Media Player Q&A: Q6 Is it more beneficial to handle upsampling in a media player or in a DAC?

Media Player Q&A: Q6 Is it more beneficial to handle upsampling in a media player or in a DAC?

6. Is it more beneficial to handle upsampling in a media player or in a DAC?

Jonathan Reichbach, President, Sonic Studio (Amarra)

Depending on the DAC we feel that a hardware based upsampling is superior. As computers get faster and as algorithms improve it is clear that software based upsampling (sample rate conversion) is quite possible. The best option is to manipulate the actual hardware sample rate directly and do no upsampling in software. If upsampling is required in software, we recommend it be done offline so as to not effect playback performance.
Damien Plisson, Founder, Audirvana, (Audirvana Plus)
It really depends on the DAC and the quality of its internal upsampler. Often x2 oversampling using a very high quality algorithm (such as iZotope 64bit SRC included in Audirvana Plus) inside the player improves the sound quality.

But I personally prefer listening to my 16/44.1 recordings using a non-oversampling multi bit DAC.

Tim Murison, Co-Founder & CTO, BitPerfect Sound Inc., (BitPerfect)
That's a complicated question. Considering only a PCM source, most DACs are sigma-delta DACs. This means they convert the PCM into a format similar to DSD before generating the analog signal. For these DACs, in my opinion, the upsampling done by the DAC is less significant to sound quality than the fact that it has converted the data. For ladder style DACs, the PCM signal is converted to analog as PCM, in these cases I think handling upsampling in the media player is preferable. The advantages being that the PC is more powerful and also might allow more choice of algorithms.
Stephen F. Booth, Founder and Developer, sbooth.org, (Decibel)
In a perfect world it wouldn't matter, but in reality it can make a difference. Also, as with all things audio it depends on your setup! I think if upsampling is necessary that handling it on the software side has several advantages. The first advantage is that software upsampling can be performed using very high quality algorithms. Most, if not all DACs resample internally. Sigma-delta DACs in particular upsample to very high rates. It isn't always possible to know whether the DAC's internal filters have been carefully designed for audio quality or whether they have been designed for economy in manufacturing and power use. In the latter case, upsampling in software can be beneficial because some of the DAC's internal filtering artifacts are moved to outside the audible range. The second advantage to software upsampling is that it can be done in advance. Disk space is so cheap that it is practical to upsample all of one's music to the maximum sample rate accepted by the DAC. This eliminates any overhead caused by upsampling at the time of playback.
Jussi Laako, Owner, Signalyst, (HQ Player)
In the player, since modern computers have more processing resources, and computer is something that already exists if playback software is used. It also allows constant improvements to the algorithms and completely new algorithms without replacing the hardware. Some DACs contain "DSP computers" to perform these operations, or even file playback from a memory card - these are practically blend of the two.

I believe performing all possible operations in the playback software offers best price-performance ratio.

Software-defined radio already changed radios and made current cell-phones possible, "software-defined DACs" can change DACs drastically.

Josef Piri & Marcin Ostapowicz, JPlay (JPlay)
In our opinion using good upspampling algorithm on PC is superior. Also upsampling offline provides better results than upsampling online (in real-time during playback).
Jim Hillegass, Founder and CEO, JRiver (JRiver Media Center)
The place it occurs is not as important as the quality of the algorithm used.
Dr. Rob Robinson, Director of Engineering, Channel D, (Pure Music)
In the media player, undoubtedly.

It’s the goal of sample rate conversion to only shift the sample rate, without adding any distortion or altering (in the case of upsampling) the signal in any way.

Pure Music’s upsampling, in particular, uses a 64 bit wide pipe for the data stream and the essential reconstruction filters (which we designed from scratch in the previous century and have painstakingly optimized for fast performance and a low CPU hit on each CPU platform). Upsampling hardware typically is limited to 32 bit precision, though there are some DSP chips that use 40 or 48 bits. The highest performance is only possible with 64 bit precision. Provided the upsampling algorithm has been correctly designed and achieves the theoretical performance limit, all upsampling algorithms will then give equivalent performance, at least as far as sound quality is concerned (whether designed by Channel D or anyone else), but I think we have an edge in terms of lower CPU footprint. Then, the DAC may operate at a much higher sample rate, and artifacts of the digital to analog conversion process become less destructive to the audible frequency range.

