Theoretical New Formats

I want to discuss some of the new formats that have been mentioned
before, but not yet implemented commercially or by amateurs (that I
know of, please tell me if I'm wrong).


Specifically, the formats I am referring to are:

• Multi-bit Delta Sigma, either 5 or 6 bit at 11.2892Mhz or 12.288Mhz (256fs) sample rates or higher, which would effectively be a "RAW" capture from today's best cutting edge low SNR ADC's (there are 1gs/s+ adc's with worse SNR = less effective bits). 
• 1-bit DSD at 256fs and higher sample rates. Anecdotal reports mention that 256fs appears to be the first point in 1-bit audio where the noise and signal don't interfere, therefore removing the need for noise shaping.
• 768 Khz PCM. The highest bitrate PCM I've seen used, but at 16-bit, and only for scientific use. Not sure what ADC's support this, but the Arda AT1401 DAC supports up to 1.5Mhz PCM.
  

The reasoning behind the proposed format of 6-bit 12.288Mhz DSD, is
the same reasoning that created DSD. The highest quality ADC's of
today use a multiple order delta sigma converter, which is than
processed into either 1-bit Delta sigma, or multi-bit PCM,
theoretically loosing some information in either conversion. By
capturing the "raw" output of the ADC, all mathematical interpretation
can be done at a later point, allowing conversion to other 'future'
formats as they appear, or reinterpretation when new conversion
methods become standard.


It should allow capturing the maximum obtainable quality in archiving
priceless audio. All the disadvantages of DSD still hold true, lack of
ability to perform manipulations or process without conversion, but
now that we have moved past monolithic single-bit delta sigma ADC's,
the archival use of single-bit DSD is questionable. I believe the
"best" mathematical conversion is open to debate, as there is no
perfect solution. By capturing the "raw" output of a ADC allows the
highest amount of care in selecting which algorithm is best in down
conversion. It should allow the maximum quality to be obtained, and
allows choosing a possibly superior conversion at a later date should
another option arise.


I believe that this project has many applications beyond simply audio,
perhaps an open-source SEM (scanning electron microscope) would
benefit from a high speed interface for an ADC for capturing data for
processing, Hams could use it for extending the utility of SDR
(software defined radio's) to allow capturing/sending data, or a radio
telescope project could benefit from more precise digitization,
allowing linking up online and using aperture-synthesis to increase
the effective aperture. Ultrasound systems and other medical imaging
devices like MRI or CAT scans could be possibly utilize it. For
medical imaging lower starting costs are important for the developing
world. GPR (ground penetrating radar) could be developed for the
developing world to survey land quickly and cheaply, to find hidden
dangers or resources.


The uses are endless, as we are in a society where there is a never
ending need to capture, create, and analyze information. Open Hardware
and Software is a revolution, and everything entered into the public
domain can be reused and modified to find uses far beyond the original
intention. For this reason I hope to create some momentum behind
developing a General Purpose High speed ADC/DAC interface. Almost no
design details exist as of now, except the proposed format. This is a
similar project to the SDR-Widgit project, but with the expanded goal
of supporting the "RAW" output of modern Delta-Sigma ADC's. Possibly
SDR-Widgit or Audio-widgit will be used as a starting point in
development.


The need for a new archival and mastering format, as well as a format
for scientific research is becoming more needed every day. For studies
of psycho acoustics, echolocation, animals, sonar, ultrasound,
turbulence, aerodynamics, and virtual reality there is no lack of need
for increasingly precise digitization of audio. There are devices sold
today sampling at rates up to 768Khz for Bat studies.  As new
technologies develop, in addition to computers becoming faster at an
exponential rate, the need for easy ways to capture data accurately
have never been greater.


I am hoping to start an open-source community that can help to develop
a high quality high speed digital interface. This project will (*most
likely) require an FPGA and custom programing and hardware to take the
high speed multi-bit or single-bit data and accurately capture or
output it, without conversion to other formats.

Upgrade to Paul Hynes Regulators

Just received the full Paul Hynes regulator set for the Buffalo-II, 4 x 3.3V shunts and 1 x 1.2V series regs, and the Z1P 5.5V DC-input pre-regulator. Planning to run it off either a 60wH 12v lithium battery pack on the go, or a pair of car batteries at home. Hopefully this will increase acuity/resolution with high rate DSD, as the Paul Hynes regulators are widely considered top-notch. After seeing the regulators first hand I know a lot of work went into their design, he has experience making them for many years, and it .

Still have to install it, need to remove some ferrites from the underside of the buffalo-II, quick job with a nice iron. I am planning to simultaneously put the Buffalo-II in it's final case, a small aluminum box, making it quite portable. I got a Greenlee #732 Punch for the XLR/Ethernet jacks, and a drill press to make things as clean as possible.

The only problem will be choosing a solution to transfer the I2S/DSD signals, which are quite low level and very sensitive to placement and length. my options are either LVDS on ethernet or HDMI, ST-Optical as mentioned by Ted Smith, 3 BNC's with SDIF-2, or just plain old I2S on shielded ethernet. The advantage of using I2S would be not needing another super regulator and the associated circuitry for LVDS/SDIF/Optical, but the disadvantage would be possibly compromising integrity/accuracy the signal, or just being unreliable with unexpected problems in different environments (e.g. NYC, or other dense urban areas might cause loss of lock, or audible glitches.)

My immediate solution is to build my USB-I2S interface into a similer small box, and have the two "piggyback" each other, therfore keeping the I2S within its length spec (19cm or less I think), and not having it travel outside of the shielded case. Experimentation with different implementations of LVDS/SDIF and optical will have to happen, as I don't know which is really technically superior, I imagine some of the high speed LVDS chips will be good enough, creating a board with multiple layers and properly placed traces will be the most important aspects of  designing a bridge for dealing with signals as sensitive as this. Possibly a group buy will be in order to lessen costs, I am sure I am not the first who wants a unpopulated pcb of the Fidelix I2S-HDMI Bridge, or at least a cheaper alternative.

I am now very confident with SMD components, after successfully assembling the QRV09 from Sjöström Audio. Remember, the flood and suck method with the right flux makes even 40+ pins and a thermalpad easy. Also if your eyes are not great a stereo zoom microscope is excellent for SMD, try to get one with a boom pole. Last but not least a nice soldering station is priceless, the ease of quick warmup, accurate temperature control, and possibly 2 irons with different tips and heats is a great addition to any electronic workshop. If you can find a desoldering station even better, it will have a vacuum pump built in. I found an old (80's~) Ungar dual iron desoldering station at a hamfest, it has made every soldering job a breeze. For years I used a few Wellers, and than the cheapest weller station, not using the iron it came with, but my trusted (grounded) older weller plugged into it. Reliable, but the ungar cost me less, and somehow heats up in a minute flat to operating temp. Definitely recommend Weller irons to start out as tips and parts are always available.