settling times, and under-compensated to mean ringing. If it is second or or higher order system, a sufficiently undercompensates unit can oscillate.

We usually tune thermal loops in the classic linear style, for critically-damped small signal disturbance. They will typically overshoot at cold-start powerup because the integrator winds up and rails. Rob did a lot of fancy math to define the software control loop; I just Spice things like that.

The oven that I pictured has a 2nd order plant response, about 75 minutes tau for the huge aluminum block and about another 17 for the platform and e/o modulator inside. Spacers set the second time constant; aluminum, stainless, or plastic.

One goal was to force minimal thermal gradients along the length of the Mach-Zender electro-optical modulator.

One of the less-frequently remarked advantages of thermoelectrics is that anywhere near room temperature they can have considerably better forcing resistance than heater-only loops with a given bandwidth.

A heater-only loop gets its negative slewing exclusively from the heat leak, which of course leaks in both directions, whereas a TEC can pull as well as push, so a TEC-based loop can be better insulated for the same control bandwidth and slew rate.

Cheers

Phil Hobbs

bitrex snipped-for-privacy@example.net wrote in news:47UiK.3545$ snipped-for-privacy@fx10.iad:

Look at it. It is smaller than the can it is in. So yes, it requires mechanical support.

snipped retarded link post.

Real mature, twerp.

John Dope snipped-for-privacy@message.header wrote in news:t6hvrj$kq3$3@dont- email.me:

John Dope is still in the dark. You never knew electronics and have been a mere interloper all these years, at best.

The actual quote is "Every idiot can count to one."

So even your google skill sucks. John Dope lives up to his name every time he posts into SED.

We only run at 30c, and ambient is very reliably 22. We considered TECs but that was too complex and expensive. We sure don't want a fan anywhere close to our optics.

The dpak mosfet heaters are cheap!

he Mach-Zender electro-optical modulator.

But a TEC has a thermal low-impedance-path from inside to the outside. That is unwanted and one of the reasons for their bad efficiency. That means that heat transients on the outside get in immediately and must be activly regulated away, AFTER they have happened.

Yes, one could insulate the outside of the TEC, but that is somewhat ill?

Cheers, Gerhard

tirsdag den 24. maj 2022 kl. 20.35.50 UTC+2 skrev Gerhard Hoffmann:

and every time you stop cooling the heat you just moved come right back at you

John Dope snipped-for-privacy@message.header wrote in news:t6j8sq$61v$3@dont- email.me:

You are a f****ng total retard. ALL that was EVER needed was the guy's name. So you are dumber than a 5 year old. It does not get dumber than you John Dope.

Kids know how to use the hardware better than you do and they get better results too.

Aren't you the total retard trying to power a bicycle with a drill motor?

Sure, I'm all in favour of what works. The lower heat leak thing is a big help if there's forcing above a few percent of the loop bandwidth.

Cheers

Phil Hobbs

I invite you to compute the ratio of the bandwidth available to a TEC loop vs a heater loop as the delta-T goes to zero. ;)

You can get TECs in all sorts of sizes.

Heater loops work fine at larger delta-T.

Cheers

Phil Hobbs

You could put another temperature sensor on the outside end of the TEC, and do some sort of feed-forward as well, to remove most of the effect before it gets in.

A really important tool in high-stability temperature control is a local feedback loop around each actuator. Proportional-only is fine, but what you really want is super high speed. These days we do a lot of TEC-based controllers. Our favourite method is to put a bare ENIG pour on the bottom side of the cold plate board (which usually has a laser or one or more photodetectors). This contacts the cold side of the TEC via (usually) silver grease.(*) One end of an 0603 thermistor is soldered to the pour, and the other goes via a longish skinny trace to the loop amp.

Doing that costs next to nothing, and allows us to make a much faster simulated actuator with much much lower effective thermal conductance.

Feedforward to the outer loop can also be done, but requires much more tuning because the speed of the outer loop changes dramatically with delta-T and Qdot.

Cheers

Phil Hobbs

(*) Thermal silver grease has much much more silver than electrically-conductive silver paste.

Bad ovens sometimes sense ambient temp and feed-forward into the setpoint, to shotgun compensate the badness.

The SRS SC10 is a good example of an oven done wrong. The manual of rev L is available online.

That scheme makes a lot of sense.

With the thermistor hanging five right at the edge of the pour (right next to the edge of the cold side of the TEC) the response is faster than 100 ms.

Thermal diffusion is quadratically slow at large distances, so of course it's quadratically fast at small ones!

Cheers

Phil Hobbs

There's a lower limit set by the structure of the thermistor. It's finite lump of metal oxides,and it takes a finite time for heat to diffuse from the edges to the middle.