Which surround receivers have the best upmix algorithms?

[oivavoi]

What the title says.

I've been intrigued by multichannel audio for some time. I will probably not invest in a multichannel rig in the near future, but nevertheless I remain intrigued. My listening is about 30 percent classical, 30 percent jazz, 40 percent rock and pop, and 10 percent electronica (since this one goes to 11). For classical, there's a lot of multichannel releases. For all the other genres I listen to, not so much. I am therefore curious about whether any surround receivers have good upmixing capabilities. A recent PhD by a German chap involved listening experiments on upmixing algorithms, and the best ones seem to do fairly well. https://d-nb.info/1128906503/34

Looking briefly into the research, it seems to me that SOTA upmixing algorithms involve elaborate ways of separating "ambient" from direct sound, etc.

So: Which surround receivers do this well? I would assume Harman has some know-how on this, but what brands in the Harman family is it worth looking into? How about other companies, like NAD or Emotiva? Oppo? Emotiva offers very good value for money and has dirac live on their hardware, but I guess they don't have the resources inhouse to research upmixing and stuff like that?

And a further question to the resident multichannel gurus, @Fitzcaraldo215, @Kal Rubinson and @j_j (if I may): Do other considerations apply when buying loudspeakers for multichannel rigs, or do the same rules apply as with loudspeakers for two-channel systems? To phrase it in a slightly different way: Will loudspeakers which are ideal for a stereo setup in a given room also be ideal for a multichannel setup in the same room - or are there other things to think about?

 

[svart-hvitt]

You know I like Genelec so you shouldn't be surprised that I advise you to look at how the Finnish company approaches multichannel.

You will not find anything by Genelec to suggest that some speakers are more suited for multichannel than others as long as DSP is utilized.

To me, it seems like The Ones range (say 8341a speakers) are ideal for a 5.1 or 7.1 setup given their point source design. They recently introduced a 15 inch sub to the range of subs (7380a); it goes down to 16 hertz.

I visited a Genelec studio in London last year (https://www.genelec.com/genelec-and-hhb-group-unveil-worlds-first-genelec-experience-centre) to listen to their 11.1 Atmos setup. It is really good. I haven't heard anything like it before or after. The clarity was exceptional.

For a more complex setup, take a look at this 42 speaker system:

https://www.genelec.com/g-livelab

And this recent technical note from one of their engineers:

https://www.genelec.com/blog/immersive-monitoring-perceptive-perspective

When it comes to the filter discussion, I don't quite follow your reasoning. I know of one flexible software package that lets the user play with different digital filters:

https://www.signalyst.com/consumer.html

I look forward to reading the more competent people's contributions to this thread!

 

[oivavoi]

Thanks, svart-hvitt! Interesting. Do you know if genelec offers any upmix algorithms themselves?

With regards to the importance of the upmix filter: the idea is simple. Stereo recordings are optimized for two channels. The mastering and mixing engineers have carefully mixed the signal so that a good reproduction system can create the illusion in our brain that we are witnessing an actual acoustic event. Similarly with multichannel records, they are optimized from the very start to fool our brains by employing more channels (which has always struck me intuitively as being superior to stereo).

But what do you do if you only have a stereo recording, and want to enjoy it on a multichannel system? Unless you know an excellent recording engineer with access to the master tapes who will upmix it for you, you need to rely on an algorithm to do the hard work. But how does this algorithm know that it doesn't mess up the stereo image by sending some information to the surround and center channels? It doesn't know this. It takes a guess, based on mathematics. But the evidence to date, both in published research (Francis Rumsey and others) and from anecdotal experience which can be found online, is that upmixing algorithms often mess up two channel recordings. While native multichannel recordings can be wonderful.

But this German guy in his recent PhD seems to get very good results with the most recent algorithms, which I guess were done by using recent computer software, not home cinema hardware. This made me curious. Probably easier to do it in software!

 

[svart-hvitt]

Thanks, svart-hvitt! Interesting. Do you know if genelec offers any upmix algorithms themselves?

With regards to the importance of the upmix filter: the idea is simple. Stereo recordings are optimized for two channels. The mastering and mixing engineers have carefully mixed the signal so that a good reproduction system can create the illusion in our brain that we are witnessing an actual acoustic event. Similarly with multichannel records, they are optimized from the very start to fool our brains by employing more channels (which has always struck me intuitively as being superior to stereo).

