This is a review and detailed measurements of the Denon DCD-SA1.
View attachment 391583
As I already wrote in my previous reviews, I like testing CD Players, especially older ones. This is one more proof. My firs review of the
Onkyo C-733 here contains information about my measurements which I try to align with the AES standard. It means that, over time, you can compare the devices I reviewed.
This is my first review of an SACD player, but really my interest is into CD Players, so that’s an exception I made as I needed to know what SACD (and its DSD format) had to offer.
Denon DCD-SA1 - Presentation
Released in 2005, nearly 20 years ago, it was top of the line CD/SACD player at a staggering price (7’000€+ in Europe). This player is one of those “statements” from the big Japanese names, as they like to produce them from time to time.
The conversion is performed by two BurrBrown DSD1792 stereo DACs which were state of the art converters at the time (PCM 24bits and DSD compatible).
Being a Denon, you already spotted the in-house digital filter in the “Advanced AL24 Processing” version here. This is the internal oversampling filter from Denon. It is to note, though, that the AL24 is deactivated when playing SACD (DSD) disks, since it works only with PCM data, of course. It is of equal importance to know that the player offers a mode where we can convert DSD to PCM and so benefit again from AL24 processing, which is interesting.
View attachment 391584
As an SACD player, we can play CDs, of course, and the Denon DCD-SA1 adds to that two digital inputs to benefit from its internal DAC. I like to see that possibility on CD players.
RCA and XLR are present, as we would expect.
Weighting an incredible 22kg (48.5lbs), this alone talks about the build quality. The top cover weights 3.5kg (7.5lbs) and is made of a top aluminum thick layer with copper bottom plates, both rubber-isolated in between, and all of it is of course decoupled from the sub-frame via several silent blocks
View attachment 391585
A special attention was given to power supply and isolation from vibrations. I let you have a glimpse at the inside and you can also look up on the web for more details about the engineering that went into this player:
View attachment 391586
The two power transformers are placed in an aluminum housing filled with resin in a two-steps process so that they are decoupled from their respective casts. Additionally, the two aluminum housings are again decoupled from the chassis using rubber material of different resonance to kill mechanical vibrations:
View attachment 391636
It is an absolute delight to use this player. Each button on the front face reminds you of the price you paid. The drive is also very fast which I like a lot.
View attachment 391587
Denon DCD-SA1 - Measurements (Analog outputs - From CD)
From now on, I will be consistent with my measurements as I described them on the
Onkyo C-733 review. So over time, this will help comparing the items I reviewed.
From RCA, the Denon DCD-SA1 outputs 2.134Vrms, that is 0.6dB above the usual 2Vrms. The XLR output is very close at 2.09Vrms.
Let's start with the standard 1kHz sine @0dBFS (dithered) from my test CD (RCA out):
View attachment 391588
Left and right channels are shown but only one gets evaluated in that view. Both channels have the same performances, though. Plot is on H2 (-97.6dBr and -99.5dBr).
THD+N is limited (and so is the SINAD) by the dither noise present on my test CD. It's the best we can get. It shows 92dBr on the dashboard, but that is because I decreased the input gain of my interface to respect my own protocol which I documented with the
Onkyo C-733 review. The truth is 93dB.
And by the way, proof of that is if I measure the XLR out, and because of 1dB difference in the input gain, THD+N is now 93dB:
View attachment 391589
Note that the THD is a little better on left channel because H2 is very much lower (-116dBr vs -102dBr).
Basically, all measurements are nearly identical between RCA and XLR, so I won’t make a difference between them in the rest of this review.
Other results (not shown) are:
- Crosstalk : -133dB at 1kHz and 10kHz (thanks to conversion in mono-mode).
- SNR : 97.3dB (1kHz @-60dBFS, no dither)
- IMD AES : -94.dB (18kHz+20kHz 1:1 @-5dBFS)
If you compare with the
Denon DCD-900NE, you see that we are very close. Basically that is nearly the best we can get from CD Audio.
I suppose you saw a very silent power supply above, so all the efforts from Denon there paid off. RCA only showed a very small power supply related spike at 50Hz (I live in Europe) and at a very low -125dBr. XLR outputs showed nothing, which means better rejection of power supply related noise:
View attachment 391591
Bandwidth measurements showed a gentle roll of at 20kHz (-1dB) starting very early (10kHz):
View attachment 391593
This is exactly the same as seen with the Denon DCD-900NE and that is logical because it’s an effect of the Denon AL24 oversampling filter. I’ll come back to that later.
I was disappointed to see a 0.2dB difference between the two channels (same with RCA and XLR).
Multitone test showed no issues:
View attachment 391594
CD Audio content is more than safe from distortion, no surprise considering the previous results.
Jitter test (16bits/44.1kHz) is nailed too, exhibiting a beautiful trace:
View attachment 391595
Red trace is what is on the test CD (digital output), it can’t be better. The Denon (blue trace) does not add any jitter and really minimal noise.
If you go check the
Denon DCD-900NE review, you will see the same, meaning that the few very low level (blue) spikes come from my measurement interface, not the player.
And I forgot to add one of my favourite measurements, the THD (excluding noise) vs Frequency at @-12dBFS:
View attachment 391596
The Denon DCD-SA1 had no issue here, it is the best trace that I have in my collection of this specific measurement. This an easy test for advanced 1bit DACs. I like this measurement because it shows lack of linearity already at this level with older R2R architectures, and some lower resolution 1bit architectures too, that I enjoy testing.
