Fet as heater (one mo' time)

Ummm... when I look at that auction (by number) it's quoting a sale price of $11.99, plus $100(!) for shipping (ePacket delivery from China). Or, if you want "standard" shipping to the U.S. (which on average doesn't get here any sooner) it would be $127.80 for shipping.

It _looks_ like a great deal if you look only at the sale price (too good, really) but the small-print shipping price makes it less wonderful.

I suspect they're trolling for people who hit "Buy It Now" and don't see the shipping cost until after they're already committed.

I don't buy from vendors who do this sort of thing. 'Tis more Ferengi than Klingon.

IIRC you're hooking up the FET as a source follower and driving the gate. If so, then the transconductance is being swamped by the source resistance, and so doesn't really matter.

Threshold might matter, but only if it's big enough that you can't turn the thing fully on.

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com 

I'm looking for work -- see my website!

Ahh! The light strikes! One ohm source resistor. That's where most of the control is.

Thanks Tim!!

George h.

Hi Steve, I've started several threads on this circuit. (With lots of help from all the experts here!) I'm hoping this is the last.

I'm mounting the temperature sensor (diode connected transistor in TO-220) right on top of one of the three heater resistors. This gives me a time constant, that depends mostly on the delay between the heater and sensor.. and hardly at all on the big thermal mass that it's connected to.. or on how that mass connects to the rest of the world. (This was a BIG discovery for me... though Phil H. has been telling me for years. (I'm totally of the belief that you have to "do" something to understand it.)

The gain seems pretty constant too, (from 77 to ~400K) (I guess that speaks to the quality of the resistors.)

I'm glad you enjoy my posts.. the women at work, (who run the company) were kidding me the other day that I don't talk in complete sentences. Before my mouth gets to the end of a thought my brain has moved on.

My writing is similar, George H.

Oh I should add I wasn't blaming my boss, part of his job is to break what I make. He was the first to use it in vacuum, and like you said, my thermal design was deficient.

GH

Are you sure that's what you want to do? You are controlling the temperature at one spot in the whole assembly. What about the thermal gradient to the rest of the world, and to the item you are trying to control?

If there is much of a gradient, then you really don't know what the temperature is at the item you are trying to control.

If you monitor the temperature at that spot, there will be a time delay in the feedback. This will be difficult to handle with a simple feedback control, but the PID can do it.

I guess it depends on how accurate the temperature has to be. If it's a few degrees, and the gradient is less than that, then maybe the resistor is a good enough place to monitor the temperature. But if you need tighter temperature control, it might take a more elegant approach.

Anyway, keep the PID in mind in case you need it later. The REX-C100 is one of the best available, and is often advertised for sale at $40 to $75. The link I gave earlier sells it for $12.78 and is the cheapest I could find. One downside is it may take several months to deliver after you place your order.

LOL

you are still missing my point

the hotter the junction , the less reliable it will be, even when run within the limits

use a resistor as your heat source and keep FET Tj as low as possible

m

Right, we've (SED) talked about some of that before. In this case its a copper block, with weak thermal links to a cold bath. I get to set the temperature at one point. I futzed around putting the sensor in the middle of the block, This took a much longer time constant ~10-30 seconds. (with attendant slow wiggles and overshoots that are always there in a PID) When I just monitored the temp in the middle (w/control sensor on the R) It would still take ~30 seconds to stabilize in the center.. but all the over shoots and stuff happened on the much shorter time scale. (I'm also guessing that the time constant of the copper will change with temperature. But it's not something I looked at.) Putting the sensor right on the plant seemed to be a win in this case. (I've made several thermal loops before... mostly with TEC's)

Oh boy, thanks for the recommendation, but... I used an Omega PID controller in two of our instruments. I think ~1/2 the service calls I get about the instrument are people having trouble with the commercial controller. I've dreamed ever since of designing it out... but it's not worth it.

George h.

No I got you. If the FET burns out they (the customer) will have to put in a new one. Three wires and a screw.

I've got a lamp circuit that runs at 120 C and uses a resistor and NPN (TIP31? It might be a darlington?) as the heater. ~150 in the field no problem so far.

George H.

I'm surprised there is such improvement. Very interesting. Thanks for the info.

Wow! One ad wants CAD $510.00 for the CN77332:

But other sources such as ebay are much cheaper.