But what do you do if you only have a stereo recording, and want to enjoy it on a multichannel system? Unless you know an excellent recording engineer with access to the master tapes who will upmix it for you, you need to rely on an algorithm to do the hard work. But how does this algorithm know that it doesn't mess up the stereo image by sending some information to the surround and center channels? It doesn't know this. It takes a guess, based on mathematics. But the evidence to date, both in published research (Francis Rumsey and others) and from anecdotal experience which can be found online, is that upmixing algorithms often mess up two channel recordings. While native multichannel recordings can be wonderful.

But this German guy in his recent PhD seems to get very good results with the most recent algorithms, which I guess were done by using recent computer software, not home cinema hardware. This made me curious. Probably easier to do it in software!

You're definitely right in assuming that software filters are much better than the ones embedded in hardware. The Signalyst HQ Player is an interesting tool (toy?) to play with a specific set of filters. It lets you understand why software is better than an expensive box which is outdated in a year or two.

Some of these filters are quite processing heavy. You'd need multiple cores (up to 16) and a state of the art graphics card for processing some of the most demanding filters. Maybe these multichannel algorithms are quite heavy as well?

Why not contact the German engineer to get it straight from the horse's mouth?

 

[oivavoi]

You're definitely right in assuming that software filters are much better than the ones embedded in hardware. The Signalyst HQ Player is an interesting tool (toy?) to play with a specific set of filters. It lets you understand why software is better than an expensive box which is outdated in a year or two.

Some of these filters are quite processing heavy. You'd need multiple cores (up to 16) and a state of the art graphics card for processing some of the most demanding filters. Maybe these multichannel algorithms are quite heavy as well?

Why not contact the German engineer to get it straight from the horse's mouth?

You are probably right that doing it in software is better. Still, I'm a bit skeptical, given that I want to be able to play CDs as well. I also interact so much with computers in my daily life that I would like my hifi to be computer free. Contacting the German engineer directly is an excellent idea though. I know from experience that young academics almost never get emails which start with "hey, I just read your PhD". On the extremely rare occasion that these emails do arrive, we get very very flattered!

 

[bobhol]

I have experience with a NAD multichannel reciever with their EARS system. My main problem is that I don't have a center channel that matches my front speakers. Still on some content (mainly rock & roll) it sounds great. The NAD is one of five amp setups in my basement and I may keep it hooked up for awhile cause it does sound nice in 2 channel or multi. But eventually I'll rotate through the rest none of which are multi-channel.

 

[DonH56]

I just use PL IIx and Neo as implemented in any number of AVRs and processors... Not really sure there are all that many "other" systems around?

The person you want to hear from is @Kal Rubinson !

 

[oivavoi]

I just use PL IIx and Neo as implemented in any number of AVRs and processors... Not really sure there are all that many "other" systems around?

Cool!
That's the thing, there are newer algorithms which supposedly do a better job than the older algorithms which are imlemented in most current AVRs. But I don't know if any AVRs have them.

 

[oivavoi]

I have experience with a NAD multichannel reciever with their EARS system. My main problem is that I don't have a center channel that matches my front speakers. Still on some content (mainly rock & roll) it sounds great. The NAD is one of five amp setups in my basement and I may keep it hooked up for awhile cause it does sound nice in 2 channel or multi. But eventually I'll rotate through the rest none of which are multi-channel.

Thanks, interesting! I wasn't aware of the NAD system. Will check it out.

 

[DonH56]

Well, I played with them a little, but am actually not sure off-hand what is in my Emotiva XMC-1. The Neo music one sounds pretty decent, though I still go back and forth between that and the PLII-whatever version. There are others but those are the two I use consistently.

 

[Sal1950]

I just use PL IIx and Neo as implemented in any number of AVRs and processors... Not really sure there are all that many "other" systems around?

The person you want to hear from is @Kal Rubinson !

Cool!
That's the thing, there are newer algorithms which supposedly do a better job than the older algorithms which are imlemented in most current AVRs. But I don't know if any AVRs have them.

