SQ Shadow Vector Soundfield Mapping

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
If it's the first logic directed Matrix H decoder specifically designed for Matrix H, I think it should decode both Matrix H formats.


Kirk Bayne
 
Well...AFAIK, the (all software) Shadow Vector matrix decoding system can be applied to any surround sound matrix, I think a variable matrix Ambisonics decoder would be useful too.


Kirk Bayne
 
I must ask Kirk.... why do you care? Do you own multiple H encode formats?
Few people do Sonik - including the BBC who probably kept nothing from those few experimental months (why would they?). And it never lived long enough to be used commercially by the BBC. About the only source of H or HJ encoded material was only ever off the air recordings made by enthusiasts (and most of that is lost- I've previously recounted how I threw all mine away with the abandonment of cassettes decades ago!). Today I probably have more hen's teeth than Matrix H recordings of any flavour!
 
Last edited:
Hi Kirk,
Not sure where the 6 positions comes from, however the decoder employed is based on the Lynn Olsen's full 3-axis decode method outlined in his patent. This allowed perfect SQ constant power decoding at any position of the Scheiber sphere. By adopting different equations, the same decode method could be applied to all 2 channel matrix based systems.
The 6 positions are the ends of the 3 axes. all perpendicular to each other. These are orthogonal positions on the Poincare (or Scheiber) sphere. The sphere is shown along with phono stylus motions.

qso-ds.jpg
qim-ds.gif

The 6 positions are:

olive - in phase - lateral phono stylus motion - QS and SQ: F
violet - antiphase - vertical phono stylus motion - QS and SQ: B
cyan - left channel only - left channel phono stylus motion - QS: L - SQ: LF
red - right channel only - right channel phono stylus motion - QS: R - SQ: RF
black - 90 deg phase - clockwise* circle phono stylus motion - QS: H* - SQ: LB
brown - 90 deg phase - anticlockwise* circle phono stylus motion - QS: Z* - SQ: RB

*
- Stylus revolutions are observed from the cartridge end of a standard arm.
- H is the position of the listener's head.
- Z is overhead if the optional decoding and speaker are installed.
 
There was a discussion about Matrix H here before that contained some very interesting information. One of the main contributors was our old friend OD under another name. Sadly in typical fashion he was being disrespectful and condescending toward others, the moderators decided to remove the entire thread!

The gist of that thread was that decoding Matrix H properly is more than simply adding a 60° (or 58°) phase lead to the right channel (or phase lag to the left channel). However IMHO the simple addition of the phase shift before a QS decoder is an easy and effective way to decode H.

It strikes me how similar Matrix-H is to the "Holzer Audio Engineering-Compatible Stereo Generator". That system added a similar phase shift in order to prevent a build up of the centre image when stereo was folded down to mono. The phasey sound it produced was criticized by Neil Young who had his 1969 self titled album withdrawn and remixed without CSG.

Two separate but similar magazine articles depict the adaptor circuit. The Scheiber sphere depicting phase and amplitude from the Hifi News & Record Review is very revealing. It shows just how close we can get with this simple circuit. Swapping the input and output channels brings the H and QS loci even closer. The BBC did other modifications as well, apparently reducing the separation of the vario-matrix decoder a bit, I assume to reduce artifacts.

Obviously by using software the locus can be tracked even closer!
 

Attachments

  • Matrix H adaptor.pdf
    338.4 KB
Last edited:
Will there be 2 Matrix H decode modes - H(76) and H(77+)?


Kirk Bayne

I think you probably meant H (77) and HJ(78) , as per the BBC quad Broadcasts.
And those years, the bulk of the H/HJ quad matrices were in use.

I think I read somewhere that BBC H matrix was actually derived back in 1974 , (but I am a bit vague on that year)
 
I think you probably meant H (77) and HJ(78) , as per the BBC quad Broadcasts.
And those years, the bulk of the H/HJ quad matrices were in use.

