Help me choose a mosfet to regulate a resistive load

[Stefano]

Hello, first post here!

I would like some help for my first high power project: an hot-plate capable of heating up to 55°C (won't use it for that much, planned for more to have some margin)

The hotplate is made of an aluminum pcb with the usual snake-like trace on one side.

I'm ecountering difficulties in choosing the MOSFET to control the resistive load, but before we talk about that I'd like to explain my design and thought process so that someone can spot potential issues.

Given the conductor width and lenght I choose, I used the Saturn PCB Design tool to get an estimate of what amount of power is needed to heat the plate to the required temperatures:
- track width: 2mm
- track length: 14 meters
- total track resistance: 4.5 Ohm
- voltage drop across the heater: 15.5V
- amp draw: 3.4A
- total power needed: 52W


At this point I needed to choose a PSU that is capable of more than 15.5V and 52W, so I choose an AC/DC brick that is capable of 24V and 90W.
This might seem excessive, but being my first high power project and taking into account the uncertainties of the Saturn tool I think 90W is perfect, almost double of what I theoretically need and the price is nice. (see bottom for parts lists)

At this point I have a 24V PSU and a 2Ohm load, which would result in 12A current, which is too much for both the wanted result and the PSU to handle, so I need to add some resistance such that the current drawn by the circuit is actually 3.5A. This can be done by operating an N-Channel MOSFET in the ohmic region and regulate Vgs to have the right resistance in the circuit for the wanted power draw.

How much resistance do I need to add? V = I*R, 24 = 3.5 * X, X = 6.8 OHms, I already have 2 Ohms from the heater, so I need to regulate the MOSFET to have ~6.8 Rds resistance. All of this is just theoretical, I might as well forget about thes numbers and just crank up Vgs slowly and stop when I have the power draw that I want.

At this point I realized I was basically building a variable load with a default resistance of 2 Ohms.
So I watched a bunch of examples online and came up with the attached schematic.

The problems that I'm having are:
- finding a logic level MOSFET that can withstand the remaining (allegedly) 3.5^2 * 6.8 = 80W of power loss
- a suitable OpAmp for the application

Let's consider PJQ4448P as an example.
Vgsth is 2V, which means its a logic level MSOFET, perfect. Then I scroll down to the SOA graph and I see that if I want to switch a 3.5A @ 24V load I'm at the limit of the SOA (preferably stay away, right?). And so I'm left confused, this was supposed to be a 40V 42A MOSFET, what am I missing?


Up until now I've only used mosfets as switches, and not as resistors, so I'm having an hard time making sense of all the different graphs etc.

Can you help me sort this out?
Do some of my calculation have errors in them?
Am I approaching this the right way?


Thanks

Components:
- AC/DC brick: https://www.mornsun-power.com/html/pdf/LD90-23B24R2.html
- dac 10 bit: https://ww1.microchip.com/downloads/en/DeviceDoc/22272C.pdf
- MCU: ESP32

[sapien]

3 things:

You probably don't want waste 80W. Either change the heating plate resistance (increasing total track resistance, untill P=U^2/R is low enough), or the power supply voltage (lowering power supply voltage=U untill P is low enough), in order to get the desired power draw.

Secondly: MOSFETs are mostly designed to operate in the non-ohmic region - that's where the 40V 42A rating is coming from. Generally speaking FETs can't dissipate 80W power loss. Now your SOA shows an amperage and voltage which are related to purely the MOSFET, not the entire system (since it obviously doesn't know or care about the complete system). In your case all the current flows through the MOSFET, however, not the full 24V is across the MOSFET. So IF you want to use a FET in the configuration you mentioned, you have to look at 3.5A @ 8.5V (24-15.5V).

Thirdly: to charge the gate with exactly with the right voltage, to get a voltage drop across it of 8.5V, is really hard. Not only would it need to be fairly precise, but it is also temperature dependent.

