Facing a problem to measure peak voltage (ADC) at 50kHz

[khalis]

Currently, I have a project which is using PIC16F883 to measure peak voltage through ADC port and ADC input's frequency is at 50kHz. The input to the ADC may varies, sometimes like pulse wave, sawtooth wave or sine wave but it is at fixed 50kHz. My problem is, I always failed to measure a peak voltage ex in-rush current which is occurred only in a few ms.

I have used two method to measure it:

i - when the ADC value is above than certain value, increase a counter.
Running in a while loop.

ii - i take 20 samples where 1 sample is taken at 128us. Then i calculate the
average.

Do you guys have any suggestion to solve this matter. Thanks in advance.

* ADC clock FOSC /32
* Used internal MCU oscillator - 8MHz

[FvM]

Do you mean, the ADC input voltage is a 50 kHz sine wave, or did you rectify the 50 kHz signal before? You can't reliably detect a peak voltage if it lasts shorter than the time between two ADC conversions.

[asmboy]

This can be done with a rectifier + peak hold capacitor - and a pic output line
to reset the "peak hold" capacitor. ( metal film capacitor please )

Let me second the previous post - its NOT gonna happen with simple ADC readings. You need some SOLDER programming of external analog components to do this right.

[khalis]

Thanks for the response guys...

FVM >> ADC input is not 50kHz sine wave but 50kHz pulse wave. But it also can be a different waveform because it is depending to load (Inductive or Capacitive) which it is connected to. Actually, PIC is used to measure discharge current from a battery (Inverter). As a result, the discharge current some kind like On and Off at 50kHz.

asmboy >> I have read that kind of method in the internet before but still under consideration because if it is possible, i want to sense the peak current by using software method first.

[FvM]

Sensing the peak current by software methods isn't feasible, when the waveform has a frequency of 50 kHz and each ADC conversion takes about 50 µs. At best, you can try to sample the peak current if you know the time of it's ocurrence in advance which may be the case when the processor is controlling the inverter. The ADC acquisition time still restricts the accuracy of catching the true peak current.

[khalis]

Noted that FvM...i knew it before but i still need opinion from others like you guys. Anyway, thanks again for your explanation.

[asmboy]

you need to understand that software does NOT solve all problems

a flexible approach that combines hardware with software skill is
the way to success. as other posters have demonstrated so well -
clear mathematical analysis is the ONLY basis for a good solution.

i actually can imagine a statistical way to get the info you want - with only software - but it is NOT pretty and it would take a LONG time to get that info - when i suspect you need it "FAST" --
you see, this is one of the cases where no amount of software cleverness is going to get the answer you want in the TIME you want it

[asmallri]

[SherpaDoug]

[ckielstra]

Ok, I'm not a hardware expert but here are some thoughts:

1) What is the definition for peak current? Without a time parameter it seems like a meaningless value to me. I mean, if you are measuring in small enough intervals you can measure an infinite current on the pulse edges but very soon the current will drop down to more normal values. All kind of higher harmonics are present but total energy is low. Wouldn't it make sense to have some kind of low pass filter at, say, 200kHz?

2) What kind of input signal is this? You say it is always 50kHz, that suggests it is a bit special.
If you have a limited number of repeated signal patters, maybe you can do something clever by calculating when the largest current is supposed to occur for this signal pattern? And then take a sample at the calculated time after the zero crossing?

[asmboy]