I agree with Don on this one. I've been chasing this dog for years but IMO I don't believe any miracle software will ever come along to turn stereo into anything that approaches the results you get from a discreet source. For pure ambience extraction the ole Hafler l-r trick, maybe adding a bit of reverb worked about as good as anything. LOL. Today I use my AV pre-pro built-in PL II most of the time, I don't know why but it offers a different Music and Movie implementation, and I actually find the Movie coding more pleasing? YMMV and today's AV rigs all offer about half a dozen different coding mixes, pick what you like.
Some of the newest ATMOS, DTS-X, and Auro-3D gear will also offer further options, I am looking forward to upgrading the pre-pro to something that includes these. But again I don't expect any real groundshaking changes for stereo upmixing.
And for sure I wouldn't spend a bundle on some esoteric offering

IIRC AJ of Soundfield Audio liked to use a proprietary implementation that older Yamaha receivers included. I heard him using it at one of his speaker demos a while back and it did do a good job of simple ambience extraction. Can't remember what Yamaha called it, maybe hit him with a email if interested.

 

[Fitzcaraldo215]

What the title says.

I've been intrigued by multichannel audio for some time. I will probably not invest in a multichannel rig in the near future, but nevertheless I remain intrigued. My listening is about 30 percent classical, 30 percent jazz, 40 percent rock and pop, and 10 percent electronica (since this one goes to 11). For classical, there's a lot of multichannel releases. For all the other genres I listen to, not so much. I am therefore curious about whether any surround receivers have good upmixing capabilities. A recent PhD by a German chap involved listening experiments on upmixing algorithms, and the best ones seem to do fairly well. https://d-nb.info/1128906503/34

Looking briefly into the research, it seems to me that SOTA upmixing algorithms involve elaborate ways of separating "ambient" from direct sound, etc.

So: Which surround receivers do this well? I would assume Harman has some know-how on this, but what brands in the Harman family is it worth looking into? How about other companies, like NAD or Emotiva? Oppo? Emotiva offers very good value for money and has dirac live on their hardware, but I guess they don't have the resources inhouse to research upmixing and stuff like that?

And a further question to the resident multichannel gurus, @Fitzcaraldo215, @Kal Rubinson and @j_j (if I may): Do other considerations apply when buying loudspeakers for multichannel rigs, or do the same rules apply as with loudspeakers for two-channel systems? To phrase it in a slightly different way: Will loudspeakers which are ideal for a Not stereo setup in a given room also be ideal for a multichannel setup in the same room - or are there other things to think about?

Not sure if I follow the logic of an "upmixing" capability. If you have a Mch processor and 5 or 5.1 speakers, then just play discrete Mch sources as 5.0/.1 or use stereo for 2 channel sources. Yes, you can fiddle with artificial synthesis of Mch from 2.0. Most all processors have choices of algorithms for that. I do not use any of them and just listen to 2.0 as 2.1 using bass management. I prefer hearing the original stereo mix in that case, not artificial synthesis, though some are not terrible, like DTS Neo6 in my limited experience. I have heard good things about upmixing from 2.0 or Mch to Auro 3D, but that requires a 3D speaker layout, likely not in my future.

If you are concerned about acquiring Mch music releases while you only have a stereo system, it is mostly a non-problem. Downloads might force a choice between a stereo and a Mch version, but few Mch downloads are available at this point. There is no Mch streaming, AFAIK. Meanwhile, Mch optical discs are usually on SACD or BD for music. I have never seen one with Mch that did not also include a stereo mix, selectable on playback. So, you can be fairly confident that Mch SACDs and BDs will play now in stereo, and in Mch later if you upgrade to Mch.

Incidentally, do not worry about 7.1. It adds little over 5.0/.1 and is not worth the expense, as I discovered the hard way.

Speakers? You always want good speakers. They should be able to serve for stereo or Mch, even for center or surround channels. But, avoid the dipole/bipole HT "surround" speakers with a null facing the listener. They do not work well with music recordings.

Most Mch music recordings adhere to the ITU standard, ideally with 5 identical speakers. But, I have only ever seen and heard one audiophile-grade setup done that way. Three identical speakers across the front with smaller surrounds is more common. A subwoofer(s) and bass management give comparable results using smaller speakers. But, it is a good idea to use the same manufacturer's speakers, ideally with identical mid and tweeter drivers all around. This helps voice all channels similarly. DSP EQ helps further by applying the same target curve to all channels.

