MSP430FR6043: Is there a limit to the total shift in Absolute UPS/DNS?

Hi,

For how to set a lower sampling frequency in the other thread. You can set the parameter USS_HSPLL_FREQ_IN_MHZ to 68 and USS_SDHS_OVER_SAMPLE_RATE to 40 in the USS_userConfig.h. The ADC sampling rate will be set as 68M/40 = 1.7M. You can also try with 1M sampling frequency if 1.7M can still observe the spikes. 

Best regards,

Cash Hao

I collected some data demonstrate the issue. See attachment.   We also discovered that the spikes we are seeing are about 3.2u2 deep, corresponding to one period of the received ADC signal.  So it appears that the algorithm is getting out of phase by one period.  Sometimes, instead of spiking from the correct value to the wrong vaolue, the Delta ToF is primarily on the wrong value and spikes ocassionally to the correct value.

Data for TI forum question.pdf

Hi,

I checked your data. I know there are cycle slip issues on dTOF at 6LPM and 4LPM and on absTOF_DNS at 4LPM. 

So, first lowering the ADC sampling frequency could help solve the cycle slip issue on dTOF. So, I would suggest you to do it first. And then we will check the results after changing it. And if it not helps, please let me know and I will let you try with some other parameters then. 

Best regards,

Cash Hao

Thanks Cash.  Please see attached.

It appears that the ToF agorithms don't work at that frequency.  Our transducer is about 310kHz, so at 1.7MHz sampling frequency we only get about 5.48 samples per period.  If I understand the documentation, you need at least 3 samples per lobe (6 samplesvper period) for the algorithm to fit a parabolic curve to each lobe.  That may explain why the 1.7MHz didn't work.

So we also dried 3.4MHz.  It worked better but still spiked at certain flows.  At 8LPM, the Delta ToF was stuck on zero.  We appreciate any further direction you can provide.  In the attached document I pasted the settings we used to create the header file. Please review them to see if we have something wrong.

Data for TI forum question 2nd.pdf

Hi,

Yes, you will need 3 samples per lobe for the algorithm to fit a parabolic curve to each lobe. But it is NOT 6 samples over period, around 4 samples over period is good enough for calculation. 

Brent Weight said:

That may explain why the 1.7MHz didn't work.

This is wrong. Based on your data in attachment. When HSPLL 68MHz, OSR 20, the sampling frequency should be 68M/20 = 3.4M, not 1.7M in your description. So, setting to 1.7M can get a better result than 3.4M. 

And for error code 135, you can change the USS_ALG_MAX_SAMPLE_SHIFT into a bigger value, like 40/60/80. 

Best regards,

Cash Hao

Thanks, we will try a bigger USS_ALG_MAX_SAMPLE_SHIFT tomorrow.

Sorry I mixed up the OSR values in the document.  I presented data for both 1.7M (68/40) and 3.4M (68/20).

Okay. Then keep the 3.4M sampling frequency and try with the bigger USS_ALG_MAX_SAMPLE_SHIFT. 

Best regards,

Cash Hao

Cash we tried USS_ALG_MAX_SAMPLE_SHIFT set to 80.  We no longer recieve the error code.  However, at 3.4MHz we observed very simialr results compared to the document I sent yesterday.  At 1.7MHz, the behavior was better than yesterday's 1.7MHz results, but worse than the 3.4MHz either day.  Both 1.7 and 3.4 MHz perform very poorly compared to 2MHz.  I don't know if that means we have not configured it properly for frequencies other than 2MHz.  Are there other settings we should be playing with at 1.7MHz and 3.4MHz to tune the system to behave at those sampling frequencies?