-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Hyperion APR-16 AV Processor Review
- Thread starter amirm
- Start date
Not sure where did you get that? Probably from the internet
. So please go and ask Dirac and let us know their feedback. There is no Dirac riot at all, it's just the cross support filters implementation.I'm reasonably sure these are convolution filters, but you can go ahead and check and report backNot sure where did you get that? Probably from the internet
Last edited:
I am not the one to raise rumours and riots - I on my part know what has so far being objectively determined and disclosed by Dirac and their partners.
What you are reasonably sure of is of no consequence. It either is or is not and reasonably sure standard is what exactly?
You obviously don't have to do anything and can probably continue spreading your preaching. I have absolutely no reason to correspond with Dirac. Their product is superb and I have no need to get any further information about its functions or performance. There is 218 page thread in this forum where ART was torn apart in little bits and pieces, so no further need for Dirac to tell us anything.
What? I'm just wondering whether all Dirac are the same. I'm not the one speculating on whether there would be a class action law suit ... I'm a huge fan of Dirac and I think it sounds great even on a lowly 4800h, but there is very little transparency the on details of the various implementations. There is no need to be confrontational as you have been above.
Last edited:
I am not speculating. I am taking based on established facts as they are. Dirac and their partners have shared sufficient detail about their implementation and there is still a single ART product. I am not speculating about the class action lawsuit either, just saying that it would be easy to win it if indeed there were different tiers of ART billed at the same price, yet with some level of difference in performance that was not disclosed.
Wondering if all implementations are the same is not helpful unless you have facts to prove it. It just creates another wave of speculations about what it could mean. Nothing personal, but just good order of business is not to appear to spread unsubstantiated rumors if not labeled as such.
can you direct me to one statement by Dirac attesting that all Dirac ART implementations are the same, in particular are the FIR Convolution filters using the same number of taps and the same precision in the math?
Sorry but where did you get the FIR convolution filters and taps and the same precision in math? There are plenty of Dirac guidelines out there but I might have missed this one?
That’s just something I remember from when I studied electrical engineering 50 years ago and worked for the Bell System on how the type of filters Dirac would likely use may be constructed. Possibly they are doing it differently. DSP has evolved but math hasn’t
Having installed 100s of atmos systems with fewer main+height channels than this and 1 to 2 subs I can confidently state the this is completely incorrect.
In a well treated room you can have an amazingly immersive experience with a 5.1.2 system.
As for 16 channels being inadequate at this price point, I agree with the one exception, the performance is potentially superior to the other products in this price range.
I did a little more work on Dirac. One question I always had was how does it correct the lower bass if the taps are limited to 3000 more or less. At 48khz sample rate that only gives you around 16hz of resolution, fine for midrange on up, but useless for the bass. Turn out below the Schroeder frequency Dirac uses biquad PEQ, which is computationally very efficient, does not rely on taps and has no latency which is important for video. This no latency feature is an advantage over a 60,000 raw power convolution. But biquad peq does nothing for modal null filling and resonance absorption. That’s where the ART array concept comes in, using the various speakers to “support” each other and fill nulls, if they can because they are in different locations, and also absorb the resonance. Above the modal region Dirac uses mixed phase FIR since it doesn’t need that many taps at these higher frequencies. This is a very clever, effective and efficient algorithm. You don’t have quite the phase and amplitude precision of an Audiolense 60,000 tap convolution, but you also don’t have the latency and you don’t need a PC to run it. This all said, underneath all this are quite a lot of mathematical calculations. The precision of the math matters, is it 32 or 64 bit? Biquad PEQ can become unstable if they are close to each other. Also 16hz resolution may be perfect at 1000hz, but not so at 150hz. At 150 hz ART is phasing out, so ideally we would want more than 3000 taps so we get a smooth handoff. And you can double that tap requirement at 96khz sample rates. So yes, the math capabilities of the DSP will affect things. This is why I keep asking the question about the details of how Dirac ART is implemented. Perhaps we will never know. Given how many more Dirac licenses are likely sold to D&M users vs Storm and APR16 users, I doubt Dirac has much incentive to tell us whether there is a difference in the underlying math.
Disclaimers: I got my BSEE in 1980 and I only worked as an engineer for 3 years, so I’m sure what I wrote above is at best almost correct. So take it with a grain of salt. But the precision of the math matters.
Disclaimers: I got my BSEE in 1980 and I only worked as an engineer for 3 years, so I’m sure what I wrote above is at best almost correct. So take it with a grain of salt. But the precision of the math matters.
Last edited:
I can’t agree more! My 5.4.2 system in a highly treated room sounds extremely good. The 4 subs do wonders smoothing out the bass, especially since my front mains are also very capable down to 20 hz or even less. And speaker quality matters.
