Frequency response of 192

[uno1234]

I don't really know that much about converter technology. But I was looking at the specs for the 192, and noticed that the frequency response is listed at 20Hz-20kHz. Now, this may be a stupid question, but shouldn't the frequency response be much higher than that, now that it can run at 192 kHz? Can someone (Dave...) explain this to me?

Thanks!

[Park Seward]

Good question. The frequency response of the 001 is 20 to 20k at both 44.1k and 48k. It is possible that in the HD, one low pass filter is used no matter what the sample rate.

[Jeff_Francis]

If you check the fine print, you'll see the the measurements were made @ 48kHz. I'm sure the frequency range is extended at higher sampling rates, though maybe not within the 0.05dB tolerance stated for 48kHz. It doesn't really matter if it is ruler flat to 96kHz. Having the relaxed requirements on the anti-alias filter (analog or digital) is where the big benefit comes from.

In an oversampled ADC, this filter is done digital, so it changes easily to match the sample rate. I'm sure there is a specific filter curve for each sampling rate.

A minor aside: I'd be willing to bet that the noise performance is slighty worse at higher sampling rates, if the noise were measured over the full bandwidth. This is because 96kHz and 192kHz are not oversampled as much as 48kHz. Not any fault with Digi or their design, just the laws of physics.

The higher rates insure that we capture the audible bandwidth (to 20kHz) as accurately as possible, without the ripple and ringing of inaccurate brickwall filtering.

-JF

[Park Seward]

You are right. Specs at 48k only. I would like to see the rest of the story.

[D Clementson]

<BLOCKQUOTE><font size="1" face="Verdana, Arial">quote:<HR>Originally posted by uno1234:
I don't really know that much about converter technology. But I was looking at the specs for the 192, and noticed that the frequency response is listed at 20Hz-20kHz. Now, this may be a stupid question, but shouldn't the frequency response be much higher than that, now that it can run at 192 kHz? Can someone (Dave...) explain this to me?

Thanks!<HR></BLOCKQUOTE>


The spec we provide is for Frequency Response Deviation. According to AES-17, you pick an upper and lower band edge frequency (in our case 20Hz and 20kHz respectively) and then measure the amplitude deviation within that band. AES-17 further requires the same band edge be used for all sample rates.

The actual -3dB point of the product is greater than 20kHz, though. The high-frequency cutoff scales with sample rate; the -3dB point is determined by the digital interpolation filters, and is equal to about half the sample rate. There is an additional slow rolloff on the DAC side above 200kHz so that the noise-shaped image energy doesn't affect downstream systems. That rolloff, cascaded with the digital band-edge filters give a -1dB response at 80kHz with a 192kHz sample rate. This filter has little effect at 48k and 96k sample rates.

About noise, one must always specify the measurement bandwidth when giving noise figures. AES-17 specifies that the upper noise band edge be 20kHz at all sample rates greater than 40kHz. Using that measurement standard, the noise bandwidth does not scale with sample rate.

DC