10 bit ADC possible for LPC1768?? I need 500 ksps!

[lpcware]

Content originally posted in LPCWare by Prithvi on Thu Sep 19 14:41:05 MST 2013
Hi everyone in this amazing community,
I see in the LPC1768 datasheet that it is possible to do 10 bit ADC at 500 Ksps at 3.3V, but I don't find that option in any register in the user manual. Any idea about this? It'll be great if this is possible since otherwise am external ADC would be required since 12 bit samples at 200 ksps is not fast enough.
Thanks a lot,
Prithvi

[lpcware]

Content originally posted in LPCWare by Prithvi on Fri Sep 20 02:43:20 MST 2013
Thanks a lot, that explains things. I don't know how I skipped the 33MHz in the table.
And it'll be great if this is explicitly mentioned somewhere in the LPC1768 documents.
Cheers,
Prithvi

[lpcware]

Content originally posted in LPCWare by tha on Thu Sep 19 19:29:30 MST 2013
Yes, it can do the 500ksps.  If you look at the table between the 12bit and the 10bit, the fclk is 13MHz for the 12bit, while the fclk is 33MHz for the 10bit.  So although a 12bit conversion is still being done, by increasing the fclk, the sampling rate increase by sacrificing accuracy (discarding of the last two bits).

[lpcware]

Content originally posted in LPCWare by Prithvi on Thu Sep 19 16:27:47 MST 2013
In table 20 it says that 500 ksps can be achieved if a '10-bit' ADC is used. So the SAR ADC must stop at 10 bit to be faster, not just let it do the 12 bit conversion and ignore the last two bits. I've attached the image of the table as a reference. Please let me know if 10 bit ADC at 500 ksps is possible in LPC1768.
Thanks a lot for the quick response!

[lpcware]

Content originally posted in LPCWare by tha on Thu Sep 19 15:41:50 MST 2013
The LPC1768 has a 12-bit ADC.  In Table 19 of the datasheet, the values is for the 12-bit ADC if used as an 10-bit ADC.  This just mean that the last two bits should be ignored when using the data is to be interpreted as a 10-bit ADC.

You do not need to do anything special, just set up the 12-bit ADC normally, but when reading the results, discard the last 2 bits.