PWM freq limit for TECs

[BillA]

am trying to assess the life expectancy of large (226W) TECs when controlled with PWM

googling I see 'controllers' with frequencies from 50 (!!) to 175kHz
with minimum frequencies described as 2k to 5kHz (to preclude/minimize thermal stressing)
- Motorola for example says 5kHz is a minimum

obviously filtering is required, but this is not my concern;
rather it is the life expectancy of a TEC at 'lower' cycling frequencies - say 500 cpm

anyone got some real info and or links ?
(please, no 'I think' kinda stuff)

[murray13]

Bill, here is a whitepaper from TEtech talking about the detrimental effects of PWM and it's frequency.

Whitepaper

Hope it helps!

[BillA]

thanks
am familiar with it; its conclusion is at odds with every other citation I have found
and unfortunatly this paper is cited by those mfgrs w/o data

most annoying

[pHaestus]

What does one use to PWM a pelt pulling that much power Bill?

[BillA]

Maxim has a nice design article which would have to be scaled way up
this design is at 100kHz but also needs to be scaled up

here is a commercial example

our principal TEC customer designs their own controllers

EDIT: link corrected

[murray13]

True, that paper did not come to the conclusion that I was expecting.

On another note, I am very familiar with Thermotek chillers. Disecting one for months, on and off.

They use a PIC chip to control the temp. Their power supply has a Vishay switchmode controller (9114a) and a Unitrode power factor regulator (2855a). The PIC sends a varying pulse width 10khz signal to the power supply. The output of the PS is 10khz, not very 'clean' at least to me. But then again I don't have the right equipment to look at it very well. (cheap scope, can't float the inputs to get a differential view) At 10khz though that might be high enough to not matter to a TEC very much.

Good luck in your quest. I would be very interested in anything you come up with, keep us posted.

[]JR[]

Originally posted by pHaestus
What does one use to PWM a pelt pulling that much power Bill?

Big FET's, well not so big a couple of the irfp460's i got infront of me now could cope max drain 20amps (500v)

Then all you need is a pic to take an analouge (or thermocouple depending on how your feeling) input to generate the squarewave to switch the fets.

Here was my idea from a while back, i believe its quite novel although i could be mistaken.

Take the pwm control circuitry out of an atx (or any smps) powersupply and replace it with a temperature controlled pwm generator. Then as the temp changes the pic can alter the voltage direct, and on a hysteresis curve, which then eliminates hitting the tec with pwm itself but alters the voltage quickly but smoothly instead.

The other option could be a max336cpi (also at my desk ironically) to build a digital resistor with 256bits of resolution.

Also I think i still got a link of a manafacturer you might want to contact, they do the 220w tec cheap (like $15 min order 10) aswell as the 172w module for like $11. But they also make PWM resistant module iirc.

Whats your overall goal BillA?

]JR[

[BillA]

not really to design a controller as the DIY use of TECs is normally flat out,
and our big customer designs their own

just wanting some good info on MTBF as a function of PWM freq
(though such is also related to the subsequent filtering of course)

[Since87]

Originally posted by unregistered
Maxim has a nice design article which would have to be scaled way up
this design is at 100kHz but also needs to be scaled up

here is a commercial example

our principal TEC customer designs their own controllers

The 'Maxim article' link took me to an O/C thread.

The second design you linked has an LC filter so that even though PWM is used to vary the input to the TEC, the voltage applied to the TEC is effectively DC. (There will be some small amount of ripple at 100 kHz but I'd consider that inconsequential except for EMI issues.)

The article Murray linked is the most in depth document I've seen, but is not very satisfying.

500 CPS (8.33 Hz) is way below all recommendations I've seen. I've frequently seen recommendations to keep ripple below 10% with 60 Hz supplies. (No links handy though.) Why would anyone want to run such a low frequency?

[BillA]

corrected link
"Why would anyone want to run such a low frequency?"
no idea
there are a number of 'cheap' PWM controllers running in the 600 to 800 Hz range, several also with a dual range the higher being 2500 Hz or so

but the space apps seem to be at 20kHz

[Since87]

Originally posted by unregistered
just wanting some good info on MTBF as a function of PWM freq
(though such is also related to the subsequent filtering of course)

It is much more difficult to filter low frequencies than high. With the components available these days, there is no justification (other than ignorance?) to using a low frequency signal.

Sorry I have no source of accurate data. Is suspect you are going to just have to do as the TEC manufacturer, and just say, 'Don't do it.'

[Brians256]

Since87, low frequencies are harder to filter, but are they not also lower efficiency? The higher the frequency, the more losses you get from transition time. Higher current switching elements (whether they are FETs, IBGT's or whatever) have slower response time.

Maybe at the frequencies we are talking about (100KHz) it really isn't a big deal?

[Since87]

Originally posted by Brians256
Since87, low frequencies are harder to filter, but are they not also lower efficiency? The higher the frequency, the more losses you get from transition time. Higher current switching elements (whether they are FETs, IBGT's or whatever) have slower response time.

Maybe at the frequencies we are talking about (100KHz) it really isn't a big deal?

Yes, if you are talking straight PWM without filtering, higher frequencies are less efficient. Although with the components available today, the efficiency can be kept pretty high up into the MHz range. 'PWM' itself doesn't imply any filtering.

However, when you add an output inductor (to get a 'DC' output) the resistance of the inductor winding becomes a significant issue, and at higher frequencies you can use smaller, low resistance inductors.

There are lots of tradeoffs when it comes to DC-DC converters, and the 'best' operating frequency depends on a lot of different issues.

[Since87]

Originally posted by unregistered
but the space apps seem to be at 20kHz

I would guess that for space apps there is no filtering. (Wouldn't want the added weight or loss of power efficiency.)

Perhaps 20kHz is what NASA considers a 'safe' operating frequency for powering a TEC with a PWM signal directly.

[SexyMF]

Since87, low frequencies are harder to filter, but are they not also lower efficiency? The higher the frequency, the more losses you get from transition time. Higher current switching elements (whether they are FETs, IBGT's or whatever) have slower response time.

Resonant techniques can be employed to minimise conduction and transistion losses. There are mighty FETs avaiable which have very low on-state resistance (and hence low on-state losses).