I'm now using ADXRS453 gyros. There's one problem that confuses me. In the datasheet, it is said that the rate output is SPI digital output with 16-bit data-word, and the meaningful range is ±24,000 LSB. But in the functional block diagram, ,
I can assure you that the performance in Table 1 came from actual characterization data, using actual units and industry-standard approaches. So, that tells us that while this is an interesting question, there has to be more to this than a simple 12-bit ADC, applied over the +/-300 deg/sec measurement range. With that in mind, I can offer some simple thoughts, which I hope will help. When the noise of a sampled signal is greater than 1 LSB, you can increase the resolution through accumulation and preserving the associate bit growth in the resulting digital signal path. That is a simplified explanation for how this works in the ADXRS453. I hope that helps.
go on, the ADC in ADXRS453 is a 12-BIT ADC, which can only provide 0-4095. So my question is if the 16-bit data-word output credible.
What does "how credible is 16-bit" mean? In the ADXRS453, you are not buying an A/D converter, you are buying a gyroscope that supports the performance metrics in Table of its datasheet. If you have a more specific concern, we will be glad to discuss it.
I was attempting to add a link to Table 1 in the ADXRS453 datasheet, but it looks like I accidentally cut this off:
Thanks，what I want to say is that, the range of the gyroscope is ±300dps and if the ADC inside it is a 12-bit one, the meaningful resolution will be 600/4096 = 0.146dps/LSB. According to the datasheet, the nominal sensitivity is 80LSB/°/sec which means 0.0125dps/LSB. How dose the gyroscope change the 12-bit ADC output into a 16-bit SPI output, which seems to "improve" the precision of the gyroscope.
Thank you! I think you've caught my meaning, though I still cannot understand how this product "increase" the resolution. For the reason that the performance in Table 1 came from actual characterization data, which you told me, I decide not to think about that any more. Maybe I should review the courses on electronics, especially the fundamental principles, which will help me understand these things.
Think about this way, if you add 4 12-bit numbers together, you will end up with a 14-bit number, if you capture the over-range bits. Assuming there's enough noise to dither the LSB (in other words, cause random changes) all 14 bits of the result will be valid. If you add 16-12-bit numbers together, you will end up with a 16-but result.
Is this the oversampling technique?
Not sure I would use that title (there is to it than sampling > Nyquist) but I have seen some refer to this behavior in that manner.
Retrieving data ...