I have decided to bite the bullet and use the ExpressPCB service so now the boards are on the way to me. Unfortunately I ordered the PCB with the PWM trace error in it. Fortunately it is a very easy board fix, the trace in question run right past the pin it is supposed to connect to.

Now I am assembling a bill of materials and would like some opinions on a suitable PWM frequency for driving fans. As I have it it is running between 250 and >1200Hz depending on gain setting. Is this reasonable?

Some raw specs based on current component values:
Voltmeter 0-23.34V, +/- 2.9mV
Ammeter 0-11.9A, +/- 1.5mA
Giving a Wattmeter 0-278W +/- 0.009mW
Accuracy unknown, component values will be 1%.

4 temperature channels
-40-150°C +/- <0.2°C
0-85°C +/- <0.025°C
18-55°C +/- <0.015°C

2 high resolution temperature channels
15-35°C +/- <0.0025°C

Accuracy unknown, subject to component drifts, tolerances, noise control and accuracy of calibration reference. Who knows? I'll see how the oscilliscope, data noise and repeatability look when it's done.

Nice work!

Do you have any idea if you could gain more accuracy from having two temp channels measured in differential mode?

Otherwise, I just might join you in this setup. Let us know how it works out, ok? ;)

Probably could. Differential inputs would normally be preferable but for the fact that this particular ADC does not use a true differential mode. This pseudo-differential mode I was not keen on messing about with, it can give some snapping problems at certain input values apparently.
Also I did not want to "waste" two ADC channels on a single input.

With a bit of software I could actually use two of the thermistor channels as they are in differential mode and obtain a delta T reading directly, for example for the water sensors. This might be a more accurate way to get the critical delta Tin-Tout for heat-to-water power calculation. Less sensitive to thermistor ref voltage noise. But a bitch to calibrate.

The PCBs arrived today. Three of them. $51. yikes. shipping $44. Holy SHITE!! :eek:
http://w1.863.telia.com/~u86303493/C...iesim/pcbs.jpg


Also Texas Instruments have an even better samples program than Maxim, I got six INA122's from them, three day shipping from US to Sweden. Amazing.

This mean's that all the Semiconductors which would have cost me in excess of $120 if I had bought them from a local supplier are free samples. More than makes up for the PCB cost.


:)

A little late but :shrug:

As part of testing out a function generator I just got from the junk shop, I hooked it up to give me a dial-a-frequeny PWM. The whole range of 1 Hz - 100kHz worked; the sweet spot for low noise and wide duty cycle range seemed to be ~900 Hz.

be interesting to get a ballpark on the PWM added noise vs. the reduction in fan noise
gonna be different though for all fans apparently

Silence isn't goal, controlling the water temperature is. Not having the fan screech is nice, but having it willing to run at <20% duty cycle is more useful.

Bill, my sample of three fans all had different ratios of turbulence noise vs. PWM noise. Total noise seem to decrease (measured with an uncalibrated ear) with reduced duty cycle, but noise vs. CFM certainly increased.

If control weren't the big issue, I'd say 25 kHz or more. And maybe pulse between 12V and 3V (not ground), so that the fan's electronics would stay powered. Projects for another day.

I don't think it was gibberish, that's what many report - no ?

My statement was based on "insufficient data". Was forgetting that PWM allows lower RPM than DC.

(sorry for the threadjack)

groth, my remark was more an off the cuff reaction to Bills' comment - and applying more to the general case of manual speed control. Sure, if you're using some form of PID and want accurate temperature control, then PWM is convenient + efficient and will also get you less non linear distortion.

Backing up my initial statement that PWM is a bad solution if you are a noise freak, I would add that if you need to run your fans at 20% then maybe swapping them out for lower speed versions would buy more dB :) Otherwise just filtering the PWM output with a cap would work, I guess.

Not my thread, hijack away! :D

Yeah, I figured that you (and Bill) were thinking of the general case of speed control. I don't expect everyone to keep track of the the details of every thread, hence my pointing out control was the overriding goal. Never meant to sound harsh or dismissive. I agree completely, PWM isn't for the noise freaks.

