audio ADCs with minimum sample rate specs ?

[Winfield Hill]

Adam S wrote...

Why do nearly all audio ADCs and CODECs have a minimum sample rate
specification ? I'd would like use a low cost 24bit audio ADCs and
interface to a microcontroller (SPI) at low sample rates < 4kHz.
Texas Instruments PCM**** range of audio ADCs typically have minimum
Fs of 16kHz. What happens at lower frequencies ? , does the logic stop
working, or is it something to do with capacitor charge loss in the
delta sigma modulator ?

I'm speculating that there are two possible reasons. First, it's
common to implement complex CMOS circuits with dynamic rather than
static circuitry, because it takes fewer MOS transistors. In such
a case, there will be a minimum operating frequency, although you
may go below that if you're below the maximum temperature spec.
The same should go for the charge on node capacitors leaking away.

Second, some delta sigma modulators have internal integrators that
have voltages that increase with time, and a maximum clock period
has to be specified to keep them from overflowing.

There are lots of cheap high-resolution low-frequency delta-sigma
ICs available, aimed at scales, geo-electronics, process industry
and other markets. Why make an awkward use of an audio IC.

 

[Al Clark]

Why do nearly all audio ADCs and CODECs have a minimum sample rate
specification ? I'd would like use a low cost 24bit audio ADCs and
interface to a microcontroller (SPI) at low sample rates < 4kHz.
Texas Instruments PCM**** range of audio ADCs typically have minimum Fs
of 16kHz. What happens at lower frequencies ? , does the logic stop
working, or is it something to do with capacitor charge loss in the
delta sigma modulator ?

Most sigma delta ADC will actually work at lower sample rates. I am fairly
certain a Wolfson WM8731 (basically the same as TI 320AIC23 with two pins
swapped) will work at 4k by just halving the MCLK and using its 8k mode.
You can ignore the DACs if you don't need them.

If you are interfacing to a DSP, chances are you have MIPs to burn. You
could sample at a higher rate and decimate.

 

[Adam S]

Why do nearly all audio ADCs and CODECs have a minimum sample rate
specification ? I'd would like use a low cost 24bit audio ADCs and
interface to a microcontroller (SPI) at low sample rates < 4kHz.
Texas Instruments PCM**** range of audio ADCs typically have minimum Fs
of 16kHz. What happens at lower frequencies ? , does the logic stop
working, or is it something to do with capacitor charge loss in the
delta sigma modulator ?

 

[[email protected]]

I was attempting to use an 8bit AVR micro, and talk SPI to an audio ADC
at Fs=4 kHz in order to sample two incoming 1000Hz sine waves at 24bit
resolution. The software demodulates the two signals at a bandwidth of
few Hz, before computing the phase and magnitude ratios. The DSP code
I've written to do the demodulation should in theory allow sample rates
up to about 16kHz with a 16MHz MPU clock, while consuming 100% processor
time. So , at much lower sample rates (say 4 to 8kHz) that should leave
me with plenty of spare clock cycles to play with.

The problem at those sample rates it seems too low for many audio ADCs
and too high for many the DC precision ADCs. There doesn't appear to be
much overlap.

If they offer a successive approximation ADC with the number of bits
you need, that might be an answer.

But perhaps you can use a sigma-delta audio part at it's intended rate,
but figure out a way to decimate in hardware? For example, if you can
create the clock with a timer and the L/R signal with a flip flop, and
then use a timer/interrupt generator to interrupt the micro every n
clocks and have it record a sample - that leaves you n-1 sample periods
of time to do other stuff. That does require though that the micro be
able to grab the serial data at the intended rate.

 

[Adam S]

@news.optusnet.com.au:




Most sigma delta ADC will actually work at lower sample rates. I am fairly
certain a Wolfson WM8731 (basically the same as TI 320AIC23 with two pins
swapped) will work at 4k by just halving the MCLK and using its 8k mode.
You can ignore the DACs if you don't need them.

If you are interfacing to a DSP, chances are you have MIPs to burn. You
could sample at a higher rate and decimate.

I was attempting to use an 8bit AVR micro, and talk SPI to an audio ADC
at Fs=4 kHz in order to sample two incoming 1000Hz sine waves at 24bit
resolution. The software demodulates the two signals at a bandwidth of
few Hz, before computing the phase and magnitude ratios. The DSP code
I've written to do the demodulation should in theory allow sample rates
up to about 16kHz with a 16MHz MPU clock, while consuming 100% processor
time. So , at much lower sample rates (say 4 to 8kHz) that should leave
me with plenty of spare clock cycles to play with.

The problem at those sample rates it seems too low for many audio ADCs
and too high for many the DC precision ADCs. There doesn't appear to be
much overlap.

Adam

 

[Guest]

: Why do nearly all audio ADCs and CODECs have a minimum sample rate
: specification ? I'd would like use a low cost 24bit audio ADCs and
: interface to a microcontroller (SPI) at low sample rates < 4kHz.
: Texas Instruments PCM**** range of audio ADCs typically have minimum Fs
: of 16kHz. What happens at lower frequencies ? , does the logic stop
: working, or is it something to do with capacitor charge loss in the
: delta sigma modulator ?

Could be, but more likely, the quantization noise will come up so
much in the audible band that the output will sound like crap, if you try
to listen to it. You COULD postfilter the output to remove the
quantization noise that is in the audio band, but the part will not do
that for you. Normally, the quantization noise will be guaranteed to be
out of band (the part is speced within the audio band.)

I typed that out really quick and I hope that it is clear. Please
let me know if you need further clarification....

Joe