• QuadraphonicQuad welcomes you and encourages your participation! Treat all members with respect. Please keep all discussions civil, even when you have a strong opinion on a particular topic.

    Do not offer for free, offer for sale, offer for trade, or request copies or files of copyrighted material - no matter how rare or unavailable to the public they might be. We do not condone the illegal sharing of music. There are many places on the internet where you can participate in such transactions, but QuadraphonicQuad is not one of them. We are here to encourage and support new multichannel releases from those companies that still provide them and as such the distribution of illegal copies of recordings is counter-productive to that effort. Any posts of this sort will be deleted without notification.

    Please try to avoid discussions that pit one format against another. Hint for new users: make liberal use of the search facilities here at QuadraphonicQuad. Our message base is an incredibly rich resource of detailed information on virtually all topics pertaining to surround-sound. You will be surprised at what you can find with a little digging!

Matrix Decoding Vs. Upmixing

Help Support QuadraphonicQuad:


Active Member
Jun 18, 2012
Unusually I've had a bit of spare time due to being stuck in the house (thank you Covid-19!) and I've been comparing upmixes done using DTS Neural using varying sources sources against a few decoded SQ and QS matrix quad mixes and discreet transfers and found some quite interesting results. This follows on from discussions on this thread (Listening to this surround UPMIX) which lead me to do some analysis of Money from DSOTM (since it is so familiar to us all I though it would be a good test bed) and I have since tried a couple more. As Jon points out - source is key to getting good results - rubbish in can equal rubbish out.

To further test the accuracy of a matrix decode Vs. upmixing, I analysed another example - this time 10538 Overture from ELO's - The Electric Light Orchestra. For this I have a SQ decode done by OD (DS39f) some years ago. I believe his source was this SQ encoded CD (Electric Light Orchestra - The Harvest Years 1970-1973) which I also have a copy of. For comparison, against the original Quad mix, I used this source (Electric Light Orchestra - The Electric Light Orchestra) which includes a discreet DTS 96/24 transfer of the quad mix and a high res flat transfer of the original stereo version. I am not sure if this stereo version is SQ encoded or not.

Using the ABX plugin for Foobar of the 40 second extracts of the rear channels here (ELO - 10538 Overture Samples) reveals the following:

1) Quad Mix - You can just about hear the clean guitar intro, before the horn swell kicks in at about 9 seconds. Then only strings and drums until the vocals start at 27 seconds
2) Quad Mix (SQ Decode) - The clean guitar is very prominent and the overall mix/separations bears little resemblance to the discreet quad mix. The sound is also quite muddy
3) SQ Encoded Source + DTS Neural Upmix - This is quite close to the Quad mix. The clean guitar is louder than in the quad mix, but quieter than the SQ decode and you can also hear a distorted guitar. Once the strings come in they are both pretty well buried. Vocals are a bit quieter than in the quad mix
4) Stereo Flat Transfer + DTS Neural Upmix - To my ears, this is the closest to the discreet quad mix. The clean guitar is there, but fairly low and you can just about hear the distorted guitar. The vocals are much quieter than in any of the other samples.

It would be interesting to hear these same sources through other processes - eg Penteo/Surround Master etc to see which can get closest to the Quad.