Too much noise in data using PCM4201 A/D

What's the spectrum of the noise? The specification guarantees 104dB (~17 effective number of  bits) dynamic range for a bandwidth limited (22Hz to 20kHz) and a-weighted digital output. The lower bits will toggle more or less but the device will still achieve at least 104dB of dynamic range for a signal within the stated conditions.

We sample at 48Khz, giving us a range of 10 to 24Khz.

On our test, we applied no signal, the inputs are to ground.

We have between 7 to 8 of the lower bit that appears random.

From your comments, should I assume it comply with specification?

Bye

Jacques

I can't say for sure if it complies with the specification without further details about the noise. If your noise is within the audio band and you only get 16 usable bits, it may be outside the specification.

We check performance internally with a -3dBFS 1kHz input signal, bandlimit and a-weigh the conversion results and analyze the spectrum. This will show us if the part achieves the guaranteed dynamic range.

I have a sample file with a 1Khz signal, can it be posted for you to check?

Jacques

Please post and I'll try checking it.

The file was sample at 64Khz, and is attached.

Thank you for taking the time to help me.

Good night

Jacques

Attached is a spectrum of your audio file. The noise floor looks good but you have peaks at 3,5,7,9,11... kHz. This is probably distortion because the amplitude of the fundamental is high. In any case, the noise itself is quite good.

Good Morning,

Can I ask you what tool did you used to generate the spectrum?

We have been using an old tool that do not give us similar result, so make us thought we have a lot more noise than what you got. We suspect our tool been confuse by the 24 bits format that is interpreted as 32.

The peaks at 3,5... may be a result of the strong input signal as you said, but also from our generator, we will have to investigate.

Bye

Jacques 

I use an Audio Precision AP2700, wave file import, fft analyzer.

Sadly, I can not download the AP2700 software without an account.

So. I will try to find another software to validate our result, do you have anything to suggest?

We are using Cool Edit 2000, it is out of date, and look like if it can not properly evaluate the 24 bits files format.

Bye

Jacques

I don't think that the AP software is available without purchasing the AP hardware.

Maybe there's open source software out there that can do basic signal analysis?

I am searching and I tried some Freeware from the net, up to now, all are too basic application, not really useful for what I trying to do. Our old software is better than that.

Bye

Jacques

Sorry to disturb you again, but is it possible to get the parameters you use to generate the spectrum you did?

Bye

Jacques

I used a 8192 point FFT with a Equiripple (Dolph-Chebyshev approximation) window. No weighing. Frequency sweep from 20Hz to 24kHz.

Good Morning,

I am still puzzle by the results you obtain versus what I got.

From your graphic, it look like I have more than 160db of range, while I do not think it's not possible with 24 bits.

Maybe I am wrong.

Anyway. I attached the result I got.

Bye

Jacques