I'm starting to find all kinds of uses for the PID temperature controller. DIY soldering iron, solder pot, precision OCXO, Rubidium clock temperature regulator, homemade soup - for $12 bucks you can afford to put them everywhere.

I can see where people may have problems with a PID. Maybe check some of the Youtube videos and see if REX is easier to use than the Omega. If so, maybe it might be worth considering changing to the REX, and give your customers a list of suitable youtube videos to watch.

In regard of the Rubidium clock, I have some questions I'd like to ask you. Do you mind if I contact you offline?

I think we pay ~$250 each.

Mostly something gets screwed up in the long chain of menus, and they can't find it. We give them all the parameters to program in... well we enter them here too, but the students and push all the buttons and get lost.

I've only got "book" knowledge of Rb clocks. Why not ask here? There are many lurkers, some of whom might know the answer.

But sure off line is fine too. George H.

The role of the source resistor in these applications is something that I'm not too sure about. I have a prototype oven controller that only has to manage about 0.5 watt

It works well with a small TrenchMOS FET being operated in its subthreshold regime (and well within its datasheet SOA.) The FET is mounted on the bottom side of the PCB under the IC being heated, while the drain resistors are RTV'ed to the top of the package alongside the thermistor.

I've never seen a circuit like this that didn't use at least an ohm or two of source ballast, but what's the use of a source resistor when you're not either paralleling multiple FETs or closing the loop around the current through it? At the end of the day, the transconductance doesn't vary enough with temperature (or anything else) to destabilize the loop, so it's not clear why I'd need to swamp it. I went back and forth a few times and finally decided to omit the source resistor on the grounds that I couldn't justify it on any basis other than tradition. Anyone care to second-guess this call?

-- john, KE5FX

Sorry, make that 0.5A/2.5 watts.

-- john, KE5FX

Ouch! I think I'll go with the REX

Actually, I found a way to model it in LTspice. This gave me the answer I was looking for, as well as a path for future development. So I thank you very much for your help, but I won't be needing to waste your time.

Thanks!

Hi John, I'm at home with "no" download left. (I didn't try to look at your pic.) My circuit is a simple V to I converter. The error signal (V), drives the non-inverting input of an opamp. which drives the fet gate. Feedback to the opamp is from the source and one ohm source resistor. (I divide down my ~13V max opamp error signal so that Imax ~ 1amp.)

I've been worried about what my opamp was pushing. I was thinking I wanted more dynamic resistance in my Fet. Tim pointed out that the opamp is pushing the one ohm source resistor around. Which I hadn't understood before.

Does that help at all? George H.

Huh, you are not wasting anyone's time if you ask a question here. on the other hand, I do often find that as I'm composing a question, to an intelligent audience, I find the answer/ mistake on my own. I sometimes post those anyway, more often not.

There's this "guy thing" that I've never understood. (maybe it's about getting laid?) I've never been afraid to ask questions, display my ignorance in public. "I'm lost, which way to the highway?"

In fact, I often find myself playing "the fool"*. 'cause if I'm not sure, everyone will tell you when you are wrong, but few will confirm that you've said something right.

George H.

*I hope SED will forgive me for this behavior, the latest example in mind now, is when Phil H. was telling me about how he modeled thermal loops as a time delay * time constant. Well, that's also relevant to this thread.

It sounds like the resistor makes sense in your case, given that you're using local feedback. Where does your error signal actually come from? In my case the only feedback path is the heater-thermistor connection. My G-S voltage swing is smaller than it would be with a source resistance, but I don't really see a downside to that.

(The only reason I'm using a rail-to-rail opamp in that schematic is because the circuit began life with a bipolar. Vcesat -- or rather the inability to reach Vcesat -- was causing excessive warmup time.)

I don't see any problems with your approach that can't be solved with a bigger FET, as you found. You live and die by the SOA plot with these sorts of circuits. The abs-max numbers themselves are almost irrelevant.

-- john, KE5FX

the latest example

Well, that's also

Nothing wrong with that. I use SED as a sounding board fairly often myself, e.g. a year or so ago I had that weird problem of a third-order IMD peak t hat exhibited a first-order amplitude dependence.

Since I work mostly alone, that helps a lot, and we can all learn something , which makes it a win/win.

Cheers

Phil Hobbs

I suppose this thread is dead by now, but

You might consider moving the fet and it's control circuit to some remote location, where cooling can be provided. Only the resistor and thermocouple need be in vacuo, where the heating effect is required.

RL