Conclusion: change heating plate resistance, or power supply voltage, in order to match P=U^2/R, so you can use your MOSFET as an switch (if that's needed).

[langwadt]

since you are already using an MCU, PWM the fet instead

[janoc]

since you are already using an MCU, PWM the fet instead

Likely not needed, a simple bang-bang regulator turning the thing on and off would be plenty. PWM would certainly require a gate driver.

[langwadt]

since you are already using an MCU, PWM the fet instead

Likely not needed, a simple bang-bang regulator turning the thing on and off would be plenty. PWM would certainly require a gate driver.

with that voltage and low resistance you'd probably want to limit the duty cycle and needing a gatedriver depends on the fet and how fast the PWM is

[Stefano]

Thanks for the quick response guys.

...you have to look at 3.5A @ 8.5V (24-15.5V).

You are totally right, I'm aware of that but I mis-typed. Thanks for correcting me.

Conclusion: change heating plate resistance, or power supply voltage, in order to match P=U^2/R, so you can use your MOSFET as an switch (if that's needed).

So what you are saying is: find a MOSFET with a specific power rating, then use its Rdson to calculate what Amps Volts I can safely run with that power rating, then adjust trace width/length to match those for the temperature increase required and use the MOSFET as a switch with PWM directly.

Is it correct?

[ledtester]

Is the goal to heat the hot plate up to a certain temperature? If so you'll need some sort of temperature sensor.

You can use a very simple algorithm to control the MOSFET -- when the hot plate hasn't reached the desired temp then turn the MOSFET fully on; otherwise keep it off. There are more sophisticated algorithms you can use (i.e. PID control) and you can use an MCU or analog circuitry.

[Stefano]

I indeed plan to implement a PID control in software.
I'm just trying to make the numbers work on paper before proceeding.

Update:
Ok so after fiddling (to say the least) with Saturn pcb tool and mosfets catalogs I think I may have a winner.
I changed the trace width to 2mm and length to 8 meters, still plenty to cover the needed area.
I selected a new MOSFET: PMPB08R5XN https://assets.nexperia.com/documents/data-sheet/PMPB08R5XN.pdf.
I also changed LD90-23B12R2 a 12V 6.75A PSU https://www.digikey.it/it/products/detail/mornsun-america-llc/LD90-23B12R2/16719438

The mosfet I selected can have a max Ploss of 1+W and I will have 0.2W.

On top of that I will be adding an heatsink with a thermal pad on the mosfet.

Using the MOSFET as a switch + PID should do the trick.

I would appreciate if someone could check my calculations, see below.

[langwadt]

do you have extreme space constraints? otherwise that's an odd choice of fet

[thm_w]

On an alu PCB, you won't need a heatsink with 0.2W dissipation.
Whats the maximum temperature you will want to run at?

Calculation looked ok to me (1oz copper total)

[Stefano]

On an alu PCB, you won't need a heatsink with 0.2W dissipation

The mosfet and other driving electronics is on a normal FR4 PCB, which connects with two wires to the Heater aluminum PCB.

Whats the maximum temperature you will want to run at?

I calculated for a 35°C temperature rise at an ambient temp of 20°C, so 55°C.  But I will propably need less, I'm just designing everything with some margin.

Calculation looked ok to me (1oz copper total)

Great, thanks for checking them out.

do you have extreme space constraints? otherwise that's an odd choice of fet

Not particularly. I know its a small FET but is it important? If it has the specs I need and is properly cooled shouldn't be a problem right?
Do you see any down sides to this particular mosfet?
My understanding is that it can afford to be that small despite the high volts/amps ratings because it has a very low RdsOn resistance of LESS than 0.01 Ohm.
I just scrolled the catalog on DigiKey until I found the specs I wanted with a decent price.
Also I plan to put an NTC thermistor near the FET to monitor its temperature.

I guess I'll just go ahead now.

