Onkyo TX-RZ50 Review (Home Theater AVR)

[peng]

Why do you think that a limited PEQ can match your target in a room better than a automatic approach with much higher resolution correction that does spatial averaging?

Because I start with the automatic PEQ and then run REW to see the current result. YPAO seems to subdue the bass response under 500 hz, but does a very good job over 500 hz. My Yamaha has a copy option, so I just take the YPAO measurement, copy it, and adjust the bass region until my REW results match my target curve.

I think Gene and others on AH who did the same way you did, also reported good results. In my opinion it is a logical approach. It seems to me the process of such manual tweak after an auto run is like a using the negative feedback technique in a control system. For example an amplifier is designed to amplify the input signal without distortions, but when measured, we see distortions, so we add a feedback loop to correct.

Same idea with REQ, YPAO, Audyssey may be designed to EQ a range flat, but when measured it was found not flat, so we edit the curve such as if we same bump of 5 dB at 60 Hz, we apply a cut of 5 dB to correct the bump to 0 dB, in this way the edited target curve would be non flat, as the curve would now show a dip instead of a bump, but the output curve would be flat.

So it really is not a case of manual vs auto, but more like with feedback vs no feedback. In this case it is not a feedback loop because unlike that used in the amplifier example, the feedback in the form of the manual PEQ tweak is not a continuous process. It is sort of one off, then we can measure again and may tweak with PEQ the second time, and can repeat the process multiple time, approaching a feedback loop approach, until we stop..

All these are just my own "theory", or actually "imagination" would be a better word, though it sounds logical to me.

 

[hmt]

They usually don't measure flat because they use spatial averaging. Besides that the predicted results of dirac are spot on.

 

[markus]

Same with Dirac's PC beta version, Audyssey actually did better for me, except DL's impulse response looked a little better. I will soon find out if DL3 is worth buying. I hope it is easy to setup so I can take full advantage of the very short trial window. When I beta test it, it was a frustrating exercise because most of time was spent on figure things out.

Is the goal a "better looking" impulse response? In my book this is just another red herring thrown by marketing departments audiophools fall for. Just like "flat is better".

 

[peng]

Is the goal a "better looking" impulse response? In my book this is just another red herring thrown by marketing departments audiophools like to fall for. Just like "flat is better".

Short answer, yes.., but..

I don't feel I know enough to really answer the question to your satisfaction so I would suggest you read up on expert opinions, and may be start with the REW help files. The following links may also help.

Improvements worth considering?

I've not had anyone take a look at my recent setup and results. To me, the frequency response looks great, but the impulse response could be better. However, I have no clue on how to improve the impulse response. Thoughts as to what it is and anything worthy of doing to make it better...

www.avnirvana.com

Driver time alignment

miniDSP is a leading manufacturer of Digital Audio Signal Processors for the HomeTheater, Hifi, headphone and Automotive market. Join our large community of Audiophiles, Engineers and DIYers using our innovative products.

www.minidsp.com

 

[peng]

They usually don't measure flat because they use spatial averaging. Besides that the predicted results of dirac are spot on.

May be but I doubt that's the main reason. You can take one measurement, or 8 positions within a few inches, a foot, or the same positions 8 times you won't get a flat curve. As I mentioned I will be trying Dirac again but last time I tried the beta version, the predicted results was not "spot on" at all, unless we define spot on as +/- a few dB within the EQ'ed range. In fact, Audyssey got it flatter, Dirac (emphasis again: I was beta testing so the results may or may not be meaningful) got the impulse graph a touch cleaner. If you have the actual before and after FR graphs I would be very interested to see it, but please use no smoothing or 1/14 smoothing in the range is say below 300 Hz. Thank you in advance.

 

[markus]

Short answer, yes.., but..

I don't feel I know enough to really answer the question to your satisfaction so I would suggest you read up on expert opinions, and may be start with the REW help files. The following links may also help.

Improvements worth considering?

