Hey guys, I was wondering if someone can perhaps steer me in some sort of a direction - I don't expect any spoon feeding, just some help to what route I should research.

I have two voltage signals that hangs close to zero volts. Then, at random intervals, two small very quick (+/- 20 micro sec), but different value pulses comes through with their max amplitude reading around 0.5mV. I need to somehow find the peak value of these pulses and "send" it to my microcontroller (arduino) for further processing. I have been looking at the mcp3202 http://www.mantech.co.za/Datasheets/Products/MCP3202.pdf , but I am not sure if adc is the correct way to go about this task. Are there chips/IC's that can handle this for me and maybe provide the peak value via a SPI or something?

Thanks in advance for any help!

 

20 uS pulse means sample rate should be 10 uS or less. 100Khz or faster.

12 bit A/D with ~ 1V Vref may handle the resolution needed. A PSOC 4 could
easily handle that with itrs onboard SAR. If you needed more res a PSOC 5LP
with DelSig configed at 15 bits could meet the sample rate. Thats ~ 30 uV
resolution.

Either A/D will common mode to the rail, and do it differentially.

What accuracy and resolution do you need of the peak measurement ?

PSOC 5LP has a +/- .1% onboard Vref if I am not mistaken. And all the CPU, COM,
etc....


Regards, Dana.

 

If there is a known minimum time between these two pulses, then an analogue peak detection circuit might be better option, which would avoid the need for a fast sampling rate. Since the TS is only interested in measuring the peak voltage level and not the duration/rise/fall time of the pulses.

 

If sampling with an ADC is attempted, the required sampling rate will depend a great deal on the shape of pulse. If it is cleanly rectangular, then sampling fast enough to get one sample at the peak is adequate. If the pulse is "pointy" - triangular, "sinusoidal", et cetera - then sampling would have to be done at a much higher rate to have any reasonable assurance of getting the peak. The mechanism by which the input signal is sampled and the effect of changing signal amplitude during sampling must be considered. There are lots of possibilities for big errors.

Many ADCs on-chip with microcontrollers require special precautions to assure that digital noise does not corrupt the analog signal. This sometimes means shutting down significant parts of the digital circuitry while conversion is being done in order not to lose two or three bits at the bottom (e.g. 12 bit converter being reduced to 9 or 10 bits). Converters (or any circuitry) with microvolt resolution must be designed with very careful attention to the errors that can be produced by thermoelectric potentials. Physical layouts must be done meticulously with this in mind. Many high-resolution ADCs are completely unsuitable for this sort of signal unless a fast analog peak detector is used ahead of the converter. Fast precision analog peak detectors aren't easy.

 

Applicable ref material -

http://www.analog.com/media/en/tech...294542582256114777959693992461771205AN280.pdf

http://www.cypress.com/documentatio...nd-psoc-5lp-mixed-signal-circuit-board-layout

http://www.cypress.com/documentatio...nd-psoc-5lp-mixed-signal-circuit-board-layout


Regards, Dana.

 

Wow, thanks for all the great replies.

Thanks also for the links. I will go through that.

I think it would be best if i take a screenshot of my scope and post that. I am a hobbiest so my info might be dangerous.

 

Guys

I finally found a solution, but not really the answer to my original question. I decided to go for timing difference rather than peak voltage levels. For those who might be interested, after looking at Schmitt triggers, I used a 555 timer to perfectly detect the start of each pulse. Works like a charm.

Thanks any case, hope this might help someone.