• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

New 110th Anniversary Denon products with "ULTRA AL32"

GXAlan

Addicted to Fun and Learning
Forum Donor
Joined
Jan 15, 2020
Messages
513
Likes
721
#1
With all of the talk in the Marantz 30 product about Denon Alpha processing, Denon just announced their new 110th Anniversary products which really do seem like new versions. The integrated AMP is very different from the PMA-2500NE

"The PMA-2500NE used a one-stage amplification with a high-gain amplifier, while the PMA-A110 adopted a new two-stage configuration with a variable gain preamp and a power amplifier. By incrementing or decreasing the gain of the preamp according to the volume, the noise level can be dramatically improved by amplifying only the power amplifier without amplifying the preamp within the range of commonly used volume."

Interestingly, when the PMA-2500NE came out, they said that the one-stage was better
"Murayama: First of all, we stopped the two-stage configuration of the conventional flat amplifier and the power amplifier, and adopted a high-gain amplifier that amplifies 45.5dB at a time. As a result, the number of elements and circuits passed through was reduced, and the route was able to be greatly shortened, so the configuration became simpler."

The second interesting part is that it's introducing new ULTRA AL32.
1599596068017.png



Ultra AL32 Processing utilizes data interpolation algorithms and supports Hi-Res 192-kHz/24-bit PCM signal input. These algorithms interpolate points that should exist before and after the points in large quantities of data to achieve a smooth waveform that is close to that of the original signal. These algorithms up-sample the bandwidth from the original content, smoothing out the waveform and improving the sound reproduction. Ultra AL32 Processing doubles the processing bandwidth from the previous generation up-sampling 384-kHz/32-bit content to 1.512-MHZ.
 

voodooless

Active Member
Forum Donor
Joined
Jun 16, 2020
Messages
193
Likes
145
#2
Ultra AL32 Processing doubles the processing bandwidth from the previous generation up-sampling 384-kHz/32-bit content to 1.512-MHZ.
Found some more info translated from Japanese: https://news.goo.ne.jp/article/av_watch/trend/av_watch-1275327.html

We are particular about DAC. In order to realize stronger sound and better spatial expression, we have newly adopted Quad-DAC configuration. Two stereo DACs (PCM1795), which are also installed in DCD-SX1 LIMITED, are used for each of the left and right channels (4 channels). The 1.536MHz signal upsampled by Ultra AL32 Processing is divided into 768kHz and input to two (4ch) differential current output type DACs. The parallel configuration using 4ch DAC per one channel makes it possible to obtain 4 times the current output and improve the SN ratio to 6dB. Realized a more energetic sound.
This word salad cannot only be the translation.. how can one divide a signal in half over two DAC’s?and how can they get 768kHz into the PCM1795? It only does 200kHz..
 
OP
G

GXAlan

Addicted to Fun and Learning
Forum Donor
Joined
Jan 15, 2020
Messages
513
Likes
721
Thread Starter #3
how can one divide a signal in half over two DAC’s?and how can they get 768kHz into the PCM1795? It only does 200kHz..
Like the old PCM1704 or 1702, the PCM1795 can use an external digital filter. I suspect this just operates the DAC at the system clock. (This is how the PCM1704 was able to do it as well: https://e2e.ti.com/support/audio/f/6/t/867792?PCM1704-DAC-Is-768kHz-the-maximum-sample-rate-allowed- )

So the Intel/Altera FPGA can send the data to each DAC at 768 kHz.

Now the question is if it is going from 384 to 768 kHz in real-world terms because it immediately gets divided by two. Or have they staggered the DACs like the very old CD players from the 80's that switched channels back and forth. So DAC L1 and L2 are running at 768Khz, but L1 and L2 are time offset by 50% so that your summed analog performance is 1.5 MHz.

@amirm or @restorer-john : Do either of you recall what I'm talking about with CD players that oscillated between the two channels? I can't remember if it was a high-end DAC that then became lower resolution because of that, or the other way around.
 

watchnerd

Major Contributor
Joined
Dec 8, 2016
Messages
8,579
Likes
5,086
Location
Seattle Area, USA
#5
Ultra AL32 Processing doubles the processing bandwidth from the previous generation up-sampling 384-kHz/32-bit content to 1.512-MHZ.
So it uses a bunch of computational horsepower to do upsampling?

Okay....but why bother?

Hobbyists in the computer audiophile crowd have been rolling their own upsampling for years.
 
OP
G

GXAlan

Addicted to Fun and Learning
Forum Donor
Joined
Jan 15, 2020
Messages
513
Likes
721
Thread Starter #6
So it uses a bunch of computational horsepower to do upsampling?

