| ??? 09/05/09 03:50 Read: times |
#168737 - That may not work Responding to: ???'s previous message |
Daniel Contarino said:
Hi Ghulam
At 19.200Bd, each byte takes 500us aprox. It means 2KHz sampling rate max. with one channel, 8 bits. I think that you should use at least 50Hz sampling rate for ECG. Otherwise will lost information in the QRS part. In that case, and using 14 bits (two bytes per sample), you will be able to sample 20 channels (OK, may be less, I am doing approximate calculations). Just how on earth will he get 14 bit precision from an 8-bit converter that has a conversion time of 100 microseconds? Summarizing, I think that your 8052 would be able to manage this project (may be an one-clocker, like DS89C450). And unless you think in averaging (like related/evoked potentials), you should pay special attention to noise, as previously advised. And IMHO also think that 14 bits is too much. For complete use of a 14 bit signal, you have to very carefully design board, amplifiers, etc.
Daniel The converter he says he's using, ADC0804, is slow, it's an 8-bit converter with a random error of 1 lsb. This will require lots of code for things like bitwise control and timing of nWR, nRD, etc. and I can't imagine a way to extract 14 bits from it. The Maxim/Dallas one-clocker is WAY too fast to allow the "conventional" interface to be used, as its access time is 135 ns, minimum, and based on NSC800 or i8080 timing. I'm inclined to believe he might be better off with one of the many linear codec's offered on the market for audio- and sub-audio-rate measurements of high precision. Their manufacturers offer lots of support, and they're designed to work in fairly noisy environments such as a PC, which normally has a really dirty switching power supply. I'm no audio expert, but ISTR that CD-quality is rated at 16 bits. RE |
