-
I'd like to make a small SMT PCB heater circuit that I can add to any board to heat it to around 120F (for oscillator stability purposes). The precise temperature doesn't really matter, as long as it's warmer than room temperature and more or less stable (it will be insulated). I'm using 1W 150 Ohm 2512 resistors which when exposed to a 12V supply will conduct about 80mA each (about 1W). The 10K thermistor (which is about 3.9k at 120F) is located very close to the heater elements, so the heaters wouldn't be on 100% power for more than a few seconds. I would like to design this so it can be shrunk or expanded by removing or adding additional 150 Ohm resistors in parallel (let's say a maximum of 5, 400mA, 5W).
What SMT component would you recommend for Q1? For testing purposes I'm using IRF510 because I have them lying around, but I imagine this part unnecessarily pricey and powerful for this task and isn't available as SMT anyway. I'm thinking about si2302cds (http://www.vishay.com/docs/68645/si2302cds.pdf) which has an ID >2A, but I'm not sure if power dissipation will be a problem (that component is limited to 700 mW)... I appreciate any suggestions or insights you may have!
Thanks in advance
-
Why not do this?
Much more stable (high open loop gain). Note the regulator is heatsunk to the same heat spreader as the transistor, because it dissipates a proportional amount of power.
(And no, the TL431 isn't necessary.)
Tim -
Why not do this?
Much more stable (high open loop gain). Note the regulator is heatsunk to the same heat spreader as the transistor, because it dissipates a proportional amount of power.
(And no, the TL431 isn't necessary.)
Tim
This is a great suggestion, and for a similar application too - thanks! The 7805 will indeed get toasty dropping 12V down to 5V, but for a heater circuit couldn't the voltage regulator be eliminated altogether? The heater circuit should run fine at 12V assuming the op-amp can take it, right? -
are you expecting self oscillating "bang-bang" control?
Linear would be better but you have to have the pass Q pkg handle the heat too in a Linear design
not that that's a problem with SOT-223 or 89, Dpak, D2... -
This is a great suggestion, and for a similar application too - thanks! The 7805 will indeed get toasty dropping 12V down to 5V, but for a heater circuit couldn't the voltage regulator be eliminated altogether? The heater circuit should run fine at 12V assuming the op-amp can take it, right?
It can be designed for whatever voltage you happen to have, of course!
Since that circuit needed regulation anyway, and the regulator was dropping about half the supply voltage, it was a natural candidate for placing on opposite sides of the heatsink, balancing the heat source nicely, reducing the gradient.
Tim -
Why not do this? ... Much more stable (high open loop gain). Note the regulator is heatsunk to the same heat spreader as the transistor, because it dissipates a proportional amount of power.
I fear that I'm nearly back to square 1. Running that circuit at 12V, it's basically the same as mine, it's just using a PNP darlington configuration rather than a MOSFET, and those parts aren't available as SMT components. Something in DPAK or SOT-23 would be ideal... I'm trying to avoid TO-220are you expecting self oscillating "bang-bang" control? Linear would be better but you have to have the pass Q pkg handle the heat too in a Linear design not that that's a problem with SOT-223 or 89, Dpak, D2...
I definitely do not want bang-bang control! That's partially why I knee-jerked toward a MOSFET instead of a darlington. I think my original schematic would probably be fine (right?), assuming an appropriate SMT-packaged MOSFET were used (which is sort of the original question) -
Your schematic has problems:
- Self heating of the thermistor may cause a few degrees of error. (Probably a small niggle, but it causes VCC sensitivity, and is an easily solved design issue.)
- The op-amp is fixed gain, so the accuracy won't be good anyway.
- Putting caps on op-amp inputs is a bad idea.
The darlington has the same gain as the MOSFET -- the source/emitter degeneration resistor dominates. Either is fine. I didn't have a PMOS handy for that project, and BJTs are cheaper anyway.
