User avatar
Hove
Posts: 1205
Joined: Sun Oct 21, 2012 6:55 pm
Location: Cotswolds, UK
Contact: Website

Mass and specific heat capacity of a chip

Sat Oct 04, 2014 2:41 pm

Anyone able to give me order of magnitude values for the mass and specific heat capacity of a 4mm x 4mm x 0.9mm MPU-6050 IMU chip (or to be honest any small silicon SMD)?

Why do I want to know? The sensor outputs vary by temperature - error about 2%. I've calibrated the unit at a couple of temperature settings, and that allows me to estimate calibration at intermediate temperatures. That gives me about 0.1% error values. This is as good as I can get over a broad temperature range as the sensor output is also only 2% linear against temperature. But it's not good enough 0.1% error on gravity is the difference between my quadcopter hovering, or accelerating up or down at a cm /s /s - and that limits flight times to ~10s before she hits the ground or starts whizzing up to the stratosphere!.

So instead, I'm considering adding a peltier effect heat pump directly onto the chip, and using the chips temperature sensor as feedback to maintain a stable temperature in the chip, and thus stable sensor outputs which then can be calibrated to read gravity (for example) as 0, 0, 1g across the X, Y and Z axes respectively regardless of ambient conditions.

I'd like to power the heat pump direct from a GPIO pin (3.3v 8mA) with PWM to control the power applied to the heat pump.

So I need approximate mass and heat capacity to calculate the energy required to modify the temperature, and from there the PWM pulse widths supplied to control the pump.

I could just do this experimentally, but at the risk of frying the sensor, the heat pump or both (£50 if both go!), so instead I'd like to have ballpark figures with which to start my experimentation.

Thanks

P.S. I know I could use ultrasonic range detection, but that'll on work for vertical up to about 6m. The sensor errors are also present horizontally, leading to horizontal drift and in an open space, a set of horizontal range detectors are useless as a chocolate teapot to prevent this.
www.pistuffing.co.uk - Raspberry Pi and other stuffing!

User avatar
aTao
Posts: 1086
Joined: Wed Dec 12, 2012 10:41 am
Location: Howlin Eigg

Re: Mass and specific heat capacity of a chip

Sat Oct 04, 2014 3:31 pm

Hove wrote:Anyone able to give me order of magnitude values for the mass and specific heat capacity of a 4mm x 4mm x 0.9mm MPU-6050 IMU chip (or to be honest any small silicon SMD)?
If its SMD then the values you get for those properties are not particularly relevant since the pcb has a really significant effect.

So, you are going to have to do it empirically. So as not to fry a£50 chip then you could test with other dead or cheap chips. Or you could experiment with low power first. BTW a peltier heat pump is a strange choice here, a resistor will cheaper and lighter and smaller and you can change the value (to give different power ranges easily).
>)))'><'(((<

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: Mass and specific heat capacity of a chip

Sat Oct 04, 2014 3:48 pm

I'd calibrate over more points, and use a PWL to interpolate between each calibration point. Easier than trying to keep the chip at the same temperature. More accuracy can be gained by more calibration points.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
Hove
Posts: 1205
Joined: Sun Oct 21, 2012 6:55 pm
Location: Cotswolds, UK
Contact: Website

Re: Mass and specific heat capacity of a chip

Sat Oct 04, 2014 4:10 pm

jamesh wrote:I'd calibrate over more points, and use a PWL to interpolate between each calibration point. Easier than trying to keep the chip at the same temperature. More accuracy can be gained by more calibration points.
Been there, done that although with using Excel to generate a trend line across the readings (minimal RMS error linearity). What's PWL?

Capture +/- 3 axes orientations at a given temperature is tricky as the chip is self heating. Also finding sufficiently temperature stable environments is hard - I've used a beer fridge but it's still tricky to temperature span under 0.5 degrees to measure gravity in 6 DOF.

Hence my experiment to try turning the chip itself into a temperature controlled environment and calibrate at that fixed temperature. No guarantee it'll work, but worth experimenting with!
www.pistuffing.co.uk - Raspberry Pi and other stuffing!

User avatar
Hove
Posts: 1205
Joined: Sun Oct 21, 2012 6:55 pm
Location: Cotswolds, UK
Contact: Website

Re: Mass and specific heat capacity of a chip

Sat Oct 04, 2014 4:16 pm

aTao wrote: If its SMD then the values you get for those properties are not particularly relevant since the pcb has a really significant effect.

So, you are going to have to do it empirically. So as not to fry a£50 chip then you could test with other dead or cheap chips. Or you could experiment with low power first. BTW a peltier heat pump is a strange choice here, a resistor will cheaper and lighter and smaller and you can change the value (to give different power ranges easily).
Peltier is necessary as my aim is to keep the sensor chips internal temperature constant, using the internal temperature sensor as feedback. The heat I'm dissipating is generated by the chip itself, along with ambient conditions (i.e. sunlight, shadow, seasonal variants etc).