To continue farther afield...since the fan motor has some inductance, once you add your capacitor on the output, all you need to add is a freewheeling diode to make a crude switchmode supply. :)

An update.

Basically the last few weeks have been spent ironing out little faults, (a bad solder joint, a 2 1/2 year old's intervention which blew the digital circuitry etc)

I have connected the digital ground to the analog ground. There was an offset in the raw output of about 4-5 LSB with it disconnected with noise levels the same so I figured it should be there.

I have spent a fair bit of time generating Steinhart-Hart coefficients for the thermistor calibration. I have done this by using 0.01 deg C (crushed ice and water), 100 deg C (boiling point) and ~37 deg C (water at about body temperature as measured using a medical thermometer, multiple samples taken as an average). This has generated curves for the four low resolution channels which ensure that they agree to within +/- 1LSB (~0.025 deg C) when taken as an average, over a range from ~20-55 deg C.

I am amazed at how well this unit works. I need to check linearity but I am confident that the S-H calibration coefficients are giving a good result, it is an industry standard method and typically has errors of less than 0.01 deg.

This document has a lot of very useful info.

I have made a new fluxblock. It has 3 sensors spaced 5mm apart to check the linearity of the measured heat flux. This one is 12x12x12mm, a 10x10x12mm version is planned. I have also made a matching heat die with one sensor. This will allow me to finally measure the TIM resistance with a fair amount of precision by interpolation.

Two water sensors, with a resolution of <0.005degC will let me crosscheck power to water vs power to WB vs Power to heat die. Secondary and tertiary losses will come from these numbers, curiosity only, no practical value, only applies to this setup but interesting to see the effects of insulation.

Still haven't played with the PWM circuitry, it will come.

Very cool... rest assured I'm keeping a close eye on your progress! ;)

...and sure enough.

I picked up a six-pack of AD7713AN
"CMOS, Low Power 24-Bit Sigma-Delta, Signal Conditioning ADC with Matched RTD Current Sources"
http://www.analog.com/en/prod/0,2877,AD7713,00.html

...and two AD580 and two REF192FS.
http://www.analog.com/en/prod/0,2877,AD580,00.html
http://www.analog.com/en/prod/0,,769_838_REF192,00.html

Now...

I've already got too many things going on (as if no one knew!) so this is probably going to sit on ice for a while, but I thought I'd put the offer forward; if anyone is interested in building a similar circuit, I'll supply the ADCs and the volt references. Heck, I'll even cover the cost of the PCBs.

Pretty interesting read here. ben linked it from another thread and just got around to reading it. This reminds me of another DAQ based on the maxim186 I cam acrossed called the MiniDAQ. http://www.embeddedtronics.com/Electronics.html

Anyway my question to Incoherent is what software are you using with it. Sorry if I missed it.

http://www.electronic-kits-and-proje...8_software.htm
ISEE DAQ software for the kit that the digital side of this DAQ is based on.
It is quite OK functionality-wise but a little limited in data processing power.
I have not had time to set up a Labview reader, but that is the plan longer term, that and moving to 16bits, I'll probably use one of the National instruments DAQs.

Incoherent, I have one of the 3118 kits and sucsessfuly put it to gether. I have tried getting it to monitor the volts through a 10k POT so i can adjust the volts but in the software it never changes. If i remove the power the software then shows a drop down to 0v.

Could you do a simple 3118 for dummys guide like me ? All i want to do is data log my cpu , vga , water in , water out and case temp. I dont need to be as precise as your system just to get a good idea of whats going on.

I have 100k bead thermisters for probes left over from a joystick port temp reader project will they be ok.

How are you wiring the pot? This might be stupidly obvious but, One end should be at +5V, the other at ground (0V) and you should monitor the voltage at the wiper.
The 100k bead thermistors should be fine.
Put them each in series with 40k resistor a between the +5v and 0v with the thermistor closest to the +5v and monitor the voltage at the connection. The curve will look something like the attachment. To reduce the linear temperature range increase the resistor value.

The 100k Pot works perfect thanks :)

I dont have any 40k resistors how would i use the pot ?

at the moment i put the 100k thermistor on the center of the pot and then to the 3118. all i get is a wave and the amplitude increases when i touch the thermistor.
I am going to look though this thread again i am sure i saw something like this in there.