SMTs for an oven are kind of dubious, just because you don't have any good way to spread the heat out. In my project, the regulator and transistor were soldered to a sheet of, I think, 8oz copper, which wrapped around the PCB.
D/D2PAKs could be soldered to a heat slug very nicely, but you need to arrange that somehow. (You can even get PCBs with heavy copper, and slugs, and stuff in them, but... yeah... custom $$$...)
Anyway, FETs comparable to IRF510 (which itself is stupid cheap!) are available in anything you can think of: SO-8, SOT-223, SOT-89, DPAK... Likewise, TIP32C --> MJD32C (DPAK).
LM358 will be an okay replacement for TLV2372. It obviously works fine under the 5V conditions, and is okay for 12V, but not if it's dirty 12V. Nice thing about the bipolar output: it biases the op-amp's output stage, which eliminates LM358 crossover distortion. That's something you might miss with a MOSFET output.
Tim -
Your schematic has problems:
- Self heating of the thermistor may cause a few degrees of error. (Probably a small niggle, but it causes VCC sensitivity, and is an easily solved design issue.)
- The op-amp is fixed gain, so the accuracy won't be good anyway.
- Putting caps on op-amp inputs is a bad idea.
The darlington has the same gain as the MOSFET -- the source/emitter degeneration resistor dominates. Either is fine. I didn't have a PMOS handy for that project, and BJTs are cheaper anyway.
SMTs for an oven are kind of dubious, just because you don't have any good way to spread the heat out. In my project, the regulator and transistor were soldered to a sheet of, I think, 8oz copper, which wrapped around the PCB.
D/D2PAKs could be soldered to a heat slug very nicely, but you need to arrange that somehow. (You can even get PCBs with heavy copper, and slugs, and stuff in them, but... yeah... custom $$$...)
Anyway, FETs comparable to IRF510 (which itself is stupid cheap!) are available in anything you can think of: SO-8, SOT-223, SOT-89, DPAK... Likewise, TIP32C --> MJD32C (DPAK).
LM358 will be an okay replacement for TLV2372. It obviously works fine under the 5V conditions, and is okay for 12V, but not if it's dirty 12V. Nice thing about the bipolar output: it biases the op-amp's output stage, which eliminates LM358 crossover distortion. That's something you might miss with a MOSFET output.
Tim
Tim,
Thank you for the thorough response! Your application seems to be much more high power than mine. My intent is to have a Styrofoam-insulated PCB at largest 2x2", so one or two 150 Ohm resistors at 12V (80 mA each) will be more than enough to heat it with relatively little current. The heated / oscillator portion of the project will be enclosed in Styrofoam, and the rest will be on a regular board, all inside an unheated metal enclosure. My idea is to place the heater resistors to the opposite side of the PCB as the thermistor (which will be near the crystal). With such small thermal mass, I'm not really worrying much about dissipation traces, slugs, etc. too much. The MJD32C is a good recommendation, I'll research it further! [http://www.st.com/content/ccc/resource/technical/document/datasheet/39/59/ca/a9/e5/74/4a/15/CD00164144.pdf/files/CD00164144.pdf/jcr:content/translations/en.CD00164144.pdf]
I do have a question regarding your advice against capacitors on op-amp inputs. Since the circuit will be so close to the oscillator/buffer, and there may be a long trace between the thermistor and op-amp, I added the caps there to keep RF out of the inputs. Is this really a bad idea?
-
Why not put the transistor inside too? It dissipates as much as the resistors (if not more). And that way, you only need one pair of high current wires (+V and GND) to the blob, not three.
The dissipations aren't evenly shared, though (between transistor and resistors). It might be beneficial to split the transistor into several, one for each resistor. Then each transistor-resistor pair is a linear (voltage to power) element, which helps control loop stability! (In my project, notice everything is one big current sink, so the total power is linear with control voltage.) And those can be distributed around to reduce gradient. The transistors won't need to be very big, probably the SOT-89 version of 2N3904 would be fine.
A cap from the thermistor is good, yes. A cap on the negative input pin is not a good idea, however..