I've not had anyone take a look at my recent setup and results. To me, the frequency response looks great, but the impulse response could be better. However, I have no clue on how to improve the impulse response. Thoughts as to what it is and anything worthy of doing to make it better...

www.avnirvana.com

Driver time alignment

miniDSP is a leading manufacturer of Digital Audio Signal Processors for the HomeTheater, Hifi, headphone and Automotive market. Join our large community of Audiophiles, Engineers and DIYers using our innovative products.

www.minidsp.com

I don't have any questions these links would answer I'm more interested why audiophiles fall for red herrings like "better looking IR" and "flat is better". Don't they know enough about how to interpret measurements?

 

[markus]

May be but I doubt that's the main reason. You can take one measurement, or 8 positions within a few inches, a foot, or the same positions 8 times you won't get a flat curve. As I mentioned I will be trying Dirac again but last time I tried the beta version, the predicted results was not "spot on" at all, unless we define spot on as +/- a few dB within the EQ'ed range. In fact, Audyssey got it flatter, Dirac (emphasis again: I was beta testing so the results may or may not be meaningful) got the impulse graph a touch cleaner. If you have the actual before and after FR graphs I would be very interested to see it, but please use no smoothing or 1/14 smoothing in the range is say below 300 Hz. Thank you in advance.

Results are "spot on" (unless one of the many bugs messes up the results). Check the Dirac Live thread at AVS for plenty of examples.

 

[MASKINEN]

I don't know about others, for me Audyssey is easy to tweak and it has SubEQ HT that time aligns two subs. Onkyo's DL version apparent cannot do that, and I am not what else it cannot do as DL has more than one version.

Understandable, but for me at least, I’ve never had as good sound/ in-room response as I have now with Dirac Live. It’s amazing!

 

[peng]

Results are "spot on". Check the Dirac Live thread at AVS for plenty of examples.

Thanks, too much pages to go through but I did see this one:

https://www.avsforum.com/threads/the-official-dirac-live-thread.3137722/page-2#post-59518062

If I compare his and mine after tweaking with the app for the range 20-200 Hz, mine is tighter, but his is definitely closer to spot on based on predicted vs actual, though he posted the single position mlp one only. I would have to keep reading to find one that he compared them for 8 positions, in order to do a better comparison.

Anyway as I thought when someone said "pot on" it might just be talking about +/- a few dB within the EQ's range. Edit: since @hmt referred to predicted vs actual, I can see why he called "spot on", he's right, I was wrong. The one posted as linked above confirms what I expected. So no I would not call it spot on but again that's a matter of interpretation. The high frequency part looks amazing for sure. Hopefully within a week I will have some graphs to post once I got the PC version set up.

 

[markus]

Thanks, too much pages to go through but I did see this one:

https://www.avsforum.com/threads/the-official-dirac-live-thread.3137722/page-2#post-59518062

If I compare his and mine after tweaking with the app for the range 20-200 Hz, mine is tighter, but his is definitely closer to spot on based on predicted vs actual, though he posted the single position mlp one only. I would have to keep reading to find one that he compared them for 8 positions, in order to do a better comparison.

Anyway as I thought when someone said "pot on" it might just be talking about +/- a few dB within the EQ's range. The one posted as linked above confirms what I expected. So no I would not call it spot on but again that's a matter of interpretation. The high frequency part looks amazing for sure. Hopefully within a week I will have some graphs to post once I got the PC version set up.

You can make a single point "look" perfect at the expense of other locations. The problem is how to make all points within the listening area better. With minimum phase EQ like Audyssey MultEQ the difference in the magnitude response between points will inevitably stay the same. Examples can be found in the Audyssey thread at AVS.

 

[hmt]

May be but I doubt that's the main reason. You can take one measurement, or 8 positions within a few inches, a foot, or the same positions 8 times you won't get a flat curve. As I mentioned I will be trying Dirac again but last time I tried the beta version, the predicted results was not "spot on" at all, unless we define spot on as +/- a few dB within the EQ'ed range. In fact, Audyssey got it flatter, Dirac (emphasis again: I was beta testing so the results may or may not be meaningful) got the impulse graph a touch cleaner. If you have the actual before and after FR graphs I would be very interested to see it, but please use no smoothing or 1/14 smoothing in the range is say below 300 Hz. Thank you in advance.

How do you come to that conclusion? That comparison has been conducted at AVS and the results were the same as the prediction made by dirac. There is no need for a feedback when the algorithm is good and the filters resolution is high enough. Trial and error is only needed when the user is making adjustments and tries to deal with limited PEQs (as with YPAO). Manual corrections are vastly overstated.

