km

PART 1:

In that case, do I need to use a variable resistor for my active low pass filter design? (for the 2nd opamp)

PART 2:

For my project, do I need to built an extra circuit before feeding the A2D output to the PIC 16F628? Because the data from the A2D is going to be very difficult to analyze and I might have difficulty in obtaining the results (i.e. frequency, average, peak-to-peak).

As for the external A2D I used (TLC 548), from what Nigel had said:"it should be fine to use that". (But from what I heard, the accuracy required for such applications is critical) Am I able to get accurate results with the external A2D (TLC 548)? Should I consider another A2D chip? (maybe an A2D 10-bit resolution):?:

PS:I'm getting more worry about my project day by day as I only left 2 months to do my project :cry:

Nigel Goodwin

Only if you want to make it variable.

No, the output from the A2D will feed directly into the PIC, you just need to write the software to control it.

As I see it you are going to have to perform considerable calculations on the incoming data - I would suggest you write a program on the PC to perform these calculations and feed it dummy data, that way you can confirm it works, and check what resolution will suffice. There's no point building hardware when you don't even know if it will be possible or not.

If you've only got two months to complete this, and you don't appear to have started work yet, I would suggest a lot of prayer might be in order :roll:

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

km

:shock: A lot of prayer? LOL :lol: Thanks for your advice!!

How can I do that? What software do I need? :idea:

Nigel Goodwin

Write in whatever software you are experienced in, Delphi, BASIC, C, Pascal, Prolog, Java - any language you like, just to prove it works, and to test what resolution works best. I would expect this sort of pre-planning and testing to be part of your documentation, and it all helps towards good marks.

You also have the problem of how many samples you need to give a sufficiently accurate reading - this is further compounded by the lack of RAM in mid-range PIC's. An obvious easy value (for an 8 bit processor) would be 256 samples, but you don't have anywhere to store them in a 16F84. Again, you can use the PC to test different numbers of samples, to see how few you can get away with - and again, this should all be documented.

If you look on the PICList there's a project for performing FFT (Fast Fourier Transforms) using a PIC, but it uses a high end PIC, with lots more memory and more power.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

km

Exo

Yes, 16f628 only has 224 byte of ram, still not enough (you're probabely need some of that for the rest of the software, not all ram can be used to store samples)

Nigel Goodwin

I agree with Exo, I suspect you would be better off with a high range PIC (18F series). But first, as I've already mentioned, try running it on a PC, and see how many samples you need to calculate what you are wanting to do - you 'might' be surprised.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

km

Exo, thanks for spending your time to read my post!! :wink:

If in that case, which PICs should I used? (without a built-in A2D)
How about PIC 16F87 OR PIC16F648A? Both of these have twice the program space as the PIC 16F628. :idea:

km

Exo, thanks for spending your time to read my post!! :wink:

If in that case, which PICs should I used? (without a built-in A2D)
How about PIC 16F87 OR PIC16F648A? Both of these have twice the program space as the PIC 16F628. :idea:

Nigel Goodwin

Do the preliminary work first, and find out what you need - then look for a processor which will fit those requirements. No point in choosing a processor to find out later it won't do what you want.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

km

Thanks for all the advice, Nigel!! 8)

One more question about testing the program on PC. After I have written a program that works, how am I going to test what resolution work best and how many samples I need to calculate? :roll:

PS:I apologize for my poor understanding.

Nigel Goodwin

Basically you need to generate data for different waveshapes, a sinewave is the obvious first choice, and you can generate that from the Sin() function included in most high level languages. As you're planning an 8 bit A2D, the maximum value can go from 0-255, so you need to write the generation routine accordingly - scaling it to give results in that range.

For a start I would try generating a run of 256 samples, and try your maths on that - then try 128 samples, and see if you get the same result, then 64 samples - you can do this easily by only using every 2nd sample, or third sample etc.

This simulates different numbers of samples, while keeping the data identical. You should also try doing the same with other waveshapes, to simulate real world conditions.

To try different A2D resolutions, scale the original numbers differently, this time from 0-1023 for 10 bit resolution - although this will give you even more memory problems on an 8 bit PIC. But I see no reason why 8 bit resolution won't be OK.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

km

Is it you mean to prove the programming code can be used to perform the calculations?
:arrow: For Example:
The C code for RMS calculations:

total = 0;
for (i=0; i<SMP_CNT; i++) //SMP_CNT=Sample Count
{
total += (sample[i]*sample[i]);
}
RMS = sqrt(total / SMP_CNT);

PS: Anyone has the C code for the frequency/peak-to-peak/average/duty cycle calculations feel free to send me. Thank you :wink:

Nigel Goodwin

That's the sort of idea, now generate some sample waveforms and make sure it works - then try different numbers of samples, to try and find the optimum point between small numbers of samples and high numbers of samples. Presumably there will be a point where you start to lose too much accuracy, you need to find it and keep just above it.

P-P is easy, simply search for the lowest and highest values, they are the two peaks.

__________________
PIC programmer software, and PIC Tutorials at:
http://www.winpicprog.co.uk

km

What you mean by generate data for different waveshapes? Is it means that find the most suitable number of samples for different types of waveshape? The waveforms that I need to generate is it a function in the programing part, for example: y=sin(x) and from here I test different number of samples on it?

All the test is it mainly on the programming part only :?:

What others waveshape I need to generate? (analogue waveforms)

Sorry, this is the first time I do these. So, I need to be 100% sure.