Tim -
Why not put the transistor inside too? It dissipates as much as the resistors (if not more). And that way, you only need one pair of high current wires (+V and GND) to the blob, not three.
The dissipations aren't evenly shared, though (between transistor and resistors). It might be beneficial to split the transistor into several, one for each resistor. Then each transistor-resistor pair is a linear (voltage to power) element, which helps control loop stability! (In my project, notice everything is one big current sink, so the total power is linear with control voltage.) And those can be distributed around to reduce gradient. The transistors won't need to be very big, probably the SOT-89 version of 2N3904 would be fine.
A cap from the thermistor is good, yes. A cap on the negative input pin is not a good idea, however..
Tim
Thanks for the tip! Using transistor/resistor pairs is a good compromise between minimizing expense and allowing easy expansion. I've updated the schematic to reflect this.
I'm not entirely confident in my prediction of power dissipation of the transistor, and I want to make sure I appropriately match the power handling capability of my resistor/transistor combo. Could you look over my shoulder on this one? I understand that a full open transistor would produce about 12V on the 150 Ohm resistor (80 mA / 1W dissiptation of the resistor), but when partially open what is the maximum power dissipation I should expect in the NPN? While I intended to use 1W resistors, the MMBT3904 is limited to 250mW. I'm looking at the datasheet curve "Collector current as a function of collector-emitter voltage" and assume my VCE will always be well below 1V, so I'm thinking:
PNPN = I*E = 80mA*1V = 80mW
... and 80mW is well within spec of the NPN, so we are in the clear?
EDIT: that (10) trace is scaring me, so I think my "less than 1V" assumption may be wrong. If this is the case, how would you recommend selecting a resistor value and dissipation spec to match a transistor's dissipation spec?
Thanks again for your help!
-
Why clean P controller, why not average the error control a bit with PI, since it have pretty much dead-time? Also gain should be conservative. My head says the stability would be better than clean P. Hopefully I'm not mis-interpretting the schematic, haven't much used the classic opa analog controller .
T3sl4co1l schema of controller is PI if I read it correctly.Also PTC vs NTC and the sense network makes difference since it is passive cooling active heating, but I can not get my head which way around it were.
nah go with NTC just look that you are at as linear region as possible to avoid mathematical modeling or go with RTD, which are more linear by nature.
-
Right, this part is very important:
The series 100k resistor sets the value of the other two components accurately (i.e., the 100k has much more resistance than the thermistor, and has a constant value).
The 1M has the effect of setting the proportional gain, so that for sudden changes in temperature, the power is changed by a modest amount. The 10uF takes this short-term error, and allows it to grow over time, so that a small error over a long time (a time constant in the 10s of seconds, in this case) gets adjusted out.
Tim
-
I'm not entirely confident in my prediction of power dissipation of the transistor, and I want to make sure I appropriately match the power handling capability of my resistor/transistor combo. Could you look over my shoulder on this one? I understand that a full open transistor would produce about 12V on the 150 Ohm resistor (80 mA / 1W dissiptation of the resistor), but when partially open what is the maximum power dissipation I should expect in the NPN?
Exactly 1/4W.
Work out the math yourself, it's a fun little proof.
SOT-23 would be pushing it, but SOT-89 isn't much bigger or pricier, and is quite available.
Tim -
Vtile and T3sl4co1l, thanks for your comments about improving the amplifier by making it a PI device. A 22uF cap across R5 should have a time constant of about 1s, and can stabalize the circuit. It's slightly different than the T3sl4co1l layout, but still a typical op-amp integrator and I assume will function similarly.
[Q1 would dissipate] exactly 1/4W.
I would love to work this out for myself, but that's part of my question... When I did it [a few posts above] I came up with 80mW dissipation in the NPN which is different than your 1/4W, so I'm assuming I'm doing something incorrectly... where are you getting the values to calculate power dissipation for the transistor?
