Thanks, I like to be wrong when I learn something from it.
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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!
Denon DA-500 Vintage R2R DAC Review
- Thread starter amirm
- Start date
restorer-john
Grand Contributor
BB also had a selection -S only ever seen in the Japanese flagship models of CD players and a few TOTL D/A converters.
infinitesymphony
Major Contributor
- Joined
- Nov 21, 2018
- Messages
- 1,072
- Likes
- 1,814
Yes, I believe "multi-bit" / R-2R / ladder generally refers to the same type of DAC. Popular DAC types like delta-sigma are usually running at 16 to 32-bit so in the literal sense they are also multi-bit.
Some but not all. Here are some R-2R chips, not sure which ones were available in the '80s beyond the Philips.
Analog Devices AD1861 through AD1866
Burr-Brown PCM1702, PCM1704, PCM56, PCM61P, PCM63P
Philips TDA1541, TDA1543, TDA1545
Most of the higher-end DACs from those manufacturers are 32-bit delta-sigma.
All R2R DACs are "multibit" by definition. Current ESS and AKM chips are sigma-delta designs with a multibit sigma-delta modulator (5-6 bits, typically). The era of 1-bit sigma-delta modulators in reasonably high-performance audio applications has largely been over for two decades.
Yup, they definitely could reproduce square waves without ringing. I saw it demoed several times at local Denon dealer in the mid 90s; no ringing on impulse either. Stereophile published measurements showing this for the Denon DA-S1(full review attached). I always wondering how they did that…no ringing yet the dual HF tone IMD test not showing prominent aliasing artifacts from a slow roll-off filter.
I recently ran across an old Denon CD player with the ALPHA processing logo in a thrift store and decided to pick it up to investigate. Using ARTA to make test CDs, I quickly determined that with a noise signal, it had a standard brick wall filter. But, with a single bit impulse signal it had a very slow roll off filter. After playing around a bit more, as best I can tell the sharp and slow filters are both operating in parallel and the processor decides based on looking at the 3 previous samples of the LSB whether to pass along the output of the slow filter or the sharp(LSB not changed in last 3 samples = use slow filter). If there is a large change in amplitude, the processor continues to use the filter it had been using before the large change. So, you can force which filter is being used by dithering or toggling the LSB on and off when you want the brick wall filter, and not dithering when you want the slow roll-off filter.
Here are a few examples:
Example 1: Single bit impulse with or without LSB dithering illustrating the time and frequency domain responses of the two filters.
Example 2: Single bit impulse with dithering turned on/off during different parts of the signal.
Example 3: 1kHz square wave with dithering turned on/off during different parts of the signal.
See Section V. in the attached ALPHA Processor description article.
“…ALPHA processor includes an Automatic Low Pass Filter Harmonic Adjustment feature-the world's first of its kind-which is able to identify the type of incoming signal and automatically select the shut-out band. Because the filter automatically varies according to the incoming data, reproduction of musical data is remains flat very close to the super-high frequency of fs/2. And in the case of pulse data reproduction, the pass band is widened so that the impulse response can be reproduced as intended without ringing…”
So with music CDs, the brick wall filter would always be used. When playing Denon made test CDs the slow filter is used, giving “pretty” impulse and square wave pictures on oscilloscope displays in dealer demos to help sell CD players.
Pulling a page each from an English and Japanese brochure for their later AL24 version of the ALPHA processor. Looks like they moved on to having 3 filters that they switched between based on the abruptness in amplitude change of the signal.
The documents you attached are really helpful! It would be interesting to see how Ultra AL32 performs in the latest Denon products. I still think ALPHA adds a lot of benefits when playing older non-dithered CDs from the 80's and early 90's.
restorer-john
Grand Contributor
How many undithered discs do you acually have (apart from the Denon Alpha demo disc)?
Probably two dozen? I assume that “Super Bit Mapping” was when Sony started dithering? So I assumed that everything DDD produced in the 80’s was probably undithered?
Some of the original stuff from DMP and Telarc is incredible and there are certainly 80’s classic pop as well.
The Pacific Microsonic ADC & HDCD encoder use the same trick : switching between two anti-alias filters depending of high frequency signal content. Then, it encodes the filter used in the HDCD hidden data stream.
It seems that the HDCD decoder (PMD100) don't do anything with this data.
About Denon alpha processing, I don't think that it was just a marketing trick for dealer demos. In AL24 they use 3 different filters so they really try something with variable filters.
I am surprised that, today, nobody try some deep learning adaptativ filter
The Pacific Microsonic ADC & HDCD encoder use the same trick : switching between two anti-alias filters depending of high frequency signal content. Then, it encodes the filter used in the HDCD hidden data stream.
It seems that the HDCD decoder (PMD100) don't do anything with this data.
About Denon alpha processing, I don't think that it was just a marketing trick for dealer demos. In AL24 they use 3 different filters so they really try something with variable filters.
I am surprised that, today, nobody try some deep learning adaptativ filter
People used to say that this digital filter was one of the reasons why HDCD enabled CD players sounded better even with non HDCD encoded discs.
In Ultra AL32, they are using a FPGA with the same power as an old Pentium CPU. They are really coy about what algorithms are used but in blog interviews do say that it’s built from databases of music in Denon (Nippon Columbia) music.
