Computer Audio Asylum Music servers and other computer based digital audio technologies.

---

[ Follow Ups ] [ Post Followup ] Thread: [ Display  All  Email ] [ Computer Audio Asylum ]

---

[ Alert Moderator ]

---

Bit Perfect Measurement & Analysis

Posted by cics (A) on December 20, 2007 at 04:06:06

In Reply to: Update v0.3: The Art of building Computer Transports posted by cics on September 14, 2007 at 15:07:18:

Bit perfect transfer measurements can be easily done using Computer Transports. I figured out a simple way to measure this. Only when I received an email recently on the Bit ‘Perfectness’ of XP over Vista for USB audio devices I decided to do the measurement. Reference is made to this manufacturer's web page.

Measuring Equipment

• Source: Computer Transport as specified in this topic. (Mine has a few more enhancements which will be published soon).
  
• Target: Computer with soundcard that acts like a DAC.

Source Computer (Transport) setup:

1. Foobar2000 plays CD track @ 24/96 (giving bit rate of 6.144Mb/s) via SRC upsampler. Choose a loud track to ensure all 24 bits are tested.
   
2. Digital SPDIF output via coax using breakout cable from RME’s HDSP 9652 soundcard
   
3. Provides Master clock

Target Computer (DAC) setup:

1. Slaved to Source
   
2. Digital SPDIF input via ESI’s Juli@ soundcard (with clock set to External)
   
3. Recording software: Steinberg’s Cubase LE (set for recording 24/96 stereo and its hard wired volume control set to 0.00db).

ASIO 2.0 is used in both Source and Target computers. For good measure, I chose the worst possible RCA cable for digital connection. Cable has no shielding and is 2m long with poor quality connectors (I doubt you could get anything worse).

Procedure

1. Target: Start Recording in Cubase
   
2. Source: Play Track (Target computer records nothing in first few seconds until Source starts playing).
   
3. Target: Stop recording after 1 minute or less. Audio .wav file (~35MB) is created in Audio sub-folder of project folder.

Analysis

1. Using foobar’s foo_convert.dll component, convert source track to .wav file (right-click track > select Convert). In converter’s settings, make sure SRC is set for 96k and foobar’s output resolution is 24 bits.
   
2. You now have the source track in a .wav file and the target .wav file as recorded by Cubase via soundcard’s ASIO 2.0 drivers. Both at 24/96.
   
3. Compare .wav files. My preference is not to use wave file analyzers (too much clutter) and instead prefer doing actual sample comparisons at 24 bit level for both channels. Using the cicsWave utility, I extract samples to a csv file which I can view in a spreadsheet. I’ve had to update cicsWave as Cubase uses a slightly different .wav file format and I also wanted more samples to view (increased from 32000 to 64000). (If you need this revised utility, let me know). Make sure source and target waveforms are aligned (as target's .wav has a few seconds of nothing recorded, i.e. the time it takes to hit play on source computer).

Results were brilliant. Plotting the source and target waveforms (overlays) will not reveal any differences! Here’s only the source left and right channels:

Comparing at actual sample level is best. Here’s a snapshot of source and target samples together with differences (delta):

As can be seen, the difference lies in the least significant bit. This I attribute to Cubase’s volume control which although set to 0.00db still has a minor rounding error. Cubase doesn’t offer a volume bypass option (if someone knows how this can be done please let me know). Otherwise, samples match perfectly – this snapshot is typical of the extracted 64000 samples.

Excluding this ‘rounding / volume’ issue, all other (including higher order) bits are absolutely perfect! That’s 23 of 24 bits delivered perfectly with only the least significant bit being compromised (say 50% of the time) due to other factors. IF bits were mangled, lost or just corrupted, then we would NOT gain this consistent level of accuracy. Put another way, mangled bits would cause any random bit to be inverted – this would result in vastly different signal amplitudes which is not the case.

Anyone with a spare computer and soundcard can do this. Much simpler recording software would be ideal (Cubase LE is ‘heavy’ and forces use of its volume control). Please test this for yourself and let us know!

This post is made possible by the generous support of people like you and our sponsors:

Kimber Kable

---

[ Top of Page ] [ Contact Us ] [ Support/Wish List ] [ Copyright © 2024, Audio Asylum, all rights reserved ]

---

[ General ] [ Speakers ] [ Tubes ] [ Vinyl ] [ Digital ] [ Hi-Rez ] [ Video Asylum ] [ Cables ] [ Tweaks/DIY ] [ Music ] [ Films ]

---