Thank you so much for your advice! -
I would love to work this out for myself, but that's part of my question... When I did it [a few posts above] I came up with 80mW dissipation in the NPN which is different than your 1/4W, so I'm assuming I'm doing something incorrectly... where are you getting the values to calculate power dissipation for the transistor?
After some napkin calculations, I think I figured something out. I found some sources that indicated a rule of thumb that a transistor's power dissipation usually maxes at 1/4 of the restive load it's controlling, but I couldn't figure out how people kept getting that number. After a little chicken scratch, I think I figured out where these values are coming from. The key I was missing was this assumption:
VNPN = VCC - VCE = VCC - RI
then when I set the first derivative of I to 0 (assuming that's where it maxes out?) and plug and chug, I end up with P=V2/(4R) which matches what other people seem to be saying, so I'm happy to leave it there.
From now on, equations can be thrown out the window, and I guess I can assume maximum power dissipation of a transistor is approximately 1/4 of the power dissipated by its resistive load. This will help me choose a transistor for this application.
Thanks everyone!
-
Bingo!
(And Dave says there's no application of calculus in electronics. If you don't care about understanding anything, sure!)
Tim -
Why not do this?
Much more stable (high open loop gain). Note the regulator is heatsunk to the same heat spreader as the transistor, because it dissipates a proportional amount of power.
(And no, the TL431 isn't necessary.)
Tim
Tim, I've got one more question about your design... What is the rationale of using the 1 Ohm resistor above the TIP32? At first I thought it was for current limiting (limits to 5A), but the LM7805 self-regulates current output to 1.5A. It may act with the 47uF capacitor on the output of the regulator to provide some low-pass filtering, but 1 Ohm is such a low value it's probably not the intent. Why is it there, and could it be omitted?
Thanks for your input! -
It certainly can't be omitted!
5A would be the short circuit figure, if the TIP32 died. But since that doesn't happen,
The op-amp output can only go as low as ~0V. It has a 1k + 1k voltage divider on its output, so the transistors see a minimum 2.5V.
Stack two Vbe's on top, for the Darlington transistor configuration, and you're at a worst case ~3.9V, which is only 1.1A. Quite cozy!
Indeed, if the transistors ran away and the 7805 limited instead, 1.5A would be a fine limit, too.
I found this circuit took about a minute to come up to temperature (saturated at ~1A), and idled at about 200mA (12V supply), using some plastic to insulate it.
The resistor is also required because it sets the transconductance of the output stage. Without a resistor, not only would the range be nearly unlimited (more than enough to pull the 7805 into current limiting), but it might never stabilize, because the loop gain would then vary exponentially with setpoint. (This is still true at very low power, where there is a threshold before the transistors conduct (~3V on the op-amp output), and where the transition from "off" to "linear" occurs exponentially. But the gain can only be lower than the resistor-limited linear gain is, so this will not cause instability.)
Matching up the operating ranges (the op-amp's output to the transistors' input) is the first goal towards stabilizing a control loop.
Tim -
It certainly can't be omitted!
5A would be the short circuit figure, if the TIP32 died. But since that doesn't happen,
The op-amp output can only go as low as ~0V. It has a 1k + 1k voltage divider on its output, so the transistors see a minimum 2.5V.
Stack two Vbe's on top, for the Darlington transistor configuration, and you're at a worst case ~3.9V, which is only 1.1A. Quite cozy!
Indeed, if the transistors ran away and the 7805 limited instead, 1.5A would be a fine limit, too.
I found this circuit took about a minute to come up to temperature (saturated at ~1A), and idled at about 200mA (12V supply), using some plastic to insulate it.
The resistor is also required because it sets the transconductance of the output stage. Without a resistor, not only would the range be nearly unlimited (more than enough to pull the 7805 into current limiting), but it might never stabilize, because the loop gain would then vary exponentially with setpoint. (This is still true at very low power, where there is a threshold before the transistors conduct (~3V on the op-amp output), and where the transition from "off" to "linear" occurs exponentially. But the gain can only be lower than the resistor-limited linear gain is, so this will not cause instability.)
