Digital PI schematic

[Nicko]

Re: Digital PI schematic

Robert,

The ADuC841 is probably good enough for a metal detector - the move to a single cycle instruction clock is a great improvement - others have being doing this for a while, and its good that AD will be doing this too. The device is still short of SRAM, but looks promising...

The dsPIC is a strange device, predominately for motor control, and whilst it has buckets more SRAM and I/O, it has no on-chip DAC, which is a bit of a pain...

I presume you are blocking interrupts during the main timing cycle - when I used this device for a robotics project I was generating stepper motor timing whilst interrupts were going on - not a good idea (loads of jitter).

Are you using the AD development board for the 812 (ISTR that its about $200), or your own board...?

Nicko

[Carl]

Re: Digital PI schematic

The input bias current won't cause any loss of resolution in the ADC. If it's constant, it will offset the input; if it has a dependency with signal level, it creates a gain error, and potentially harmonics.


Where input resistance can affect resolution, is in thermal noise contribution.

- Carl

[Nicko]

Re: Digital PI schematic

Hi Carl,

With +/- 1uA input leakage, a 610ohm input impedance will generate 610uV of error. WIth a 2.5V reference (the internal reference), this equates to 1 LSB (2.5V/4096 - they are 12bit ADCs). This issue is specifically raised in the ADuC812 FAQ.

So, for relative measurements, it will make no difference if, as you say, the leakage is constant - the error will then be constant and will cancel out. For absolute readings, high input impedances will create fixed errors. Again, you are correct regarding thermal noise...

The principal is still important, and very easy to miss. The whole issue of using ADCs in a low noise environment is critical. As this application requires close attention to noise and signal loss, all these issues need to be taken into account.

Nicko

[Robert Hoolko]

Re: Digital PI schematic

Nicko

I am using the Analog Devices evaluation board. However I am not using the external memory that is available on that board because I am trying to evaluate what I could do with the schematic I posted here which only uses the ADuC812 chip.

In the tests I am doing on Eric's forum I am connecting the evaluation board to a free running analog PI detector, so jitter is unavoidable. The digital board is only synchronized to the analog board by an interrupt, and the currently executing instruction has to complete before the interrupt can be serviced. That can give as much as 4 usec of jitter. Switching to an 841 would reduce that jitter to less than 1 usec just because it is a faster chip.

I do not have enough information about the 841 to know if the new SRAM they added can be used for burst mode A-D conversions. If it can, that will take care of the jitter problem when the uP is running everything. But if burst mode cannot be used without external chips then I will always have jitter problems because I cannot afford to wait in a timing loop for 100 usec.

Robert

[Carl]

Re: Digital PI schematic

True, you will lose precision, but not resolution. But with up to 1.5 LSBs of INL, the 812's ADC is probably not meant for precision measurements anyway. And in a metal detector design, precision is unimportant. Resolution is what matters.

- Carl

[Vasil]

Re: Digital PI schematic PIC16f877

PLS let mi shematic.