Share | |
COMMENTS
deckeda's picture

Looks like most of these guys think converting in software at least has an equal chance of success, if not better, than converting in hardware. I wonder how that additional task gels with the goal of having the computer do LESS processing while playing back re: not having to decompress lossless files etc.

I once ran Pure Music on an old G4, Dual 1.0 tower. Upsampling 16/44.1 to the DAC's native 24/192 usually resulted in a few stutters and sharp clicks that sounded a lot like speaker voice coils bottoming out (scary!)

"But I personally prefer listening to my 16/44.1 recordings using a non-oversampling multi bit DAC."

Is that a DAC which considers more than one sampling rate to be "native?" (There isn't such a thing, right?) I'm having a vague recollection of Philips or Sony making "1-bit" DACs at one time, which probably has nothing to do with any of this ...

earwaxxer's picture

Personally, I have found that using a SRC to my redbook before sending it to the Transporter yealds better sound. I have played quite a bit with the settings using Sox SRC in foobar. What seems to sound the best, so far, is to upsample to 24/96 at 'best' quality, 95% passband, minimum phase (0%), dither, and allow ailiasing. I also apply album replay gain -7.8db. I use a tube preamp, and these settings seem the best, although I have tried using the Transporter as a preamp with similar results.

Keep up the good work with SRC software! Its game changing IMO.

labjr's picture

I would think the best way is to use software for upsampling and do it ahead of time. Not on the fly while the music is playing.

Larry Ho's picture

This is one of several questions that I have strong believe in mind. ;)

Vincent Kars's picture

In the end all resampling is done in hardware.

You can’t do it without a processor.

Hence the power of this processor is important as you have a very small time slot (1/44100 of a second in case of Redbook).

Each calculation has a finite precision. Doing many calculations means a cumulative quantization error.

Simple DSP chips are often limited in both performance and precision.

On the other hand, they are very efficient as they are designed for a specific task (simply repeat the same calculations all the time, no multi-tasking here).

 

How it turns out in practice is a matter of experimenting.

If you have a upsamling DAC to 24/192 you can feed it Redbook or Redbook upsampled by the media player to 24/192 and check if you hear a difference.

SiCHIPS's picture

Thanks so much for all the valuable info! Many great points cited.

After having heard a good number of Transports as well as CD players, it is overwhelming how each has its own sonic signature! At one point in time, I was convinced that I could tell the difference whether a Ladder DAC or a Sigma-Delta DAC was included. The former being superior in Soundstage and the latter being much more detailed. With the technology advances today, I am holding back making this comparison as it is more difficult.

With my particular system and taste, I can definitely tell a difference between the various media music players and their SRC implementation. I am leaning towards believing the best is to use a software based SRC and also I like the ability to hear the differences created by changing the parameters associated with the filtering. Seems to me I can create sonic signatures that exemplify Transparency and Soundstaging versus Smoothness through the audio range which suits specific albums or tracks. Also, having a High End Transport and a Logitech Transporter using the same ripped AIFF and CD as comparison, I would have to give the edge to the software SRC in that the sonic character can be optimized to taste.

Having said all that, there is a lot to be said with industry experts with ears much better than mine in having them do the fine tailoring of the algorithms! Compute power and extremely low disk space storage cost allows AIFF or FLAC ripping as the major components to achieving this...

For the record, I prefer Upsampling 2X and using the latest versions of either Pure Music, Fidelia Advanced or Audirvana (for now anyway, as others are worthy of consideration) depending on what CD or track I play. Each has their own sonic character and thus advantages. Using these software based programs has allowed me to hear Transparency and Soundstaging that I never though possible from my system. As always, the source must be of a High Quality to demonstrate this which is the #1 limitation of many CDs.

One things for sure, the state of the art now possible has definitely benefited us all!

X
Enter your AudioStream username.
Enter the password that accompanies your username.
Loading