I think I read somewhere that BBC H matrix was actually derived back in 1974 , (but I am a bit vague on that year)
A couple of articles from my library! From the May 1977 Wireless World. First is on Matrix H and I attached the modified to H/45J QS decoder 2-pager, the second is from the News pages with short bits on the ITN on Surround Broadcasts Page 2, & Matrix UD45J on pages 3 & 4.
Wireless World - May 1977 - Cover.jpg
 

Attachments

  • Wireless World - May 1977 - BBC Matrix H .pdf
    5.5 MB
  • Wireless World - May 1977 - Surround mentioned in New Of The Month .pdf
    3.9 MB
Last edited:
Just to clarify;

Actually HJ was readied in August 77 , but I don't think the BBC utilized HJ until early in 1978 .

I believe 45j was the designated term for early ambisonics , later to become UHJ sometime after the BBC H and HJ Test Broadcasting....79 or 80 perhaps.

Now I must read Duncan's Wireless World downloads. 😊
 
Well...AFAIK, the (all software) Shadow Vector matrix decoding system can be applied to any surround sound matrix, I think a variable matrix Ambisonics decoder would be useful too.


Kirk Bayne
My understanding of Ambisonics is that it just doesn't work that way. Forget the idea of discrete separation. Ambisonics uses phase plus filtering to replicate the original soundfield, vario-matrix would not work as intended, it relies solely on the amplitude of the signals.

As Fizzy just stated H transformed into Ambisonics with 45J and then to UHJ.
 
Last edited:
https://www.abbeyroad.com/de-mix
De-Mixing used for Matrix (any system) Decoding:


It seems to me that de-mixing tech could be used to separate out the sounds in each direction without using variable matrix decoding.

Anyone tried SQ (for example) encoded test signals to see if de-mixing can do what no (logic directed) matrix decoder can do - provide more channel separation than the basic matrix does for all (nearly equal amplitude) simultaneous sounds in different directions in the same frequency band (of a multiband matrix decoder)?


Kirk Bayne
 
My understanding of Ambisonics is that it just doesn't work that way. Forget the idea of discrete separation. Ambisonics uses phase plus filtering to replicate the original soundfield, vario-matrix would not work as intended, it relies solely on the amplitude of the signals.

As Fizzy just stated F transformed into Ambisonics with 45J and then to UHJ
My understanding of Ambisonics matches yours, it's a completely different approach and when it is working the speakers completely disappear as point sources. I have experienced this with several of OD's UHJ decodes on my system, it's impressive closing my eyes and realising I have lost any sense of where any speaker is. But the way it works with phase and filtering as Par4ken says makes me dubious it can be decoded with shadow vector. All OD did was extract the extra channel that is in the UHJ phase matrix, and then feed the 3 channels (UHJ is missing the height information of full 4 channel B format) into Bruce Wiggins software decoder that converts it to whatever speaker layout you specify, in this case 4.0.
 
My understanding of Ambisonics matches yours, it's a completely different approach and when it is working the speakers completely disappear as point sources. I have experienced this with several of OD's UHJ decodes on my system, it's impressive closing my eyes and realising I have lost any sense of where any speaker is. But the way it works with phase and filtering as Par4ken says makes me dubious it can be decoded with shadow vector. All OD did was extract the extra channel that is in the UHJ phase matrix, and then feed the 3 channels (UHJ is missing the height information of full 4 channel B format) into Bruce Wiggins software decoder that converts it to whatever speaker layout you specify, in this case 4.0.
If it makes the speakers disappear, why do you need a shadow vector decode?

I want to make the speakers disappear with RM, QS, and discrete.
 
If it makes the speakers disappear, why do you need a shadow vector decode?

I want to make the speakers disappear with RM, QS, and discrete.
Speakers can dissappear with very good speakers, in fact that is my test. If you can stare at a speaker and are convinced there is no sound coming from it, them I say its a good speaker.
 
If it makes the speakers disappear, why do you need a shadow vector decode?
Because as ever with OD's decode processes, he's kept his UHJ decode process to himself and it will likely go to his grave with him. I have no ability to perform a decode myself.
 
Back
Top