At 10:39 1998-05-02 -0400, you wrote:
>Morgan,
>
>I think (puzzled look) I understand your method, but you have me thinking
in a
>diferent direction.  What if you use a PWM that you write yourself (12 or
14 or
>16 bit count off of the TMR0, pre-stuff for high speed?) to generate a target
>output, and use the A/D input to qualify it?  This assumes you have filtered
>your home made PWM to the quality you wish.  

I need to have stabel output level between output changes.  I will have a
loop which make about 50 A/D conversions per second on a few channels, and
I want to update the analog out 25 times per second if possible, and keep
it stable in between.  It must make a full change every time or else I will
have oscillation in the whole system (a kind of balance system)

But I believe 8-bit (linear) resolution is enough.

>I personally have had trouble
>getting '16' bit performance out of the part.  I am curious to know what
usable
>performance you get.

Right, the noise level is higher than I expected, and right now i begin
implementing a filter for the A/D-read.  Strangely it make very little
effect if cpu is on or in sleep during conversion.  Right now I am not sure
where the noise come from: ramp generator, comparator (probably) or my
preamp (probably too (it«s cheap). Supply is very stable.

>
>Chris Eddy
>Pioneer Microsystems, Inc.
>
>Morgan Olsson wrote:
>
>> PIC14000 has A/D circuitry consisting of ramp generator and a 16bit fast
>> counter which capture the count when a comparator say the measured input is
>> the same as the ramp.
>>
>> As it is possible to backward calculate the exact count for a desired
>> voltage we can make a high resolution D/A if we trigger a sample/hold
>> cirquit to lock the value of the ramp at the desired count.
>>
>> It is easy just to continuously read teh counter and send a trigger pulse
>> at a desired value.  BUT the datasheet say we should not read the counter
>> while it is running.
>> Why?
>> Anybody tried?
>> Will the read make it *count* false?
>> Or do we risk *readig* erroneous value?
>> Probably the latter, as the counter is normally driven very fast, by the
>> oscillator directly (4MHz typ), and the counter is probably not synchronous
>> binary counter type (more silicon need), so the count might not have
>> propagated through the whole counter when it is read by the cpu?  That is
>> of course dependant on mainly clock frequency, but also the individual
>> chip, temperature, supply voltage...
>>
>> But I believe there should be a osc frequency below which it is guaranteed
>> to work OK.
>>
>> Ideas?
>>
>> We can also use the 8-bit timer to generate interrupts, but i need it for
>> other stuff, and it is more complicated to get it all work.
>>
>> /Morgan
>>
>> /  Morgan Olsson, MORGANS REGLERTEKNIK, SE-277 35 KIVIK, Sweden \
>> \  mrt@iname.com, ph: +46 (0)414 70741; fax +46 (0)414 70331    /
>
>
>
>Attachment Converted: "C:\EODORALT\Attach\vcard2.vcf"
>
/  Morgan Olsson, MORGANS REGLERTEKNIK, SE-277 35 KIVIK, Sweden \
\  mrt@iname.com, ph: +46 (0)414 70741; fax +46 (0)414 70331    /