Skip to main content
Topic: Myth on sound chips and sample rates? (Read 885 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Myth on sound chips and sample rates?

Hello,

I'm writing a book on ePub (in ePub too) and part of it includes encoding and embedding audio files.

I do not remember where I heard it but I am having trouble finding a source for it, so it may be a myth? Figured HA was best place to find out, but for a long time I've believed that many consumer sound cards are designed with sample rates of 2^n * 12kHz in mind (where n is non-negative integer) and that when these chips encounter 44.1kHz or 20.5 kHz sample rates, they resample on the fly to 48kHz (or 24kHz) for the DAC to decode into the analog wave form.

If the resampling is done poorly it can degrade the result, thus it is better to resample CD quality audio to 48kHz with SoX or similar tool if distributing to be played on random devices.

Is that belief of mine a myth or is that the case?

Re: Myth on sound chips and sample rates?

Reply #1
I know some Creative Soundblaster cards did this, at least the Live and Audigy cards did, possibly others.

On the other end of the scale, the relatively popular FiiO Taishan D03K DAC only has a 44.1KHz clock, and thus has quite bad 48KHz (and multiples) performance.

Re: Myth on sound chips and sample rates?

Reply #2
My sound card works best with 44.1 KHz and it's an ASUS brand.

The 48 KHz thing probably comes from the fact that some Creative sound cards infamously resampled everything very poorly to 48 KHz.  As now a days it's either your sound card works best with 44.1 KHz or 48 KHz multiples or you don't have any audible issues at all with anything that's supported.

Re: Myth on sound chips and sample rates?

Reply #3
Didn't integrated chips based on AC97 standard also resampled everything to 48 kHz wth poor quality?

Re: Myth on sound chips and sample rates?

Reply #4
Hello,

I'm writing a book on ePub (in ePub too) and part of it includes encoding and embedding audio files.

I do not remember where I heard it but I am having trouble finding a source for it, so it may be a myth? Figured HA was best place to find out, but for a long time I've believed that many consumer sound cards are designed with sample rates of 2^n * 12kHz in mind (where n is non-negative integer) and that when these chips encounter 44.1kHz or 20.5 kHz sample rates, they resample on the fly to 48kHz (or 24kHz) for the DAC to decode into the analog wave form.


DACs don't have a preferred sampling and aren't designed for any specific frequency. The chip just takes a clock and returns samples. For cheaper hardware however there may be a fixed clock which is divided down to create the sampling clock however. In that case you will only get dividends of that clock.

This is mostly cheap junk though. A PLL costs less than a dollar and avoids this issue.

Re: Myth on sound chips and sample rates?

Reply #5
Didn't integrated chips based on AC97 standard also resampled everything to 48 kHz wth poor quality?

I certainly remember having a Compaq Presario Windows XP era desktop computer that had AC'97 for it's onboard sound.  It was the worst system I ever owned.  The power supply failed after 2 years.  The DVD ROM drive started having troubling reading CDs and stopped reading DVDs before it completely died in the less than a year.  The onboard graphics could barely handle a game from 1999.  It's onboard sound was noisy as well.  After a couple of years, I wasn't able to reinstall Windows because the RAM or the motherboard started failing.  The system came with so much bloatware that you could mess up the Windows installation trying to remove it and have to do the system recovery all over again.  It was a nightmare to maintain and do much of anything on it.

Windows XP allowed you to configure the resampling quality and I don't remember anything other than the constant whine of the onboard sound driving me crazy when nothing was playing.  To be fair my Windows XP experience overall was more a miserable one compared to the much better experience I had with Windows 7 on a newer, better quality machine from a different manufacturer that didn't really have any problems with hardware reliability through out it's lifespan.

I know this got a little off topic but if I'm recalling things correctly that maybe it was down to drivers more than anything else as Creative probably said it's card supported in it's own drivers 44.1 KHz and Windows sent it that instead of saying I don't support 44.1 KHz and letting Windows do the resampling to 48 KHz?  How Windows handled resampling before Windows Vista came along was also part of the problem.  With the issue of poor quality resampling (not sure which post XP version of Windows it was) being an issue up until a feature update of Windows 10 finally fixed that.

Re: Myth on sound chips and sample rates?

Reply #6
I remember some years ago trying to take a hearing test on my PC from a website. All of the frequencies to high for me to hear were aliased to frequencies that I COULD hear. Clearly that audio chip had serious problems.

Re: Myth on sound chips and sample rates?

Reply #7
Didn't integrated chips based on AC97 standard also resampled everything to 48 kHz wth poor quality?

Integrated chips typically only have fixed output rates, like 48000Hz, and possibly 44100Hz and 32000Hz, and some higher end also have 88200Hz and 96000Hz or even higher.

All other rates are resampled by the operating system. Windows XP and older defaulted to a low quality resampler, unless you changed the quality to Best, which would be okay for most uses.

It was only the Sound Blaster Live! and Audigy series that included a sound chip capable of resampling in hardware, and they suffered from using a resampler that didn't give enough headroom for inter-sample peaks, so it could suffer from clipping those, causing bad distortion on specially crafted test samples, but not likely anything too noticeable on most normal audio.

The original X-Fi cards, on the other hand, featured high quality sinc resampling in their DSP, using upwards of half of its processing capability to resample and mix streams. Unfortunately, you still need to use OpenAL to take advantage of this on Vista and newer. Fortunately, though, the software mixing and resampling on Vista is pretty top notch, so the only thing you'll be worrying about is processing time.

Most competing cards, like the Asus Xonar series, did everything in software anyway, even their EAX simulation. The cards themselves were just simple DAC/ADC interfaces, and some of the circuitry was pretty bad at least in the Xonar DX I own but no longer use. For instance, when coupling with my case's front panel jacks, it has horribly noticeable crosstalk between the microphone and the headphones, which has never been present when that same interface was connected to the onboard audio.

Re: Myth on sound chips and sample rates?

Reply #8
Kode,

Do you know of a utility that can query the soundcard's actual physical/electrical capabilities?    The driver has to "know" so it seems like there should be a way, at least for hardware that uses the standard Microsoft-supplied (or Apple-supplied) drivers.

Re: Myth on sound chips and sample rates?

Reply #9
On linux lsusb will show that (for usb devices only, obviously):
Code: [Select]
lsusb -v
...
Device Descriptor:
  Configuration Descriptor:
    Interface Descriptor:
      AudioStreaming Interface Descriptor:
        bBitResolution         24
        bSamFreqType            4 Discrete
        tSamFreq[ 0]        44100
        tSamFreq[ 1]        48000
        tSamFreq[ 2]        88200
        tSamFreq[ 3]        96000

Re: Myth on sound chips and sample rates?

Reply #10
On windows you can click on sound settings and select the device properties to show what formats it supports. Note that this is just what the device claims to support. There is obviously no utility that prove the device actually runs at a given setting without loop back testing.

 
SimplePortal 1.0.0 RC1 © 2008-2019