 

[SimpleTheater]

How do you come to that conclusion? That comparison has been conducted at AVS and the results were the same as the prediction made by dirac. There is no need for a feedback when the algorithm is good and the filters resolution is high enough. Trial and error is only needed when the user is making adjustments and tries to deal with limited PEQs (as with YPAO). Manual corrections are vastly overstated.

I'm a little confused by this response. Are you saying DIRAC is perfect?

 

[hmt]

No.

 

[peng]

How do you come to that conclusion? That comparison has been conducted at AVS and the results were the same as the prediction made by dirac. There is no need for a feedback when the algorithm is good and the filters resolution is high enough. Trial and error is only needed when the user is making adjustments and tries to deal with limited PEQs (as with YPAO). Manual corrections are vastly overstated.

You asked me to go there and look so I did, and the one I linked did not show the results spot on with the target. If you see one that did, please do me a favor, link it or post the curves. I did say I have not read all the pages so there may someones posted better results that I haven't seen yet.

To be clear, of course the results would follow the target, but not "spot on", and I am not saying it has to be spot on either. I never used that term, I would be over stating if I did.

Again, if you can tell what you mean by spot on, may be we would see that we actually have the same conclusions on something. I want to avoid splitting hair so I said in the beginning that if it means in a few dB difference here and there within the selected range then we are all good and happy.

 

[peng]

You can make a single point "look" perfect at the expense of other locations. The problem is how to make all points within the listening area better. With minimum phase EQ like Audyssey MultEQ the difference in the magnitude response between points will inevitably stay the same. Examples can be found in the Audyssey thread at AVS.

May be may be not, I checked as much 10 positions, also tried different angles to see how much worse the other locations could get. Surprisingly nothing remarkable really. I do have explanation for that but it will be a long one, can do it via PM though. I lost count how many threads I read over at AVSF over the years lol. I haven't visited the Dirac thread for a while and now it has over 400 pages, numerous different opinions out there for sure, many supported theirs with graphs, there are so many variables..

I am at the point now I think for just enjoyment, I would run Audyssey may be a few time just to confirm the results are consistent , and if I get it flat enough, I would just spend an hour or two with the App to take care some obvious bumps that the auto run did not fix, then check a few positions to see if things are reasonably good there too, and then leave it alone until I have to do it again for other reasons. As I used the THD/amplifiers analogy, surely the lower the better but also surely 0.05%/66 dB SINAD is good enough for me. I know I can get FR for my front speakers and subs to within even less than +/- 1 dB 20-200 Hz or higher if I spent enough time on it, and without making other positions worse, but that's just to have some nice looking graphs. There is no way I can hear a difference either way.

 

[NirreFirre]

Phew, that's a good time reading 19p of discussions that mostly were civilized and pretty balanced on the topic. Appreciated.

I have €0.02 and that is Amir's reviews are not classic reviews, and most are aware of that but the term "review" is perhaps, simplified, a bit broader in its scope (using the device's features, letting the kids use the remote etc)? Imho, that makes the following discussion a bit muddy on the relative weight the poor (or "safe", depending on your perspective) protection circuits for the amplifier. "Real world listening" and dynamic (>18dB crest, ~<5Wrms) program material has its place as inputs in a review. But, as Amir and ASR has shown time and time again, so does strict, objective and repeatable tests to enable us consumers to compare options, find the worst offenders but also the more honest and better performing manufacturers. In long run, we all benefit from both but need to always remember these contexts and write our comments with them in mind.

As a help with that, I suggest an adjustment to the titles to "technical test" or "technical review" (similar to the "side 2" that John Atkinson et al does over at Stereophile). But, it get's a bit redundant when you see the site's name though


Now, on the actual unit. Having no good options to my old university's (Uppsala, signal and systems dep, signal.uu.se) software company offspring, multichannel Dirac to purchase here in Sweden, I do have to add a note on a possible killer feature that no one seems to have mentioned:

Using the web UI with ciuser/ciuser let's every owner have a full (or even more than 100%) access to the system's settings, status, hdmi cable tests and all! The convenience aspect if this works as intended is huge imho. Phone apps will stop being updated and probably not too nice to use in ~5 years but http and REST services would work. And who knows, some guys reverse engineer the API and creates a custom WebUI at a later stage (similar to HPT-1), with even better functionality..