Okay....but why bother?

Hobbyists in the computer audiophile crowd have been rolling their own upsampling for years.
See my posts here
https://www.audiosciencereview.com/forum/index.php?threads/marantz-model-30.15724/#post-503532

https://www.audiosciencereview.com/forum/index.php?threads/marantz-model-30.15724/page-2#post-504197

With Audacity it definitely cannot make a -90 dB sine wave look like a sine wave as the old 90's era ALPHA can do. The claim which Amir will test hopefully in Nov or December is that the Denon algorithms are superior having been built around real music from Nippon Columbia. The fact that Denon keeps doubling down on the algorithm and spending extra money on CPUs suggests that they believe it is worthwhile.
 

voodooless

Active Member
Forum Donor
Joined
Jun 16, 2020
Messages
193
Likes
145
#7
Like the old PCM1704 or 1702, the PCM1795 can use an external digital filter. I suspect this just operates the DAC at the system clock. (This is how the PCM1704 was able to do it as well: https://e2e.ti.com/support/audio/f/6/t/867792?PCM1704-DAC-Is-768kHz-the-maximum-sample-rate-allowed- )

So the Intel/Altera FPGA can send the data to each DAC at 768 kHz.
Ah, yes, I see, that indeed seems to be possible with the DAC. But why all the effort? The 0.0005% THD (which is only a 106 SINAD) that is claimed gets quite a bit worse at higher sampling rates.. at 192 Khz it's only 0.0015%, so just about 96 SINAD (at least we clear CD quality..). Now an external filter might fix this, but why not just use a DAC that already can do fine on it's own.. It's not that there are none to be found..

See my posts here
https://www.audiosciencereview.com/forum/index.php?threads/marantz-model-30.15724/page-2#post-504197

With Audacity it definitely cannot make a -90 dB sine wave look like a sine wave as the old 90's era ALPHA can do. The claim which Amir will test hopefully in Nov or December is that the Denon algorithms are superior having been built around real music from Nippon Columbia.
Like I said in the response to that post: I just two squiggly lines that look the same.. just show the actual sample points. The samples are the only actual data points. The rest is meaningless. You cannot derive from this what the actual waveform on a scope would be.

The fact that Denon keeps doubling down on the algorithm and spending extra money on CPUs suggests that they believe it is worthwhile
Sure, financially it might be.. That does not mean it actually brings anything objectively. If so, why do they have to shroud all of this in such big mystery? If it is really that good.. then just prove it! Saves us the trouble. So many people buy 1000 dollar USB/network/speaker cables, or go to sharlatan doctors.. The companies and individuals selling those products and services also thinks its worthwhile :facepalm:. It's proves absolutely nothing.
 

restorer-john

Major Contributor
Joined
Mar 1, 2018
Messages
5,012
Likes
11,091
Location
Gold Coast, Queensland, Australia
#8
@amirm or @restorer-john : Do either of you recall what I'm talking about with CD players that oscillated between the two channels? I can't remember if it was a high-end DAC that then became lower resolution because of that, or the other way around.
The CDP-101 used a single D/A and toggled between L/R. Many of the later BB PCM-56 based players did the same.

The earliest Philips used 14 bit 4x OS and two D/A converters as the TDA-1540 couldn't be taken up to 352KHz- it coud do 176KHz max. So, they had no delay between channels.

I can't remember the first D/A in a CD player that went differential- I think it was a Technics using 4x 18 bit BB D/As. Plenty of "one bit" machines went differential (balanced) in the early 90s. Most were using one bit and an NPC 8x OS filter chip which was ubiquitous and really good.
 
OP
G

GXAlan

Addicted to Fun and Learning
Forum Donor
Joined
Jan 15, 2020
Messages
513
Likes
721
Thread Starter #9
just show the actual sample points. The samples are the only actual data points. The rest is meaningless. You cannot derive from this what the actual waveform on a scope would be.
...
If it is really that good.. then just prove it!
I agree. That's why I've bought with my own money and sent to Amir with my own money, an original 16-bit Denon ALPHA external DAC that can turn on/off the feature, the Denon test track data, and the request for Amir to compare to actual modern DACs to see if this is something that all DACs do well anyway.

I always put my money where my mouth is.

Amir is really backlogged and this is potentially a very time intensive analysis, so I think our thought process is November or December even. Lots of people seem interested in Denon AVR's and wondering if there's any benefit to AL32 and now that Denon's reinvested in Ultra AL32, maybe new interest will bump it up a few spots in line.