(edit) i put the thermistor across the pot 5v and the output and it works ok.
I tried running the power of the pc psu and a switch mode psu but they are both the same.

Is it a 10K or 100K pot? If it is a 10k you need to get a larger value resistor because no matter how you do it it will be very non linear in the temperature range you are working with.
Wired the way you have it your range is limited, you are not using the full resolution that you can achieve. If its a 100k pot wire it as attached, set it at half way and do not touch it. Then you can start thinking about calibrating.
You will find that you can change the value by adjusting the pot, it might feel like a way to calibrate it but it is not, it is just going to destroy the linearity of your readings.

they are 100k and they are working ok. I have 4 sensors running but the pots need to be replaced with resistors.

thanks again :)

I used this website before for calibrating the thermistiors.

http://www.benchtest.com/gp_Temp3.html

there is a xls spread sheet for calculating the curve of the thermistors. Will the mv reading work in place of the joystick port readings ?

steinhart.xls

http://www.benchtest.com/images/s&h1.jpg

I am trying to find an old program i wrote that read from the joystick port and had the xls functions built in to.

Absolutely.

You just need to calculate the resistance of the thermistor at the three given temperatures from your readings. Knowing the +5v, the voltage at the thermistor (your DAQ reading) and the resistance of the series resistor, you can figure this out. (5V-Vtherm)/(Vtherm/R)=Rtherm I think. Plug in the Resistance and calibration temperature for the three temperatures and you're good to go.

Good excel sheet. I did all my calculations myself but I just checked it against that sheet and it give the exact same answers for the SH-H coefficients using a slightly different method.

The ISEE software control language lets you put these equations into it so you can get a calibrated temperature output directly. It's a bit buggy but works, just don't feed it divide by zero problems or it will crash.
A sample might look something like this:

@RES1=(5 + (@CH1/1000)*-1)/((@CH1/1000)/40000) =====> (the thermistor resistance.)
@LOG1=LOG(@RES1) =====> (natural log of the resistance)
@TEMP1=0.00147408 + 0.00023704*@LOG1 + 0.00000010839*@LOG1^3 =====> (A,B and C coefficients, the steinhart-hart equation )
Tags: @CH1 is ADC input, @RES1, @LOG1, @TEMP1 are analog variables

There are nicer ways but this works.

i have 4 probes working using 100k@25C thermistors.
I am working on the calibration and have 2 very close but 1 is a little off

this is the calibration temps

http://www.ljsnet.co.uk/temps.gif

I left the probes in the hot water until they cooled down and this is the results.
you can see the pink line is higher then the other 2 even though i calibrated them all at the same time.

Some think not right with the pink reading they are a bit all over the place compared to the others. I wonder if i put the correct ground on this one.

http://www.ljsnet.co.uk/temps2.gif

I have a couple of probes from a lian li 5 1/4 drive bay temp unit. they look like they are 50k at 25C will be getting them working soon so i can use them on the cpu as the end of the probes are very thin comapred to the bead thermistors

thanks for all your help i am working an a vb probram to capture the probe data and apply the curves automaticaly plus have a calibration screen.

Looks good apart from the ringing. What time period are you logging at?
Try putting an RC filter on the ADC inputs, i.e. connect a 1k resistor in series with your signal line with a 0.47µF cap to ground after the resistor. This won't affect your average reading and will stabilise the signal, its a lowpass filter.
The temp reading difference might be caused by errors in the real resistance of the resistor in your voltage divider (the one in series with the thermistor). Measure it if you can and adjust the value in the resistance calculation. Otherwise, if there is no major linearity problems i.e. the difference is constant, you can make small adjustments to the A coefficient value to tweak the offset away.


Adding a pic of a clip.

Not relevant, just don't want to start a new thread or bump this one

I have this working a lot better after i checked the circuits and found a couple of mistakes. Also i found that if i have a certain power supply switched on the ringing gets worse.
I have the vb program working and can calibrate the reading very easy with it just need to add things like logging to a file or spreadsheet.