Jump to content
IGNORED

Player resampling & filtering -- implications for the DAC's processing


Feanor

Recommended Posts

I'm aware that some players, (e.g. HQ Player, BeanHead Emperor), or player plug-ins, (e.g. Resampler-V for Foobar2000) do upsampling and HF filtering of lower resolution, (e.g. 16bit/44.1kHz), files before sending the data stream to one's DAC.

 

Presumably the effect in case of NOS DACs, (provided they can handle the higher sample rates), is that there is no conflict or redundancy of processing between the player and the DAC. However what about DACs that oversample internally? I aware that some DAC oversample, let's say at even multiples of the input rate, to enhance their own filtering. Some DACs over sample up to 8x, (44.1 to 352.8).

 

What are the implications of using a resampler player of plug-in with an oversampling DAC? Is the some sort of conflict or redundancy? Or is the DAC's internal processing ineffectuated?

Link to comment

The DAC's internal resampling takes the input from whatever sample rate it is to 352.8 (44.1 and multiples) or 384 (48 and multiples). That bitstream is then run through a sigma-delta modulator to convert it to a DSD-like format before final conversion to analog. If the DAC accepts DSD input it may avoid all of these steps except the final conversion to analog.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment
The DAC's internal resampling takes the input from whatever sample rate it is to 352.8 (44.1 and multiples) or 384 (48 and multiples). That bitstream is then run through a sigma-delta modulator to convert it to a DSD-like format before final conversion to analog. If the DAC accepts DSD input it may avoid all of these steps except the final conversion to analog.

 

Even if DSD isn't supported, by upsampling to the highest rate the DAC will accept, the effect of its filters will be reduced, especially in the audible band. For example any pre-ringing from a linear phase filter will have a much higher frequency if the input has been upsampled part of the way using a minimum phase filter.

Link to comment
Even if DSD isn't supported, by upsampling to the highest rate the DAC will accept, the effect of its filters will be reduced, especially in the audible band. For example any pre-ringing from a linear phase filter will have a much higher frequency if the input has been upsampled part of the way using a minimum phase filter.

 

Yes. The DAC has no way of "knowing" if the input sample rate is from player software or the original recording. So the effect on the DAC, if it is one of the vast majority that internally runs 352.8/384KHz PCM into a sigma-delta modulator, is to replace the internal resampling with the player software resampling.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment
Even if DSD isn't supported, by upsampling to the highest rate the DAC will accept, the effect of its filters will be reduced, especially in the audible band. For example any pre-ringing from a linear phase filter will have a much higher frequency if the input has been upsampled part of the way using a minimum phase filter.

 

So this seems to make sense to me. In the case you explain, the DAC's internal processing, say a linear phase filter, causes less pre-ringing by virtue of the fact that input has been process by a minimum filter with no pre-ringing.

Link to comment

I am not sure that filters always combine in such a linear fashion.

 

There is also the effect of the ADC's filters to consider.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment
I am not sure that filters always combine in such a linear fashion.

 

Linear filters do. It's easy enough to try and see. In the figure below, I have upsampled a single-sample impulse at 48 kHz to 384 kHz in three different ways: linear phase (top), minimum phase (middle), and minimum phase to 192 kHz followed by linear phase to 384 kHz (bottom).

 

filter.png

 

When resampling, the ringing has a frequency of half the lower sampling rate (whether resampling up or down). By using a minimum phase filter to go as high as possible, any damage by further upsampling within the DAC is limited to higher frequencies.

 

There is also the effect of the ADC's filters to consider.

 

True, but there's nothing we can do about those.

Link to comment

 

True, but there's nothing we can do about those.

 

Are you sure? At least one developer I know of favors apodizing filters to counter ADC filter ringing.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment

Thank you by the way for the discussion of how filters work in combination. It's something I am interested in learning more about.

One never knows, do one? - Fats Waller

The fairest thing we can experience is the mysterious. It is the fundamental emotion which stands at the cradle of true art and true science. - Einstein

Computer, Audirvana -> optical Ethernet to Fitlet3 -> Fibbr Alpha Optical USB -> iFi NEO iDSD DAC -> Apollon Audio 1ET400A Mini (Purifi based) -> Vandersteen 3A Signature.

Link to comment

My dac up samples everything to 96/24 using an AK4122 sample rate converter. It sounds much better is this is done with software. This is not quite the same as the dac chip performing the upsampling internally, but the end result should be the same.

 

2012 Mac Mini, i5 - 2.5 GHz, 16 GB RAM. SSD,  PM/PV software, Focusrite Clarett 4Pre 4 channel interface. Daysequerra M4.0X Broadcast monitor., My_Ref Evolution rev a , Klipsch La Scala II, Blue Sky Sub 12

Clarett used as ADC for vinyl rips.

Corning Optical Thunderbolt cable used to connect computer to 4Pre. Dac fed by iFi iPower and Noise Trapper isolation transformer. 

Link to comment
Are you sure? At least one developer I know of favors apodizing filters to counter ADC filter ringing.

 

There is no contradiction what he says... Practically first step gets to define how the impulse response look likes.

 

Here's an example of upsampling result of 44.1k source that is originally down-conversion from 96k PCM dirac pulse using linear phase filter. On top there's upsampled result to 352.8k using apodizing minimum phase filter and the second in the middle one is upsampling result with non-apodizing linear phase filter. On bottom there's the original 44.1k. You can see that the apodizing filter has replaced the impulse response of original filter with it's own.

 

Even if DAC runs this through some upsampling filter again it is very unlikely to be changed at all.

 

apod.png

 

 

P.S. Upsampling 44.1k to 48k likely wouldn't take out the DAC's filter from the figure. But going to 88.2k should be enough that further steps don't alter the impulse response. However, DAC filters could still impose some rounding and other errors and those are best avoided by going to as high rate as possible (least recursive/cascade filters in the DAC).

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...