[langwadt]

do you have extreme space constraints? otherwise that's an odd choice of fet

Not particularly. I know its a small FET but is it important? If it has the specs I need and is properly cooled shouldn't be a problem right?
Do you see any down sides to this particular mosfet?
My understanding is that it can afford to be that small despite the high volts/amps ratings because it has a very low RdsOn resistance of LESS than 0.01 Ohm.
I just scrolled the catalog on DigiKey until I found the specs I wanted with a decent price.
Also I plan to put an NTC thermistor near the FET to monitor its temperature.

I guess I'll just go ahead now.

Fet is probably fine, but it the PCB and soldering will be more hassle than needed unless you really need the tiny size

[Stefano]

Fet is probably fine, but it the PCB and soldering will be more hassle than needed unless you really need the tiny size

I agree, but I only have to do it once and I have experience soldering tiny stuff, so I will be fine.
I'm making the footprint now and I'm not having fun 

[magic]

ON/OFF control or PWM won't work if the load draws more power than the PSU can provide.
Heater resistance and PSU voltage must be matched to provide the appropriate power right away.

They may provide a little more power than necessary (and the PSU must be able to cope with it), then digital control can be used to reduce power and regulate temperature.

BTW, an MCU isn't really required for that, it could be a simple analog comparator.

[Stefano]

ON/OFF control or PWM won't work if the load draws more power than the PSU can provide.
Heater resistance and PSU voltage must be matched to provide the appropriate power right away.

Indeed my PSU can handle the required load.
I'm using LD90-23B12R2 PSU, which can deliver up to 6.7A @ 12V.
The heater is a 2.6Ohm resistive load, so when the mosfet is fully ON it is drawing 4.6A, which is 68% of the max current the PSU can deliver, effectively giving me a 32% margin. This seems safe enough to me.

The only thing that I still have to figure out is how to calculate how much capacitance I need between the PSU and the heater to be able to switch the power ON-OFF without problems, if that makes sense. The datasheet for that PSU says to put a 1uF-100V MLCC and a 330uF-35V electrolytic cat on the output, which I did, so hopefully, giving the 68% load is enough. BTW I ended up using a 470uF electrolytic instead, just to be on the safe side, although I need to study this matter more. Thoughts?

BTW, an MCU isn't really required for that, it could be a simple analog comparator.

I wanted to learn how to use an ESP module, so that's my reason

[thm_w]

470uF is fine. The output capacitance is more to reduce ripple, which you won't care too much about, as you are driving a resistive heater.

Output Filter Components:
We recommend using an electrolytic capacitor with high frequency, and low ESR rating for C2 (refer to manufacture’s datasheet). Choose a Capacitor
voltage rating with at least 20% margin, in other words not exceeding 80%. C1 is a ceramic capacitor used for filtering high-frequency noise and TVS is a
recommended suppressor diode to protect the application in case of a converter failure.

[Stefano]

470uF is fine. The output capacitance is more to reduce ripple, which you won't care too much about, as you are driving a resistive heater.

Great!

PWM would certainly require a gate driver

My understanding is that you need a gate driver when you need high current draw/sink capabilities in order to turn the MOSFET on/off very very fast and precisely.

In my scenario I will most likely use a PWM signal with a period is the order of ms, not ns, and also I don't need the turn on/off event to be super precise, it can vary of some and I wouldn't feel a difference in heat, I think.

As of now I'm using RQ3E110AJ, which has a total gate charge of 13.5 nC, and the MCU has a max current draw of 40 mA per pin.
With these informations, I can use I=Δgate-charge/Δtime  -> Δtime=Δgate-charge/I  -> 13.5 nC / 40 mA to get a gate charge up time of... 0.3 seconds?
Wait this can't be right, I was hoping to switch the FET faster than 3Hz 
What am I doing wrong? Did I mess up the units in the formula?

[langwadt]

13.5 nC / 40 mA -> 13.5e-9/40e-3 = 0.3375e-6

[Stefano]

Of course I messed up the scale, thanks for that.