 

[NirreFirre]

BTW, I found the (or part of it at least) research from 2014 that seems to be the basis of Dirac's "new" feature, I haven't read the full thesis so cannot say anything on the quality or how robust or heavyhanded the listening tests were performed but at least there's more science being done in our domain!

The second part of the thesis takes on with three novel equalization schemes, which are based on insights from human perception. First, a general filter design framework based on multiple-input multiple-output (MIMO) feedforward control is introduced. The main objective is to provide means to jointly equalize a single loudspeaker by utilizing all available loudspeakers in a multichannel sound system.

..

Evaluations based on simulations and measurements acquired in representative listening environments strongly indicate that the proposed methods successfully treat several causes which are known to impair sound quality and thus yield improved sound reproduction.

Ph.D. Thesis by Adrian Bahne | Signals and Systems, Uppsala University

Signals and systems, Uppsala University

www.signal.uu.se

Only $65 on Amazon or probably a few Google searches away as pdf..

 

[Sprint]

You probably can say the same if that 23 dB difference is 33 dB. Many people just can't tell the difference between 0.001% and 0.5 or even 1 % THD. I just don't find it unacceptable for a >$2,000 AVP to have that kind of anomaly and would not do anything about it when Gene reported it to them. My concern is, what else could be wrong with it. Amir just measured a Carver amp a lot of people raved about how good they sounded, yet it has very high distortions than any AVRs measured here on ASR. So clearly high THD amps can sound good to a lot of people for sure, but if I buy amps that have the lowest THD, then I know all else being equal (IMD, SNR, DF, FR etc.) my amp will be closer to a straight wire with gain. If I want to spice things up, then I can do it myself with EQ, intentional harmonics etc.. I would buy the CX-A5300 if Yamaha fixes that issue. Again, not about sound quality but a matter of principle, and to avoid having to wonder if something else may be wrong, or not knowing the cause of that issue, would it get worse over time.

Yeah fully understand your opinion. Luckily, that Gene did not find any other issues and due to his recommendations, helped me go for it and I thought if needed, worse case I will use RCA pre-outs for front LR. However I do not see the need as the THD is not audible either to me or also others who I asked to listen for any audible distortions. I also did a AB for pure 2ch with my Topping D10s which also uses ESS DAC (I love them by the way ). I could not hear any difference between Yamaha 5200 2ch and Topping D10s. This increased my confidence and made me decide.

Gene reports 0.05% THD+N for front LR vs 0.002% for other channels. I then tried to find any Youtube samples for distorted sound just to see what it sounds like and in what THD% I could hear. I found a good video here.

 

[dlaloum]

I think this will be referring to the "New Features" alluded to by our Dirac co contributor - and which may be part of upcoming AVR's (eg: Onkyo RZ70/90 )

 

[hmt]

May be may be not, I checked as much 10 positions, also tried different angles to see how much worse the other locations could get. Surprisingly nothing remarkable really. I do have explanation for that but it will be a long one, can do it via PM though. I lost count how many threads I read over at AVSF over the years lol. I haven't visited the Dirac thread for a while and now it has over 400 pages, numerous different opinions out there for sure, many supported theirs with graphs, there are so many variables..

I am at the point now I think for just enjoyment, I would run Audyssey may be a few time just to confirm the results are consistent , and if I get it flat enough, I would just spend an hour or two with the App to take care some obvious bumps that the auto run did not fix, then check a few positions to see if things are reasonably good there too, and then leave it alone until I have to do it again for other reasons. As I used the THD/amplifiers analogy, surely the lower the better but also surely 0.05%/66 dB SINAD is good enough for me. I know I can get FR for my front speakers and subs to within even less than +/- 1 dB 20-200 Hz or higher if I spent enough time on it, and without making other positions worse, but that's just to have some nice looking graphs. There is no way I can hear a difference either way.

Did you look for measurements that also replicated the multi position averaging. Afaik Markus did that once and found out that their were really accurate. Same with a one position measurement in dirac.