Hi,
I have a need to be able to logically reconstruct an analog signal applied to a channel of an ADS1278 using the ADS1278 digital output. In addition, I need to know how much of a delay there will be from the analog input to the digital output.
For example, say that I have a 20 kHz sine wave connected to an input channel of the ADS1278 (of course, properly conditioned) and am operating the ADS1278 at the maximum sample frequency of 144 kSPS. What I need to know is how much of a lag there will be between the input analog signal and the reconstructed digital version of it.
I am not very experienced and Delta-Sigma converters. My initial thoughts where that the digital data at a given sample reflected the average analog input over the previous sample period (1/144 kSPS; about 7uS) but after reading the datasheet section convering "SETTLING TIME", it appears that there is a significant delay and that the digital output incorporates analog input data over many sample periods.
I am probably misunderstanding this, but need to have it clarified before I can complete the design.
Thank you to anyone who can help me,
David Naviaux
You want the group delay spec, which is 39/data_rate. So 39/144,000 = 270uS. This is the delay from the input to output.The settling time is longer then the delay time, as the Delta-Sigma is running an internal FIR filter which has a specific settling time of 79/144,000 = 548uS, this is the amount of time you have to wait for the output signal to meet the A/D accuracy specs
If you need fast response you need a SAR converter.
I also like to add if you do use a SAR converter, you most likely will have digital filters running in software, so the end result of your output may have the same delay as the delta sigma.
The Delta Sigma is more a system chip as it does a lot of the filtering work for you, you don't have the option to turn it off though, whereas using a SAR you can reduce the filtering and therefore reduce the delay, the output will be noiser though. You never get something for free!
Thank you steveg,
I was planning on using this ADC in a 3-phase inverter application to measure the 3-phase voltages and the associated currents supplied to a 480V 3-phase grid. What is important is that at every ADC sample, I can determine what the intantaneous 3-phase voltages and currents are so that I can accurately calculate the instantaneous input or output power of the inverter with a deterministic uncertainty. A 270uS delay is not critical since it will be the average of the instantaneous power samples over one 60Hz cycle that is important. However, what I would be concerned about is if the ADC sampled data does not accurate represent the time varying analog inputs at the ADC (a fixed delay of 270uS would be fine).
If the FIR filter has a setting time of 548uS, does that mean that an ADC sample represents the average of the analog input over the last 548uS or is the analog data to the FIR filter weighted in some way (e.g. the most recent data is more significant)?
Thanks again steveg,
Any input would be greatly appreciated.
-Dave
Steve,
I have reviewed my FIR filters and although all is not 100% clear, I think that I have a better understanding of how the ADC digital outputs model the analog inputs. Is it fair to say that the FIR filters essentially filter the analog inputs of the ADS1278 with a fixed delay of the "group delay" with a stopband frequency of about 0.4 of the sample frequency?
David Naviaux Thank you steveg, If the FIR filter has a setting time of 548uS, does that mean that an ADC sample represents the average of the analog input over the last 548uS or is the analog data to the FIR filter weighted in some way (e.g. the most recent data is more significant).
If the FIR filter has a setting time of 548uS, does that mean that an ADC sample represents the average of the analog input over the last 548uS or is the analog data to the FIR filter weighted in some way (e.g. the most recent data is more significant).
It is not a simple average, the weighting is more complicated, if you look at Figure 58 and 66 it shows the frequency response and step response. A simple average would have much poorer stop band rejection
Yes that is an idealized description of it, If you inject a sinewave into the front end the digital version of the sinewave coming out of the A/D will have a fixed amount of degree shift in it defined by the group delay. It will be constant and known so normally I just adjust for the error (if I am measuring phase for instance). The passband also has ripple associated it, there is a stopband slope, the stopband has ripple too.
Thank you Steve; your input has been very helpful. It looks like the ADS1278 will work fine for my application.