The obvious question is if a deep learning filter results in an audible improvement or not.
There's a lot of active research work being done on applying deep learning to DSP. I haven't been following it closely enough to be able to comment on whether anyone's improved meaningfully on conventional low-pass filter design, but one very recent paper I can recommend showing how good deep learning can be at replacing a traditional filterbank for audio encoding/decoding is this one: https://arxiv.org/abs/2107.03312I am surprised that, today, nobody try some deep learning adaptativ filter
Would you describe this as a nice sounding dac. Considering using one for streaming seeing most I listen to is 16/44 it seems.
As I said, yes to me it's good sounding as a CD Player. But what do you mean exactly by using one for streaming? there is no digital input other than the CD Player. There are digital outs. It's to play CD either with the included R2R DAC or any external one.
I apolagize, I thought you were referring to the DA 500 dac.
The PCM1702 is a monolithic Chip. It requires a minimum of Two for stereo operation. That Sony player used 2x PCM1702-J chips...
Last edited:
Mind numbingly stupid to use such an ancient Digital Receiver, limited to 16 Bits, and poor Jitter Rejection performance, when employing Top of (then) line 20 Bit R2R chips. Even at the time, there were far better alternatives (Crystal, Burr Brown). Quite a nonsensical choice...
Hi everyone,
Same results as you @bolserst, with a Denon DCD-S10 (1994). With dither, a slow filter is active. But with undithered simple test tones, the filter is simply deactivated.
Example below of a square wave as we should see it (output from a Teac VRDS 25x):
The Gibbs phenomenon explains the waves. Now, same square (1kHz @0dBFS) from the Denon DVD-S10:
As you can see, nearly no oscillations.
To better analyze what was going on, I used a square signal at 100Hz to get more odd harmonics and I did an FFT analysis up to 96kHz with the two players (Teac and Denon):
You can see the sharp filter of the Teac, as it should (red). And the Denon shows typical sinc envelope of the conversion, which means it’s running in Non Over Sampling (NOS) mode, only to show a nice square wave.
That’s not all, in this player, but I’m having issues to publish, I’ll complete my message ASAP.
As written in the Stereophile report published by @bolserst, this Denon increases the bit depth before conversion. This can be demonstrated, as stated in the previously attached documents, when using the lowest possible 16bits digital signal (@-90.31dBFS). This signal, in 2’s complement sign magnitude PCM representation, will show, on a scope, the three basic levels of the conversion, -1, 0 and 1. Here is an exemple of it, taken from the digital output of the Denon:
In a standard player, through its DAC, it should show the same, only disturbed by low level noise, such as with the previously mentioned Teac VRDS 25x:
You can still recognize the 3 levels, containing a low level noise. Now, when performing the same measurement with the Denon, we see:
A sine wave clearly appears.
This mechanism will not improve a dithered signal though. I did not bother testing, an exemple is shown in the Stereophile report ans Amir explained in more details in his review.
I have more recent Denon players with Alpha processing, I’ll test them too and report.
It was somewhat a nice tentative of the past to make a product difference. Was it useful? I doubt, but as already stated here, maybe the bit depth increase would have improved low level signals of undithered recordings. I think I still have a lot of these, I should give them a try.
————
Flo
Same results as you @bolserst, with a Denon DCD-S10 (1994). With dither, a slow filter is active. But with undithered simple test tones, the filter is simply deactivated.
Example below of a square wave as we should see it (output from a Teac VRDS 25x):
The Gibbs phenomenon explains the waves. Now, same square (1kHz @0dBFS) from the Denon DVD-S10:
As you can see, nearly no oscillations.
To better analyze what was going on, I used a square signal at 100Hz to get more odd harmonics and I did an FFT analysis up to 96kHz with the two players (Teac and Denon):
You can see the sharp filter of the Teac, as it should (red). And the Denon shows typical sinc envelope of the conversion, which means it’s running in Non Over Sampling (NOS) mode, only to show a nice square wave.
That’s not all, in this player, but I’m having issues to publish, I’ll complete my message ASAP.
As written in the Stereophile report published by @bolserst, this Denon increases the bit depth before conversion. This can be demonstrated, as stated in the previously attached documents, when using the lowest possible 16bits digital signal (@-90.31dBFS). This signal, in 2’s complement sign magnitude PCM representation, will show, on a scope, the three basic levels of the conversion, -1, 0 and 1. Here is an exemple of it, taken from the digital output of the Denon:
In a standard player, through its DAC, it should show the same, only disturbed by low level noise, such as with the previously mentioned Teac VRDS 25x:
You can still recognize the 3 levels, containing a low level noise. Now, when performing the same measurement with the Denon, we see:
A sine wave clearly appears.
This mechanism will not improve a dithered signal though. I did not bother testing, an exemple is shown in the Stereophile report ans Amir explained in more details in his review.
I have more recent Denon players with Alpha processing, I’ll test them too and report.
It was somewhat a nice tentative of the past to make a product difference. Was it useful? I doubt, but as already stated here, maybe the bit depth increase would have improved low level signals of undithered recordings. I think I still have a lot of these, I should give them a try.
————
Flo
Last edited:
Similar threads