Me? I use Martin Logan 'stat dipoles all around with Dirac Live EQ and a JL f113 sub. But, my main fronts are large, inherited from my previous stereo when I had no sub. My surrounds are much smaller versions that use similar drivers. My center is a horizontal ML to fit above my TV front and center, but it is also a stat hybrid, albeit with a dome tweeter. But, it works pretty well, and I am happy.

If I were to create a new system someday, I would not use large main fronts, just quality smaller speakers all around, because subs+bass management work so well.

 

[oivavoi]

Thanks @Fitzcaraldo215 ! Very useful input!

As for upmix/downmix algorithms: I'm not concerned with downmixing multichannel. My quesion is about upmixing stereo to multichannel. Yes, I know that many people give the same recommendation as you: Just listen to native stereo. But, I was under the assumption that there had been made progress in the algorithms in recent years. But so far, it seems as if most AVRs just use the old standards for upmixing, and not the newer fancy-schmancy stuff.

Speakers: Thanks. I'm aware about the need for identical front speakers, with the somewhat laxer requirements for surround speakers. What I was wondering was actually more whether particular speaker directivities are better suited for mulitchannel than for stereo. But it seems like good results can be had with different kinds of speaker directivities in surround, just as in stereo.

 

[Sal1950]

My quesion is about upmixing stereo to multichannel. Yes, I know that many people give the same recommendation as you: Just listen to native stereo. But, I was under the assumption that there had been made progress in the algorithms in recent years. But so far, it seems as if most AVRs just use the old standards for upmixing, and not the newer fancy-schmancy stuff.

Don't let them brow beat ya, use what you enjoy. I rarely shut off my upmixing any more. Used wisely it can add a lot of enjoyment.
No, it's not the same ole same ole, there has been progress on the coding over the last 2 decades. Personally I'm anxious to hear what the latest immersive tech can bring to the plate. I've heard a lot of good things of what can be done with ATMOS and the others. The bottom line here is that now we are in a purely subjective realm, there are many options and you need to find the one that fits your grove.

Speakers: Thanks. I'm aware about the need for identical front speakers, with the somewhat laxer requirements for surround speakers.

But, avoid the dipole/bipole HT "surround" speakers with a null facing the listener. They do not work well with music recordings.

Agreed, I didn't think they worked well with anything. I used Paradigm dipole surrounds for many years and never did grow to love them. Mainly the only place I hear dipole/bipole being promoted is by their manufacturers. Dolby, etc all call for direct, Stick to direct radiators.

 

[Fitzcaraldo215]

Don't let them brow beat ya, use what you enjoy. I rarely shut off my upmixing any more. Used wisely it can add a lot of enjoyment.
No, it's not the same ole same ole, there has been progress on the coding over the last 2 decades. Personally I'm anxious to hear what the latest immersive tech can bring to the plate. I've heard a lot of good things of what can be done with ATMOS and the others. The bottom line here is that now we are in a purely subjective realm, there are many options and you need to find the one that fits your grove.


Agreed, I didn't think they worked well with anything. I used Paradigm dipole surrounds for many years and never did grow to love them. Mainly the only place I hear dipole/bipole being promoted is by their manufacturers. Dolby, etc all call for direct, Stick to direct radiators.

To each, his or her own on upmixing. When I had a Mch prepro with what seemed like a gazillion upmixing choices, naturally, I played with that. I compared notes with friends who similarly listened to classical music. They preferred Neo6, and I tended to agree. But, after awhile, we all grew tired of it, especially compared to discretely recorded Mch, which was clearly much better for the classical genre.

It might be different for rock/pops, which is a different ball game because of major differences in engineering, recording venues, etc. And, that might also benefit in some ways just from having more drivers working in the room in the mid-bass, which might suit the music, even if the "envelopment" is artificially concocted.

I just do not think there is sufficient information in a typical stereo recording, especially not from multi-mic mixes, from which to recreate a sense of envelopment with any degree of fidelity to the original. But, if you like the effect, er gimmick, then go for it. I'll stick with the plain stereo, thanks. It just seems a little crazy to me to invest so much in a system's fidelity to the source, then overlay that with an artificial algorithm to create something not there in the original source.

As for Atmos, I have not heard much good about this very much movie-oriented system when used for music upmixing. For example, Doug Blackburn in Widescreen Review has reviewed a number of newer 3D processors over the past year or so. He has been very down on Atmos upmixing with music, but he likes Auro upmixing very much. Auro was developed by a big Belgian music recording studio for both music and movies. But, it is subjective, of course.