Then Amir will ship the DAC back to me on his own dime.

Amir can always use donations -- maybe we need sponsors for the ALPHA processing tests.
 

voodooless

Active Member
Forum Donor
Joined
Jun 16, 2020
Messages
193
Likes
145
#10
I always put my money where my mouth is
While I full heartedly applaud the sentiment, and will be anticipating the outcome of the experiment, the comment wasn’t really meant towards you ;) I’d rather that Denon proves it, and ASR verifies it.
 

stevenswall

Senior Member
Forum Donor
Joined
Dec 10, 2019
Messages
410
Likes
231
Location
Orem, UT
#12
Does their "upsampling" also do some other kind of processing? Hopefully their engineers understand that you only need so many samples to make a sine wave... It's more like a vector format than a raster one, and adding more points along a defined curve seems like a waste.
 

Beershaun

Active Member
Forum Donor
Joined
Oct 3, 2019
Messages
185
Likes
156
#14
I'm confused by the limited specs Denon published on their website. The anniversary amp, the anniversary AVR, and the x8500 AVR all seem to have similar THD and power levels yet they have different amp designs. So I'm not sure what the measured benefits of these anniversary amps will be over the less expensive x8500.
 

voodooless

Active Member
Forum Donor
Joined
Jun 16, 2020
Messages
193
Likes
145
#15
Does their "upsampling" also do some other kind of processing?
Other than adding samples? Like what?

Hopefully their engineers understand that you only need so many samples to make a sine wave... It's more like a vector format than a raster one, and adding more points along a defined curve seems like a waste.
Traditionally up or over sampling will push the analog filter far up in frequency, so that it can be designed easier, shallower and also has less influence on the pass band. It also allows for optimizing the stop band, giving more control over the brick wall filter. Every modern DAC already does this without the need for external processing.
 

stevenswall

Senior Member
Forum Donor
Joined
Dec 10, 2019
Messages
410
Likes
231
Location
Orem, UT
#16
Other than adding samples? Like what?
Something that used machine learning and was fed a ton of music to understand the correct dynamic range of any instrument playing so that it could undo the destruction of generations of art carried out by studio engineers and labels would be good. Could also be used to reduce noise in an automated way (like MQA claims to do with some remasters when the original was made on a machine that had a known issue.)

I'd pay quite a bit for that.
 
OP
G

GXAlan

Addicted to Fun and Learning
Forum Donor
Joined
Jan 15, 2020
Messages
513
Likes
721
Thread Starter #17
Something that used machine learning and was fed a ton of music to understand the correct dynamic range of any instrument playing so that it could undo the destruction of generations of art carried out by studio engineers and labels would be good. Could also be used to reduce noise in an automated way (like MQA claims to do with some remasters when the original was made on a machine that had a known issue.)

I'd pay quite a bit for that.
So it's interesting because THEORETICALLY, ALPHA wasn't machine learning but it was tested using real recordings of music from the Nippon Columbia era. The demo disc that I have mentions that Denon took Bruel & Kjaer microphones fed straight into a PCM digital recorder. They did minimal mic'ing technique where you just had a pair of microphones placed in an optimal position in a real-world space (to capture the reverberations). They made recordings in operat houses, cathedrals and concert halls in France, Germany and Italy and "DENON developed the ALPHA circuitry by carefully monitoring these recordings with recording engineers." It was all classical music.

Around the same time of ALPHA's development (which was during the Japanese economic bubble), Denon some wild things like this ANECHOIC recording of orchestral pieces which was done by going a concert hall in Japan and then surrounding the entire orchestra in an absorptive enclosure specifically constructed for this project.

https://www.stereophile.com/j_gordo...suitability of auditoria as orchestral venues.

This was done so that they could gain data on what an anechoic orchestra sounded like and then what it would sound like with DSP for add reverberation, etc.

So Denon definitely spent a lot of money in the 80's and 90's on this.

What I cannot determine is if
a) the same way people expect JBL SDP-55 or Arcam's to beat "consumer grade" Denon AVRs, that Denon 2-ch hifi fell out of favor due to sighted bias against mainstream products, but it'll turn out that Alpha is useful technology for 16/44 content the same way a Denon receiver from Best Buy can beat "audiophile grade AVRs"?

b) ALPHA became synonymous with Denon in Japan, so even though the technology is outdated/ineffective for 16/44, Denon still invests in it the same was Porsche 911 is associated with a rear engine design because it's an identify. We don't need to care about ALPHA.

c) ALPHA really works for 16/44 but is no longer needed with even 48 khz mastering and distribution. So it's useful for CDs but not for high-res or movies.

d) ALPHA in receivers is the same thing as ALPHA in their hi-fi setup.