Matching up the operating ranges (the op-amp's output to the transistors' input) is the first goal towards stabilizing a control loop.
Tim
Ah, that makes sense. It's important in stabilizing the loop in this op-amp / analog control configuration. Curiously, if the base of the Darlington were driven with TTL levels (slow PWM or even bang-bang temperature regulation), would the 1 Ohm resistor then become unnecessary? [appreciating it may not control temperature as finely, and may be undesired for RF applications, but also that the power dissipation of the transistor would be greatly diminished, and that the 7805 would become the primary heating element, and the 1.5A internal regulation of the 7805 would always be used while heating] -
I wonder if an LM317 and an NTC couldn't do almost all of it and at the same time be procted against over temp
-
I wonder if an LM317 and an NTC couldn't do almost all of it and at the same time be procted against over temp
This seems ideal because it minimizes component count, has current limiting and thermal protection built in, and the LM317 (which would be the primary heating element) is in a package easy to mount to a metal enclosure... It feels weird to short the output of the LM317 (or LM7805) to ground to use the regulator as the primary heating element, but it should be able to take it right? -
I know we have deviated somewhat from the initial design, but I'm really appreciating all your responses. The last couple posts made me consider a minimal-case bang-bang design. In this circuit, heating would be controlled by an ON/OFF microcontroller pin (perhaps using PWM). Is it okay to rely on the internal current limiting of the LM7805 to use the regulator as the primary heating element? Driven with TTL levels, I assume the Darlington wouldn't get hot. I understand this may be noisy and sub-optimal when in close approximation with RF traces, but for bang-bang temperature regulation would this work, or is this a terrible idea? The goal here would be to utilize the easy TO-220 7805 package as an easy-to-mount heating element, rather than spending money on expensive power resistors with carefully-selected values (and radial ones are difficult to mount to an enclosure).
-
I wonder if an LM317 and an NTC couldn't do almost all of it and at the same time be procted against over temp
This seems ideal because it minimizes component count, has current limiting and thermal protection built in, and the LM317 (which would be the primary heating element) is in a package easy to mount to a metal enclosure... It feels weird to short the output of the LM317 (or LM7805) to ground to use the regulator as the primary heating element, but it should be able to take it right?
rethinking, shorting the output of an lm317 won't work the minimum output voltage is ~1.25V
maybe an lm337 with the input to ground and output to supply...
-
My restlessness got the best of me. I took a LM7805 and soldered the output pin to its ground pin
I then used some thermal compound to attach it to an LM75A (i2c temperature sensor) board I had laying around, and used a bus pirate to pull temperature as I applied 12V on the input. It looks like after a few minutes it stabilized around 225 F. I'm assuming this is where the thermal shutdown began to kick in (it's graded, not an all-or-nothing shutdown right? It's hard to tell from the datasheet) and current draw from my 12V supply dropped to about 0.25A (3W). I'm guessing this means an LM7805 shorted to ground makes a not-awful option for a heater in hack projects (not suitable for production), where temperature can be controlled (bang-bang, or perhaps even PWM) by controlling current to the permanently-shorted device, or by shorting it through a transistor when desired...
-
Ah, that makes sense. It's important in stabilizing the loop in this op-amp / analog control configuration. Curiously, if the base of the Darlington were driven with TTL levels (slow PWM or even bang-bang temperature regulation), would the 1 Ohm resistor then become unnecessary? [appreciating it may not control temperature as finely, and may be undesired for RF applications, but also that the power dissipation of the transistor would be greatly diminished, and that the 7805 would become the primary heating element, and the 1.5A internal regulation of the 7805 would always be used while heating]
Why would you go to the trouble of buying and placing all the components that make an analog regulator, then smash it all up with a digital hammer until it's no better than a Klixon?
AFAIK, an overloaded three-terminal regulator has the same on-off behavior.
I don't get where you're going with this.
Tim