Fidji
Addicted to Fun and Learning
- Joined
- Feb 17, 2019
- Messages
- 595
- Likes
- 1,162
You can not get beyond performance Lvl 1 or “basic artistic intent” with 5.1.2. There is just not enough angular coverage and spatial resolution for immersive formats. I also do not see other immersive audio performance metrics going above Lvl 1.
Most of 5.x.x systems are result of a compromise - typically sofa against the wall, or open plan or other reason that does not allow to place at least 7.x.x properly. So you have wrong or borderline angles, near wall placement etc. Nothing wrong with it, but let's not pretend, this is real high performance. It is best it can be under conditions, if done properly.
I have built my current system from 4.0.0, through 5.2.2, 7.3.4, 7.4.6, 9.6.6 and 9.8.7. Each of the additions improved overall immersion. going from 5.x.x to 7.x.x was big step forward. Also FW get quite often dismissed as not necessary - but if done properly at +- 60dg it makes a lot of difference in creating the bubble around you.
4 subs is absolutely a correct decision. In my opinion 4 subs are the sweet spot in terms of bass quality - smoothness, coverage etc. 3 and less - smoothness suffers, more than 4 - you basically just add SPL and headroom.
Why did you settle for 5.x.2?
Last edited:
Like you explain above, the room doesn’t lend itself to more than 5.4.2. And I’m not much into video or immersive audio. Mostly a 2 channel stereo boomer. That said over 5 decades I’ve tried almost everything regarding 2 channel stereo. After owning many of the usual suspects (Quad, Wilson, Dunlavey, Cello, Magico), I settled on TAD R1s. And I’ve tried everything to get these sounding as good as possible, including having Andrew Jones visit, bi amping the TADs with Andrew’s help, having RPG professionallly treat the entire room …. The only two things that really made a big difference were first having Mitch Barnett design an Audiolense convolution filter that runs on Roon. And then setting up ART with four subs. Of the two I would say ART has much better bass although on a Denon 4800h as a prepro it’s not quite as transparent as the Roon/RME DAC set up. Looking forward to installing my APR16 that Amir recently tested.
I doubt that's true but one never knows, and if one wants to know more, then there is the option to contact Dirac for the information, though they may or may not disclose the info. Until then, I would think that Dirac does the calculations and implements the filters they see fit, based on the information collected by the user/mic, not the manufacturer. Devices that have processing powerful would naturally handle more channels and offer mor "taps", but "higher resolution filters", again I doubt that if you mean in terms of each individual filter, but more filters/taps would naturally mean higher resolution. This is just me trying to think logically, not speculating anything, so again, hopefully someone will contact D+M and Dirac and share their responses to the questions.
The Tide and the APR have more modern and powerful DSP. It does not seem a stretch that these could do quicker (more) math than the DSP in a 3800H receiver which I understand is the same as in an AV10.
Now whether the Dirac app in the Tide and APR takes advantage of this is another question. Who decides how precise the math is, how many taps, how many filters ... is not clear to me.
These are my questions.
I would not assume all Dirac ART are the same until we get a statement from Dirac on this. So far the transparency on what's happening under the covers in the DSP is "clear as mud".
The proposition by other posters that Dirac would be subject to a lawsuit if all ART are not the same seems extremely far fetched.
Also would not be a shock if in a year or two we see something superior to ART or even ART HD, and that this would only run on processors that have greater DSP capabilities. This has been the pattern with Dirac continuously improving their offering. It would not make business sense for them to tie their hands by insisting on backward compatibility, and so far they have not.
As I understand even the AV20, which is a lower priced unit, has a more capable DSP than the AV10. Maybe this tells us something.
Last edited:
welwynnick
Major Contributor
- Joined
- Dec 26, 2023
- Messages
- 1,271
- Likes
- 1,633
According to this video and this video by Sound United Training, both the AV10 and the AV20 use the ADI SHARC+ ADSP 21593 (2-core , 1GHz) with 2,000 MIPS processing power.
Recent Denon AVRs like the A1H, A10H, X4800H and X3800H also use the same DSP.
23:09
09:48
Last edited:
welwynnick
Major Contributor
- Joined
- Dec 26, 2023
- Messages
- 1,271
- Likes
- 1,633
Interestingly, this DSP comparison by Trinnov seems to show that the single ADSP-21593 used in the Denon X3800H and others has slightly more processing power than the 4 x ADSP-21489 used in the StormAudio ISP 32 channel:
You're right. I stand corrected, I had read somewhere that the 20 and 30 had a newer chip. From what I can tell now, it may be newer but not necessarily more MIPS. And we still don't know much about what's in the APR or Tide. Again, would want more transparency.
OP
- Thread Starter
- #240
You didn't say this but to be clear, the latency cannot be dealt with using faster processors. You still have to wait for samples to arrive to process them. Therefore, the longer the filter, the longer the latency.
What you can do of course process more channels, or more accurate filtering, etc.
