Maximum Sampling Rate of DC voltage sensors

Thread Starter

rcvkumar2000

Joined May 27, 2021
6
Hey there,

I wonder what is the maximum sampling rate used to measure voltage.
That is, how accurately can we measure a voltage. For example, for implementing MPPT algorithms for PV arrays.

Will be really helpful, if you can inform me of this....
Thank you
 

Yaakov

Joined Jan 27, 2019
3,155
There is a specification of ADCs called conversion time which is the total time for converting the analog signal to a digital one. It depends on the type of ADC and the physical circuitry (propagation time for the signal through it).

In a SAR ADC, the specification is in SAR clocks, that is, how many clock cycles it takes. Multiplying the SAR clock interval by this will get conversion time. In a pipelined ADC you will find a samples per second specification.

Conversion time isn't the total overhead but it's the major portion of the time between starting to sample and delivering the results

In the case of an MPPT device, where do you expect the sample rate to be a consideration? Modern ADCs are much faster than I can imagine any useful algorithm to care about.
 

Ian0

Joined Aug 7, 2020
2,220
It depends on how fast the clouds blow across!
But seriously, @Yaakov is perfectly correct in saying that the A/D is unlikely to be the bottleneck. Even delta-sigma devices that do 15 samples a second would probably be fast enough for solar power monitoring.
If you are intending to implement a P&O algorithm, your limiting factor would be how long it takes the perturbation to take effect bearing in mind the capacitances in the circuit. You might also need some IIR filtering after the sampling to remove noise.
 

Thread Starter

rcvkumar2000

Joined May 27, 2021
6
There is a specification of ADCs called conversion time which is the total time for converting the analog signal to a digital one. It depends on the type of ADC and the physical circuitry (propagation time for the signal through it).

In a SAR ADC, the specification is in SAR clocks, that is, how many clock cycles it takes. Multiplying the SAR clock interval by this will get conversion time. In a pipelined ADC you will find a samples per second specification.

Conversion time isn't the total overhead but it's the major portion of the time between starting to sample and delivering the results

In the case of an MPPT device, where do you expect the sample rate to be a consideration? Modern ADCs are much faster than I can imagine any useful algorithm to care about.
Thank you for the reply, Sir.

I have been designing a PV system in MATLAB. I have used the Perturb and Observe Algorithm for MPPT. I used a sampling time of 20 microseconds for the algorithm and used a small step size. I am not sure about the magnitude of the sampling time which is generally used. I am afraid that what I use might be impractical. That is why I wanted to make sure.

I checked some datasheets of several ADC and the maximum sampling rate observed was 800 SPS. So, I guess I cannot achieve what I have implemented in the simulation. Any further advice would be really helpful.
 

Thread Starter

rcvkumar2000

Joined May 27, 2021
6
It depends on how fast the clouds blow across!
But seriously, @Yaakov is perfectly correct in saying that the A/D is unlikely to be the bottleneck. Even delta-sigma devices that do 15 samples a second would probably be fast enough for solar power monitoring.
If you are intending to implement a P&O algorithm, your limiting factor would be how long it takes the perturbation to take effect bearing in mind the capacitances in the circuit. You might also need some IIR filtering after the sampling to remove noise.
Great! Those are some valuable points I would note.

I am trying a new strategy that would require 20 micro-second sampling time of the PV voltage. I guess that is impractical. What are your thought, sir?

Thank you!
 

Ian0

Joined Aug 7, 2020
2,220
I am trying a new strategy that would require 20 micro-second sampling time of the PV voltage. I guess that is impractical. What are your thought, sir?
Most microcontrollers these days have A/d converters that can approach 1M samples/second, so I doubt it would be a problem. For instance I’m currently using a LPC5502 which can has a 16Bit converter operating at that speed, and it only costs £1.85. That’s cheaper than some standalone a/d converters.
 

Thread Starter

rcvkumar2000

Joined May 27, 2021
6
Most microcontrollers these days have A/d converters that can approach 1M samples/second, so I doubt it would be a problem. For instance I’m currently using a LPC5502 which can has a 16Bit converter operating at that speed, and it only costs £1.85. That’s cheaper than some standalone a/d converters.
Amazing
Thank you for you valueable time and knowledgeable response....
Cheers!
 

Tron Jockey

Joined May 3, 2020
17
Hey there,

I wonder what is the maximum sampling rate used to measure voltage.
That is, how accurately can we measure a voltage. For example, for implementing MPPT algorithms for PV arrays.

