Reconstructed signal from DAC0808

[itzme]

Hi all,
I would like your advice on something. Im currently doing a project requiring me to test the accuracy of conversion for my ADC. The ADC chip im using is a ad7819. Im using the dac0808 to check the reconstructed signal. The output from the ADC is directly connected to the DAC. However the Vpeak to peak of the reconstructed signal from my adc is owiz a few mVs higher. Is it the correct output Im obtaining? And Im also required to test the SNR for my signal. Do i meansure it at the input of the ADC or at the DAC?
Thanks

 

[Nigel Goodwin]

For a start you need suitable anti-aliasing filters on the input of the A2D and the output of the D2A, do you have those?. Both also obviously need to use the exact same reference.

 

[Nigel Goodwin]

Yes both are using the same reference. As the input to my ADC is from a Bluetooth headset it is already filtered internally if im not wrong. Do u have a reference to the anti-aliasing circuit? Or do u have a reference to it?
Thanks

Google for it, you shouldn't be playing with A2D/D2A if you don't even know what anti-aliasing is.

 

[Sceadwian]

You'd need to check the voltage into the ADC on a multimeter (or scope) that's in known calibration and compare that with the data the ADC is giving you simultaniously. Or feed it a fixed voltage from a known calibrated reference. The DAC is going to introduce it's own error, so you're not directly measuring the ADC's error unless the DAC itself is in known calibration, which requires approximatly the same steps as calibrating the ADC itself so is a bit silly =)

The equation for calculating SNR can be found here, and would be at the ADC without the DAC, again unless the DAC is in known calibration, otherwise you're calculating the total error of the ADC and the DAC together, not just the ADC.
Signal-to-noise ratio - Wikipedia, the free encyclopedia

 

[itzme]

You'd need to check the voltage into the ADC on a multimeter (or scope) that's in known calibration and compare that with the data the ADC is giving you simultaniously. Or feed it a fixed voltage from a known calibrated reference. The DAC is going to introduce it's own error, so you're not directly measuring the ADC's error unless the DAC itself is in known calibration, which requires approximatly the same steps as calibrating the ADC itself so is a bit silly =)

The equation for calculating SNR can be found here, and would be at the ADC without the DAC, again unless the DAC is in known calibration, otherwise you're calculating the total error of the ADC and the DAC together, not just the ADC.
Signal-to-noise ratio - Wikipedia, the free encyclopedia

Hi thanks so much for your advice again! =] So do you mean I input the ADC with a constant voltage first and observe it via the DAC? Yes, Ive seen the SNR formula! [= But I'm just abit confused on where to measure it [= So i would have to measure the SNR at the input into the ADC you mean? Plus when referring to my ADC data sheet, I understand that my ADC has a SNR of 50dB.

Thanks for the advice. And the diode connection u mentioned previously is working already. Thanks!

 

[Sceadwian]

Don't use the DAC at all, throw it out.
You input a KNOWN voltage to the ADC. You get a digital number on the ADC.
For example if you have a 0-5 volt reference for your ADC every increase in the byte for the ADC value is going to be in approximately 0.0196078431 volt increments. So If your voltage refrence is exactly 2.5 volts you should be getting a binary value of 127.5 which is right in the middle of byte transition so you might be getting fluctuations between 127 and 128 very rapidly this is the aliasing Nigel was mentioning. It's not so easy as you might think to calibrate it not if you want precision. The main problem with calibration is it's only as good as the standards you use. Precision resistors can be used as a voltage divider to feed your ADC for reference voltages, but they're only as accurate as the resistors are (times two cause it's a pair of them)

 

[itzme]

Don't use the DAC at all, throw it out.
You input a KNOWN voltage to the ADC. You get a digital number on the ADC.
For example if you have a 0-5 volt reference for your ADC every increase in the byte for the ADC value is going to be in approximately 0.0196078431 volt increments. So If your voltage refrence is exactly 2.5 volts you should be getting a binary value of 127.5 which is right in the middle of byte transition so you might be getting fluctuations between 127 and 128 very rapidly this is the aliasing Nigel was mentioning. It's not so easy as you might think to calibrate it not if you want precision. The main problem with calibration is it's only as good as the standards you use.

Ah yes. I've already perform testing using a known voltage. And yes there is the error u mentioned at times. But not in all the values. So now when I input the signal from my headset im using the DAC to check the reconstructed signal.

 

[Sceadwian]

itzme then what's the DAC for? If you've tested it against a known voltage they you already know it's accuracy! You're not making sense.

 

[itzme]

itzme then what's the DAC for? If you've tested it against a known voltage they you already know it's accuracy! You're not making sense.

Oh that is caused Im required to check the accuracy of the reconstructed signal. And Im required to check the number of points sampled using the DAC, check the resolution, calculate the conversion time and check the accuracy of it and other analysis as well.

 

[Sceadwian]

itzme, you can't do that with an ADC fed into a DAC, they interact. You're measuring both not just the ADC which invalidates the supposed ADC's accuracy and SNR ratio observed...