Thoughts on a flowbench computer

[Moriniman]

I'm just putting together some preliminary ideas for a flowbench DAQ/controller.

I've got some Honeywell sensors to hand; 26PCBFA6D 0-5psi differential temperature compensated.

Analogue Devices AD7705 16 bit differential 2 channel A2D converter with an input span down to 25mv full scale so it will interface straight to the sensors.

Microchip PIC 16F876 becuase I'm familiar with it, have the compiler and an ICE.

[106-1194218389]

I personally think you are on to it, good work!

[Tony]

Yup, have to agree. Everything there you need to do the job well.

[Moriniman]

How important do people feel it is to have air temperature measurement?

Measuring temperature to a resolution of 0.1 Celcius is fairly easy, measuring to an accuracy of better than 1 degree is much harder (quite a lot of scientists seem ignorant of the difference between accuracy and resolution).

If the purpose is to correct readings taken on one day to those taken on another, then resolution rather than accuracy is more important.

Thermocouples are quite cheap and Maxim do an integrated cold junction compensator and 10 bit A2D converter (MAX6674) giving a resolution of 0.125 degree and a basic accuracy of +/-2 degrees.

[86rocco1]

With a "normal" orifice type bench, temperature measurement isn't at all critical because you don't need to do air density corrections. However with Pitot tube benches or with a central blower orifice benches like the SF110 where air density corrections are required, temperature is important.

In the past, I've used thermistors to measure air temperature, with the proper transfer function, they are easily accurate within + 1 degree. After calibrating the transfer function with the help of an ice bath, I compared one of mine to a meterological grade mercury thermometer that reads to + 0.1 degrees, the thermistor reading was never out by more than 0.3 degrees over the normal outdoor temperature range.

[Tony]

The easiest way to electronically measure temperature is with something like an LM34 or LM35 temperature sensor.

www.ece.osu.edu/~passino/LM35.pdf
www.investigacion.frc.utn.edu.ar/sensor ... AN-460.pdf

The LM35 outputs ten millivolts per degree Celsius directly, and needs no corrections or scaling. Quoted accuracy is half a degree.

There is also the LM34 which does the same thing, but it outputs millivolts directly in the Fahrenheit scale.

With either of these sensors, (plus a battery), a digital multimeter, or data acquisition system, and you have simple reliable temperature sensing.

It really is the easiest way, and these sensors are not expensive.

[Moriniman]

I've had the pcb built for a while, but hadn't got any code done. I've brought the board home and borrowed one of the spare PIC emulators (we're using ICEPIC IIIs now).



I'm using one ADC per pressure sensor for faster data aquistion, but two more sensors can be fitted. My thinking it that you're either reading the drop across the orifice, or using a velocity probe. There is provision for a stepper motor and two PWM channels. The currently unused internal PIC ADC channels (10 bit) are brought out to an edge connector.