I appreciate the significant impact of the PCB sapping heat, which is why I only want ballpark figures.

Not sure how a resistor would be using instead of the peltier?

I think you're right, empirical test could be done safely. Starting with a > 500 ohm (~ 3.3v / 8mA) resistor powering the Peltier from PWM GPIO pin, I can check the cooling effect of the pump at various PWM pulse widths without drawing more than 8mA from the GPIO.

Thanks
www.pistuffing.co.uk - Raspberry Pi and other stuffing!

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 22733
Joined: Sat Jul 30, 2011 7:41 pm

Re: Mass and specific heat capacity of a chip

Sat Oct 04, 2014 5:02 pm

Hove wrote:
jamesh wrote:I'd calibrate over more points, and use a PWL to interpolate between each calibration point. Easier than trying to keep the chip at the same temperature. More accuracy can be gained by more calibration points.
Been there, done that although with using Excel to generate a trend line across the readings (minimal RMS error linearity). What's PWL?

Capture +/- 3 axes orientations at a given temperature is tricky as the chip is self heating. Also finding sufficiently temperature stable environments is hard - I've used a beer fridge but it's still tricky to temperature span under 0.5 degrees to measure gravity in 6 DOF.

Hence my experiment to try turning the chip itself into a temperature controlled environment and calibrate at that fixed temperature. No guarantee it'll work, but worth experimenting with!
Piecewise linear interpolation - in effect lots of short linear interpolations between each data point in the set. Approximates a non-linear graph. Used a lot in the GPU camera software. I'd certainly consider that as opposed to keeping the chip at a stable temperature. In actual fact we used it in the Nokia 808. The black level changes with temperature, and we used a PWL to set it according to the internal sensor temperature. Ring any bells?
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
"My grief counseller just died, luckily, he was so good, I didn't care."

User avatar
Hove
Posts: 1205
Joined: Sun Oct 21, 2012 6:55 pm
Location: Cotswolds, UK
Contact: Website

Re: Mass and specific heat capacity of a chip

Sat Oct 04, 2014 5:30 pm

jamesh wrote: Piecewise linear interpolation - in effect lots of short linear interpolations between each data point in the set. Approximates a non-linear graph. Used a lot in the GPU camera software. I'd certainly consider that as opposed to keeping the chip at a stable temperature. In actual fact we used it in the Nokia 808. The black level changes with temperature, and we used a PWL to set it according to the internal sensor temperature. Ring any bells?
Ding dong!

Definitely worth trying, thanks, though a little fiddly to get a wide range of temperatures to keep the PWL as accurate as possible across a broad range - easier as winter draws in and the heating in the house is on. I'll still try the peltier option as it seems a cool approach ;)
www.pistuffing.co.uk - Raspberry Pi and other stuffing!

User avatar
aTao
Posts: 1086
Joined: Wed Dec 12, 2012 10:41 am
Location: Howlin Eigg

Re: Mass and specific heat capacity of a chip

Sat Oct 04, 2014 5:35 pm

Hove wrote:
aTao wrote: If its SMD then the values you get for those properties are not particularly relevant since the pcb has a really significant effect.

So, you are going to have to do it empirically. So as not to fry a£50 chip then you could test with other dead or cheap chips. Or you could experiment with low power first. BTW a peltier heat pump is a strange choice here, a resistor will cheaper and lighter and smaller and you can change the value (to give different power ranges easily).
Peltier is necessary as my aim is to keep the sensor chips internal temperature constant, using the internal temperature sensor as feedback. The heat I'm dissipating is generated by the chip itself, along with ambient conditions (i.e. sunlight, shadow, seasonal variants etc).

I appreciate the significant impact of the PCB sapping heat, which is why I only want ballpark figures.

Not sure how a resistor would be using instead of the peltier?

I think you're right, empirical test could be done safely. Starting with a > 500 ohm (~ 3.3v / 8mA) resistor powering the Peltier from PWM GPIO pin, I can check the cooling effect of the pump at various PWM pulse widths without drawing more than 8mA from the GPIO.

Thanks
All depends if the chip needs cooling to be stable or heating. I would be surprised if its natural temperature left no head room for a bit of reliable heating. For instance if its natural operating temp is 30~40 C then IIRC from the spec you have the range 40~70 C to pick a temp to maintain by heating. Peltiers are ludicrously inneficient

But, to be honest, if James's method works then that would be the best solution by a long chalk.
>)))'><'(((<

User avatar
Hove
Posts: 1205
Joined: Sun Oct 21, 2012 6:55 pm
Location: Cotswolds, UK
Contact: Website

Re: Mass and specific heat capacity of a chip

Sat Oct 04, 2014 6:52 pm

Don't get me wrong, PWL interpolation should be able to produce a better accuracy solution. But it just doesn't float my boat.

