Sample rates for multichannel DACs?
- Thread starter niklasni1
- Start date
Help Support GroupDIY Audio Forum:
mulletchuck
Well-known member
fix yer links.
jdbakker
Well-known member
In an audio context, the AD5348 isn't the most usual example you could have picked.
A multichannel audio DAC, like the PCM1681 or the CS4384, acts like several stereo DACs in a single package, generally with shared clock lines and separate data lines. The DACs run in lock-step, but for a 192ksps-part each output will update at that rate. (In fact, I2S being stereo by default, each data line will carry 2 * 192ksps in this situation).
The AD5348, which you'd not normally use for audio, has one shared data interface for all eight DACs. The data sheet doesn't explicitly mention a maximum sample rate: you are limited by the settling time of the DACs on one side and the write cycle time of the data interface on the other side, with the former looking to be the limiting factor in most cases.
JD 'why'd you ask?' B.
A multichannel audio DAC, like the PCM1681 or the CS4384, acts like several stereo DACs in a single package, generally with shared clock lines and separate data lines. The DACs run in lock-step, but for a 192ksps-part each output will update at that rate. (In fact, I2S being stereo by default, each data line will carry 2 * 192ksps in this situation).
The AD5348, which you'd not normally use for audio, has one shared data interface for all eight DACs. The data sheet doesn't explicitly mention a maximum sample rate: you are limited by the settling time of the DACs on one side and the write cycle time of the data interface on the other side, with the former looking to be the limiting factor in most cases.
JD 'why'd you ask?' B.
Yes. For a variable clock system it's more appropriate, though, since you can actually vary the clock 
I think you answered my question, though: the write interface is more than capable of keeping up with 8x 125ksps, but the DACs take 8µs to settle -- which is where the 125ksps figure comes from, I'd think --, so it should be fine for 8 channels of <50khz.
I think you answered my question, though: the write interface is more than capable of keeping up with 8x 125ksps, but the DACs take 8µs to settle -- which is where the 125ksps figure comes from, I'd think --, so it should be fine for 8 channels of <50khz.
Most DAC's are capable of variclock. What you don't have in the A5348 is a continuous analog reconstruction filter. It may work if you wanted restricted range of variclock, but if you want -30/+40%, it's gonna be a problem.niklasni1 said:Yes. For a variable clock system it's more appropriate, though, since you can actually vary the clock
I think you answered my question, though: the write interface is more than capable of keeping up with 8x 125ksps, but the DACs take 8µs to settle -- which is where the 125ksps figure comes from, I'd think --, so it should be fine for 8 channels of <50khz.
For example, the Cirrus 4362 is capable of variclock AND has continuous analog reconstruction filters. It may or may not be suitable for your intended application, though. A typical audio DAC has a 1:2 variclock range.
That's exactly what I wrote. You're gonna need an external filter. This filter will be fixed at the lowest Nyquist frequency. Is that suitable for your application?niklasni1 said:
OK. Your application was not clear so it was a tad difficult to answer. Now it's better.niklasni1 said:
No problem. Everything's fine.
Andy Peters
Well-known member
niklasni1 said:
Without knowing how you're going to control these DACs, or exactly how you'll implement the logic, or what the output of the DAC connects to, an octal audio DAC (like the PCM1680 which I just found through a simple search) is a whole lot less expensive than the AD5348.
-a
Andy,
don't forget, audio converters do little to spec DC performance.
if its a control voltage for a VCA etc, then it'll need decent DC performance.
/R
don't forget, audio converters do little to spec DC performance.
if its a control voltage for a VCA etc, then it'll need decent DC performance.
/R
Andy Peters
Well-known member
Rochey said:
True! I'm not quite sure what he's trying to do with the DAC.
-a
The DAC is for the audio output. Pristine quality is not the goal
The system consists of a microcontroller, RAM, and a DAC and filter and VCA for each voice. I'll ignore the voltage control of the filter and VCA for now.
A tone (ie., a vector of single samples) is stored in RAM with a note of the rate at which it was recorded.
On triggering a tone, the playback rate is calculated as a function of the desired pitch offset and the rate at which the sample was recorded.
The microcontroller runs timers at a high frequency (>10 megahertz), and the playback rate is derived by an integer divison of this, ie. HF/PBR = N. Every N HF clock cycles, the microcontroller moves the next sample from the RAM to the DAC.
This is basically the way the old Akai and E-MUs worked. No sample interpolation.
The system consists of a microcontroller, RAM, and a DAC and filter and VCA for each voice. I'll ignore the voltage control of the filter and VCA for now.
A tone (ie., a vector of single samples) is stored in RAM with a note of the rate at which it was recorded.
On triggering a tone, the playback rate is calculated as a function of the desired pitch offset and the rate at which the sample was recorded.
The microcontroller runs timers at a high frequency (>10 megahertz), and the playback rate is derived by an integer divison of this, ie. HF/PBR = N. Every N HF clock cycles, the microcontroller moves the next sample from the RAM to the DAC.
This is basically the way the old Akai and E-MUs worked. No sample interpolation.