Yes, the dipole/bipole surround idea started with the now obsolete, unnecessary THX surround scheme and was developed initially for low bit rate DVD audio, well pre-BluRay. They might have their uses deployed in large numbers around large theater venues, but not in most listening rooms these days as your only surround speakers, especially not for discretely recorded Mch music.

 

[oivavoi]

Thanks, @Fitzcaraldo215 and @Sal1950 !

Useful input. Concerning upmixing: in my personal system, I'm in a certain sense a subjectivist all the way. I just listen to the things that sound good to me. Still, I find it useful to be guided by findings in published research in what I choose to explore, given that I don't think my sensory apparatus functions fundamentally different from the sensory apparatus of other humans.

I'll upload some screenshots from the PhD which are illuminating. The first is a comparison he made, using songs which had both a multichannel mix and a stereo mix. He then compared the artistic/original surround mix to the upmix based on the stereo version. The ambient-driven upmix is actually just as well-liked as the original artistic surround mix!

Also, he did comparisons of music which was mixed for "3D audio", including height channels, with down mixes for ordinary surround and stereo. Pretty clear that people preferred more channels by a significant margin.

Now I tend to agree with @Cosmik that preference testing has lots of limitations and isn't the final word in audio. Directions and tendencies in the material are more interesting, also, than absolute numbers. For me it's more of a hint in a certain direction: this confirms something which seems very intuitive to me, that more channels will be better at rendering replicas of acoustic events. Seeing results from such preference tests confirms, in a way, that what works in theory also works in practice. And that it may be worthwhile to explore whether surround soujd or 3D Sound and the holy grail of stereo upmixing may be to my liking.

 

[svart-hvitt]

FWIW,

when stereo material plays on a multichannel setup you lose the solid left-right orientation, but wouldn’t people agree that the speakers play more effortlessly in a way when summed together in a multi-group? You can play with less volume, but it sounds louder.

If you play music for a group of people, I wonder if multi is the way to go because stereo is intended for the sweetspotters.

Just some thoughts. I would gladly be corrected

 

[Kal Rubinson]

I'll upload some screenshots from the PhD which are illuminating. The first is a comparison he made, using songs which had both a multichannel mix and a stereo mix. He then compared the artistic/original surround mix to the upmix based on the stereo version. The ambient-driven upmix is actually just as well-liked as the original artistic surround mix!

Too many variables.

FWIW, when stereo material plays on a multichannel setup you lose the solid left-right orientation, ..........................

Would that not be a function of the particular up-mixing algorithm?

but wouldn’t people agree that the speakers play more effortlessly in a way when summed together in a multi-group? You can play with less volume, but it sounds louder.

Would that not be a function of the particular up-mixing algorithm?

If you play music for a group of people, I wonder if multi is the way to go because stereo is intended for the sweetspotters.

Not usually. While a good multichannel setup with good multichannel source material may offer a wider sweet spot, it is not likely that it will work for a group of people (more than a handful) because the sweet spot is not that big and because the individuals inevitably block line-of-sight (hearing) from each other. I was never sold on multichannel at the original large demos at industry shows. I didn't buy in until Meridian set one up in my living room.

 

[svart-hvitt]

Too many variables.

Would that not be a function of the particular up-mixing algorithm?

Would that not be a function of the particular up-mixing algorithm?

Not usually. While a good multichannel setup with good multichannel source material may offer a wider sweet spot, it is not likely that it will work for a group of people (more than a handful) because the sweet spot is not that big and because the individuals inevitably block line-of-sight (hearing) from each other. I was never sold on multichannel at the original large demos at industry shows. I didn't buy in until Meridian set one up in my living room.

Regarding sweetspot in multisetup: When walking around in the Genelec 11.1 showroom I think the sound was likable even in the corners. Maybe I’m not a picky guy...

 

[Fitzcaraldo215]

Regarding sweetspot in multisetup: When walking around in the Genelec 11.1 showroom I think the sound was likable even in the corners. Maybe I’m not a picky guy...

I agree with Kal, and sweet spot size is not really an important characteristic to me. I do think it is larger in Mch than in stereo., for whatever that is worth. But, honestly, it does not do corners well, just like anything else. And, even if it did, who sits there, except me occasionally with my dunce hat on because I have not done my chores.