Anyway it's interesting. There is a blog post from Denon Japan that asks the question how they do their algorithm where a symmetrical set of data is interpolated asymmetrically and they just say that it's part of their trade secrets.
https://www.denon.jp/ja-jp/blog/3817/index.html
 

voodooless

Active Member
Forum Donor
Joined
Jun 16, 2020
Messages
193
Likes
145
#18
So it's interesting because THEORETICALLY, ALPHA wasn't machine learning but it was tested using real recordings of music from the Nippon Columbia era. The demo disc that I have mentions that Denon took Bruel & Kjaer microphones fed straight into a PCM digital recorder. They did minimal mic'ing technique where you just had a pair of microphones placed in an optimal position in a real-world space (to capture the reverberations). They made recordings in operat houses, cathedrals and concert halls in France, Germany and Italy and "DENON developed the ALPHA circuitry by carefully monitoring these recordings with recording engineers." It was all classical music.

Around the same time of ALPHA's development (which was during the Japanese economic bubble), Denon some wild things like this ANECHOIC recording of orchestral pieces which was done by going a concert hall in Japan and then surrounding the entire orchestra in an absorptive enclosure specifically constructed for this project.

https://www.stereophile.com/j_gordon_holt/recording_of_september_1989_idenon_anechoic_orchestral_recordingsi/index.html#:~:text=Denon suggests that the anechoic recordings, presumably when,evaluating the suitability of auditoria as orchestral venues.

This was done so that they could gain data on what an anechoic orchestra sounded like and then what it would sound like with DSP for add reverberation, etc.
Very nice story, but so what? What's the actual point of all of this? 99.9% of music most people listen to is not this, so how could it ever serve as a reference? You would think that a better reference would be a collection of high resolution sound tracks of any kind of music, since that should still contain the information that was lost by the downconversion to CD quality.

Anyway it's interesting. There is a blog post from Denon Japan that asks the question how they do their algorithm where a symmetrical set of data is interpolated asymmetrically and they just say that it's part of their trade secrets.
https://www.denon.jp/ja-jp/blog/3817/index.html
A lot of talk, but no real information.. we knew that they upconvert, upsample, and digital filter.. none of these concepts are actually explained, just a lot of unproven assertions are made. I love it how the picture suddenly changes from dB scale to bit scale, without the waveform ever changing. I also love the clear errors in the "upconverted" picture. A kid could do a much better job doing this manually ;) I understand that these are not real examples.. but at least be consistent.



Anyway, yes, in the early days, this technology might actually have been useful. Nowadays however.. Quantization errors are really a non-issue with 24 bit audio, the last 4 bits are noise anyway at the very best of times. And even with 16 bit audio, proper dithering fixes the issue for the most part. It's also not clear why one would need to select different digital filters based on the content..
 

restorer-john

Major Contributor
Joined
Mar 1, 2018
Messages
5,012
Likes
11,091
Location
Gold Coast, Queensland, Australia
#19
Anyway, yes, in the early days, this technology might actually have been useful.
The early Denon classical recordings were all 14 bit, so they could certainly do with some help. Classical recordings highlight the problems very clearly in the low level passages. In my opinion, it was a partial solution to all those "PCM" 14bit classical recordings they already had in their catalog and needed to make money on in the CD era.
 
OP
G

GXAlan

Addicted to Fun and Learning
Forum Donor
Joined
Jan 15, 2020
Messages
513
Likes
721
Thread Starter #20
Anyway, yes, in the early days, this technology might actually have been useful. Nowadays however.. Quantization errors are really a non-issue with 24 bit audio, the last 4 bits are noise anyway at the very best of times. And even with 16 bit audio, proper dithering fixes the issue for the most part. It's also not clear why one would need to select different digital filters based on the content..
This is my belief too. But sometimes music you really enjoy is only available in 16/44. The question is if today’s DACs do a good enough job with up sampling and over sampling that ALPHA is pointless, or is the technology still useful today for 16/44?

The switching of digital filters is what is argued by the figures I posted in the Model 30 thread. Sharp filters have one artifact and slow filters have another. Alpha was supposed to know how to make the adjustments in real time. So you get the best characteristics of both. Today we have a lot of digital filters built into an ESS Sabre and AKM Velvet DAC.

More questions than answers, but going off the AVR performance of Denon flagships against NAD, Arcam, Anthem, etc. I am hopeful.
 
Top Bottom