Webcontrol - timing errors?

[CAI_Support]

Curious how the min readings on the red and the max readings on the blue correspond to what appears to be the true readings.

Lou,

WebControl coutner is very accurate. However, since CPU is doing a few things all the time, it may miss a clock advance from time to time. The frequency measurment is divide the counter value by the time lapsed. Since we know the counter is accurate, the time could be actually longer by tiny bit, F = Counter/Clock, so the lower F is more accurate.

The variation is very small. When Ross talked about 20Hz jitter over his 120KHz clock source, the inaccuracy is 1.6/10000 or 0.016 percent.

 

[rossw]

The variation is very small. When Ross talked about 20Hz jitter over his 120KHz clock source, the inaccuracy is 1.6/10000 or 0.016 percent.

In the current field my source is only running at around 67KHz, so its actually much worse than that - but 0.03% is still pretty close

 

[CAI_Support]

Ross,
If you don't need that frequency measurement, you can use it with a Voltage to Frequency converter chip to make that an analog input.

 

[rossw]

Ross,
If you don't need that frequency measurement, you can use it with a Voltage to Frequency converter chip to make that an analog input.

Hardly! The analog input is only 10 bits. Even with no jitter at all, that's 0.1%

 

[fwd03]

Burr Brown's VFC320 is 12-14 bit accuracy 0.03% at 100KHz
National LM331 is 0.01% accuracy also up to 100KHz
Both are part of TI now.

 

[rossw]

Burr Brown's VFC320 is 12-14 bit accuracy 0.03% at 100KHz
National LM331 is 0.01% accuracy also up to 100KHz

The VFC320 will convert voltage to frequency. I have a variable frequency source. It has a very wide dynamic range, from approx 48KHz to in excess of 150KHz. The problem is, depending on the field, the variation I'm actually looking for is in the order of 0.5Hz

So it could be from (say) 48123.0 to 48124.0Hz, or it could be 145220Hz to 145221Hz, or anywhere in between.
We're not talking "trivial" data acquisition here.

Even if I were able to read within 0.03%, at 150KHz that's 45 Hz uncertainty - almost 2 orders of magnitude more uncertainty than the data I'm trying to measure!

 

[fwd03]

So it could be from (say) 48123.0 to 48124.0Hz, or it could be 145220Hz to 145221Hz, or anywhere in between.
We're not talking "trivial" data acquisition here.

Just curious what is the frequency source? To have that kind of measurement accuracy, you will for sure to have constant temperature for the measurement equipment, or you have to write a program with calibrated curve against the temperature change.

 

[rossw]

Just curious what is the frequency source? To have that kind of measurement accuracy, you will for sure to have constant temperature for the measurement equipment, or you have to write a program with calibrated curve against the temperature change.

The instrument is a very sensitive flux-gate magnetometer.
At the moment its just in my lab - I had it in a null-chamber initially, but have now taken it out and just got it in a minimally-used area where normal activity doesn't bother it. I do have a temperature sensor with it, and have some approximate compensation curves programmed in now. Drift across all components - its own power supply, the FGM etc, works out at around +0.05%/degC. I still have problems with local fields disrupting things - if an automobile parks within 30 metres of it, it makes a substantial change. I took the lawnmower out, it was about 10 metres away but it made a substantial change too. In order to minimise this, I will be burying it several feet down for thermal stability, probably 50 metres or so away from the house away from vehicles etc. But I need to get the noise as low as possible before then, otherwise I'm just not going to see anything of relevance.

I have adjusted the one-wire bus timing, but temp sensor 2 still refuses to behave. I will try another sensor just in case, but it wasn't a problem on the original board....

 

[CAI_Support]

Ross,

You just prompted me to think your design can be used for room occupancy detector. If a piece of metal as capacitor for the frequency generator, anybody walking into a room, that will cause the frequency changing a lot. Once the person left, the frequency will go back to its stablized value. You could write a PLC code to detect longest time period with no frequency change, that must be the room empty time.

Normally the higher the frequency, the bigger the change can be. Just need to figure out a low cost stable frequency generator.

What do you think?