Will be really helpful, if you can inform me of this....
Thank you
=======================================================================
I think we can assume that A-to-D Conversion times are inconsequential for this application. You're essentially measuring a varying DC voltage that transitions relatively slowly. No need to consider Nyquist limits. Just about any A-to-D converter will convert far faster than your application demands. I will assume then that what you're mostly concerned with is "sample rate". Dismissing conversion speed what you then have are essentially two variables, samples per unit of time (likely minutes for your application), and sample depth (bits per sample). There are other variables but at this point in the design sample rate and sample depth are all you really need to be concerned with, interfacing circuitry and signal conditioning can then be used to make the selected A-to-D converter fit virtually any application. Satisfying these two variables depends upon the requirements of your application and what it needs to effectively capture as data samples and what it needs to do with those data samples (averaging, level detecting, power spectral density, duty cycle, controlling feedback, etc...). Once you know what your particular MPPT controller requires for data input it should narrow the field of prospective A-to-D converters. I have no experience with PVs, MPPT controllers or solar applications of any sort so take this with a grain of salt but, just winging it, I'd guess that one sample every 10-30 seconds would suffice. PV output voltages are typically less than 40Vdc so a 12-bit converter would provide roughly 0.4V per bit resolution worst case. Signal conditioning that 40V down to something like 1V would increase that 12-bit converter resolution to 0.003V per bit. I can't imaging this application needing more resolution. Without knowing the input requirements for your MPPT controller its difficult to even guess.
Try researching the following: Optimizing sampling rate of P&O MPPT technique | IEEE Conference Publication | IEEE Xplore
 

Thread Starter

rcvkumar2000

Joined May 27, 2021
6
=======================================================================
I think we can assume that A-to-D Conversion times are inconsequential for this application. You're essentially measuring a varying DC voltage that transitions relatively slowly. No need to consider Nyquist limits. Just about any A-to-D converter will convert far faster than your application demands. I will assume then that what you're mostly concerned with is "sample rate". Dismissing conversion speed what you then have are essentially two variables, samples per unit of time (likely minutes for your application), and sample depth (bits per sample). There are other variables but at this point in the design sample rate and sample depth are all you really need to be concerned with, interfacing circuitry and signal conditioning can then be used to make the selected A-to-D converter fit virtually any application. Satisfying these two variables depends upon the requirements of your application and what it needs to effectively capture as data samples and what it needs to do with those data samples (averaging, level detecting, power spectral density, duty cycle, controlling feedback, etc...). Once you know what your particular MPPT controller requires for data input it should narrow the field of prospective A-to-D converters. I have no experience with PVs, MPPT controllers or solar applications of any sort so take this with a grain of salt but, just winging it, I'd guess that one sample every 10-30 seconds would suffice. PV output voltages are typically less than 40Vdc so a 12-bit converter would provide roughly 0.4V per bit resolution worst case. Signal conditioning that 40V down to something like 1V would increase that 12-bit converter resolution to 0.003V per bit. I can't imaging this application needing more resolution. Without knowing the input requirements for your MPPT controller its difficult to even guess.
Try researching the following: Optimizing sampling rate of P&O MPPT technique | IEEE Conference Publication | IEEE Xplore
To be honest, I'm trying to check the validity of modifying the P&O algorithm so that I can bypass the DC-DC power stage. I'm considering the load to be a pump which doesn't have any specific speed whatsoever. So I'll run it at whatever speed possible for a given input power. Thereby, varying the speed can alter the operating point of the array. So by noticing the changes of the array voltage by perturbing the speed, the MPP can be tracked.

Even though I wasn't able to achieve instantaneous speed tracking, the average value to speed was correctly determined to make the array work at maximum power. I used a sampling time of 20 microseconds and very small perturbations in speed ( a larger value will lead to oscillations). All of these are based on simulation. I was wondering if I could achieve such a fast sampling (sample the voltage at every 20 microseconds) of voltage to a certain accuracy.
 
To be honest, I'm trying to check the validity of modifying the P&O algorithm so that I can bypass the DC-DC power stage. I'm considering the load to be a pump which doesn't have any specific speed whatsoever. So I'll run it at whatever speed possible for a given input power. Thereby, varying the speed can alter the operating point of the array. So by noticing the changes of the array voltage by perturbing the speed, the MPP can be tracked.

Even though I wasn't able to achieve instantaneous speed tracking, the average value to speed was correctly determined to make the array work at maximum power. I used a sampling time of 20 microseconds and very small perturbations in speed ( a larger value will lead to oscillations). All of these are based on simulation. I was wondering if I could achieve such a fast sampling (sample the voltage at every 20 microseconds) of voltage to a certain accuracy.
Hmm, I thought you were chasing an entirely different problem. However, finding a fast sampling 8, 10, 12, 14 or even 16-Bit ADC should be quite easy. There are many high speed A-to-D converters that will convert far faster than 20 usecs (50kHz). Just remember to design a proper anti-aliasing filter in front of it.
 

crutschow

Joined Mar 14, 2008
27,213
I used a sampling time of 20 microseconds and very small perturbations in speed ( a larger value will lead to oscillations)
If you loop stability is that sensitive to sampling time, then it would seem your loop is not sufficiently stable.
I would think the loop should be stable down to a sample time that's only a few times faster than the maximum frequency change of the PV voltage.
 
Top