The aim of my quadcopter project is not to make a toy that I can fly. The aim is wholly autonomous flight, and adding the peltier for sensor accuracy adds to the autonomy despite it being a completely OTT solution!

So peltier is the way I'm going for now, but if it sinks like a fish in prototyping, at least I know I can improve my linear calibration by using PWL interpolation - with winter approaching the available temperature range is starting to grow nicely.

Cheers for your thoughts, suggestions and ideas.

Hove
www.pistuffing.co.uk - Raspberry Pi and other stuffing!

grahamed
Posts: 277
Joined: Mon Jan 30, 2012 7:01 pm

Re: Mass and specific heat capacity of a chip

Sun Oct 05, 2014 3:27 pm

Hi

This sort of problem is commonly faced when trying to make a crystal oscillator run at a constant frequency - in general the crystal is significantly temperature sensitive, Google for "crystal oven". Always the idea is to run the crystal above ambient - maybe 40-60C- so it always needs heat and always the heat source is a resistor - never seen a Peltier device in use.

User avatar
Hove
Posts: 1205
Joined: Sun Oct 21, 2012 6:55 pm
Location: Cotswolds, UK
Contact: Website

Re: Mass and specific heat capacity of a chip

Sun Oct 05, 2014 4:11 pm

Oooh, interesting! I hadn't understood the mention of using a resistor instead of Peltier in an earlier response, but now I do, thanks. I'd been concentrating on heating or cooling with Peltier to say 20 degrees, instead of heating to say 40 degrees (i.e. hotter than even summer ambient)

Simple op-amp, feedback from thermistor, output driving resistors (via transistor) to provide heat. Simples!

Except for 1 or 2 minor details
  • chip to cool is 4mm x 4mm x 0.9mm - not much space to fit a thermistor and heating unit whereas I have found a 4mm square Peltier
  • analogue solution doesn't involve the Raspberry Pi :D
Thanks though that does give me a much simpler solution to explore.
www.pistuffing.co.uk - Raspberry Pi and other stuffing!

Ravenous
Posts: 1956
Joined: Fri Feb 24, 2012 1:01 pm
Location: UK

Re: Mass and specific heat capacity of a chip

Mon Oct 06, 2014 8:59 am

I'm a bit late, but just an additional note, I've heard of the "oven" idea used a lot in older-style analog synthesizers. Some of these are voltage controlled and through an accurate log-linear converter. Temperature stability is (was) critical in the converters.

So it's not a completely outlandish idea. I think it was just a power resistor epoxied to the relevant semiconductors in a little cluster. Not sure if they were precisely temperature controlled or not.

(More modern ones I've seen use a temperature compensated idea with a tempco/thermistor but that's a different thing.)

User avatar
Hove
Posts: 1205
Joined: Sun Oct 21, 2012 6:55 pm
Location: Cotswolds, UK
Contact: Website

Re: Mass and specific heat capacity of a chip

Mon Oct 06, 2014 10:09 am

Ravenous wrote:So it's not a completely outlandish idea. I think it was just a power resistor epoxied to the relevant semiconductors in a little cluster. Not sure if they were precisely temperature controlled or not.

(More modern ones I've seen use a temperature compensated idea with a tempco/thermistor but that's a different thing.)
I'm actually going to be using a 'digital' version of the op-amp / thermistor / heated resistor feedback model.

The MPU-6050 already has a temperature sensor at it's core (i.e. where it's needed). The heating element will be an SMD resistor (20 ohms or so) fixed by thermally conductive tape (used for attaching heatsinks) with a MOSFET switching current driven by PWM from the Raspberry Pi - a little bit of PID code to tie these together and Bob's Your Mother's Brother!
www.pistuffing.co.uk - Raspberry Pi and other stuffing!

Ravenous
Posts: 1956
Joined: Fri Feb 24, 2012 1:01 pm
Location: UK

Re: Mass and specific heat capacity of a chip

Mon Oct 06, 2014 10:41 am

It's funny really :) - we're in the 21st century and we still, sometimes, need a heater to make sure electronics are at the right temperature.

In the 50s it was valve equipment that had to warm up, in the 60s it was a fixed resistor to raise the temperature to roughly the right region, in the 2010s it's computerised PID loops to regulate the applied heat. :lol:

I'm not complaining, I just find it entertaining that we still sometimes have an excuse to do this. :)

Return to “Automation, sensing and robotics”