TMS320C5535: Is it possible to connect MIC to SAR ADC of C5535?

Hi Sung-IL,

The SAR ADC on C5535 supports upto 64 kilo samples per second (ksps), which is more than sufficient for audio applications.
However, the number of bits with this SAR is 10-bits (effectively 9-bits for linearity), so the resolution is not nearly as good as 16-bit that is available with most codecs like AIC3204.
Additionally, the reference voltage of the SAR ADC is 1.3V max (ignoring the 3.6V tolerant input which goes through a 1/8 voltage divider), so all input voltages must be within 0V to 1.3V. Most electret microphones require a bias supply of 2V or greater, but some microphones can work with 1.3V.

The real issue is the output peak to peak voltage swing from an electret microphone. Normal conversations occur at about 65dB SPL, and the signal from the microphone can be expected to be just 1-2 mVpp. The loudest yelling would register about 100dB, still producing less than 50mVpp from the mic. 9 bits of SAR ADC resolution and a 1.3V Vref gives about 2.5mV per step (ie. between code 0x00 and 0x01) (Note: internal Vrefs of about 1.0V and about 0.8V are available). So at full yelling volume, the un-amplified microphone input would utilize less than 20 ADC codes (about 4%). This may be enough for detection of loud amplitude signals, but it will not sound good at all - especially if amplified after conversion.

Clearly this signal needs to be pre-amplified, and you can achieve good amplification two stages of op-amps (OPA134 for example), but considering the surrounding discrete components required, it might just be easier for you to use the pre-gain amplifier on a CODEC like AIC3204. These codecs also use sigma-delta sampling. You can review this presentation for the pros and cons of sigma-delta vs successive approximation (SAR): http://www.ti.com/europe/downloads/Choose%20the%20right%20data%20converter%20for%20your%20application.pdf

Lastly, I'll leave you with an alternate proposal. Digital microphones produce a digital stream of 1-bit samples at an over-sampled frequency using pulse density modulation. It is possible to directly connect a digital microphone to one of the I2S ports of the C55xx processor. Usually, the C55xx I2S will be the clock master and operate its bit clock and word clock at an oversampling rate (ie. 64x) of the desired sampling frequency. Refer to this app note for details: MEMS Microphone Direct PDM Input via I2S to a C5515 EVM With Software Decimation - http://www.ti.com/lit/pdf/sprac51

See also this presentation I found on PDM audio format:
http://users.ece.utexas.edu/~bevans/courses/realtime/lectures/10_Data_Conversion/AP_Understanding_PDM_Digital_Audio.pdf

Hope this helps,
Mark