Denon DCD-SA1 – AL24 measurements
As with the
Denon DCD-900NE, I think the Denon proprietary oversampling filter deserves a specific section, as it finds its roots back
3 decades ago.
As
@bolserst wrote
some time ago ago about Denon filtering, the first iteration of ALPHA processing featured an automatic filter selection based on LSB toggling, and which I could replicate too. Subsequent version of ALPHA processing included further intelligence in terms of filter selection.
I'll try to keep this section as simple as I can, but it's a challenge (again). The situation is actually the same as with the
Denon DCD-900NE. So, I’ll use a different standardized test, to show the same behaviour.
First, this is the filter response (from white noise) overlaid with the standard CCIF IMD test (19kHz+20kHz 1:1) which a lot of reviewers like to use:
View attachment 391598
At the moment, please forget about the filter response (in blue) between 60k and 72kHz.
Now, those of you used to perform and/or look at these tests will see an impossibility here. It is an obviously slow filter response (as we can see from the white noise), and so it's not logical to see total absence of aliases of 19kHz and 20kHz which would show up at 25.1kHz and 24.1kHz respectively.
Well, that is because the Denon (its AL24 filter) recognizes the typical test tones and switches to a sharp filter in that case, which would make people like me (theoretically) happy. Fail
To counter the test detection by the filter, it is enough to add a third test tone with this standard CCIF test. So adding a 9kHz test tone defeats the detection of the AL24 filter, and here below we get what we should:
View attachment 391599
Very much different, indeed! This time we see what's logical with a slow filter response: aliases of 19kHz and 20kHz replicate around 22.05kHz. So you find them at 25.1k and 24.1k respectively. The same goes with other standardized tests such as AES, DIN. When “detected”, the AL24 filter switches to a sharper mode, to exhibit what testers like to see : absence of aliases out of band.
Also, as with the first iteration of ALPHA processing, it detects square signals and switches to NOS (Non Oversampling) mode in that case. This allows Denon to show perfect square waves. When looking at the same in frequency domain, we get this:
View attachment 391600
This is beautiful and could be used by a teacher at the university to talk about D/A
conversion and its effect on creating aliases, enveloped into a sinc function. This garbage is on purpose, only to show perfect square waves, when requested.
The two filter modes I showed, Sharp and NOS, are not activated during music playback. Their purpose is only to shine during very specific tests.
As I commented about the
Denon DCD-900NE, this has been ongoing for decades and continues today. If you want to go deeper into this, I recommend you to read the
Denon DCD-900NE review, as it is exactly the same, and I already compared this DCD-SA1 with the DCD-900NE.
About the situation between 62kHz and 72kHz, that has minimum effect when playing music. Again, please check the
Denon DCD-900NE review, I’ve put measurements, they are the same with the DCD-SA1. Even if it's not clean, we shall not worry too much about that specific noise (it will remain at very low level, ie -100dBr).
Denon DCD-SA1 - Measurements (Optical Out - From CD)
Some of you like to know if the player can be used as a transport. So I measured the digital output, from my test CD.
The below view shows what's on the test CD (1kHz @0dBFS):
View attachment 391601
It can't be better than that, this is what's recorded on my test CD.
Equally, and again about the digital output, I already
suggested here the use of an
undithered 1kHz sine at -90.31dBFS to verify the quality of the drive, should we have doubts. With 16bits, the signal should appear (on a scope) as the 3DC levels of the smallest sign magnitude digital signal, which is what we get with the Denon:
View attachment 391603
This, and other measurements I performed on its digital outputs, made me confident that the Denon is a prefect transport for those who want to use it with an external DAC (for CD audio only, of course).
Denon DCD-SA1 - Measurements (Digital In - RCA out)
Very quick feedback about using the digital inputs of this player with higher resolution PCM input.
The THD does not change (of course) and the noise improves because of the bit depth increase. Unfortunately, with "only" 2Vrms output, my interface (Motu ultralite mk5) reaches rapidly its limits (because of no auto ranger like an AP, it would be much more at ease with 5Vrms at least) and so I essentially measure its noise floor:
View attachment 391605
The noise floor is lower, but again I can’t measure it reliably. What I see, from playing with dither at lower levels, is that the noise is below -105dB, and so that is very good for the time. Power supply spike appears from the reduced noise floor (and because of 256k FFT length + 32 averages) at below -130dBr (I say negligible
).
One thing I did not mention, because I’m keeping it for the SACD measurement section, is that below full scale (0dBFS), the THD performance increases. Let’s have a look.
Denon DCD-SA1 - Measurements (Analog outputs - From SACD)
For these tests, I used the Denon Audio Check SACD. There are only so few test tones on that disc, but it’s informative anyways.
The test SACD of Denon contains test tones at -16dBFS. It’s far from the 0dBFS we (and I) are used to use. That does not help for comparison, so I had to adapt to that.
First our regular 1kHz test tone at -16dBFS (XLR out):
View attachment 391607
Pay attention to the fact that the trace in
dB relative (dBr). Since we are at -16dBFS, this has the effect to visually increase the noise floor by 16dB.
At this much lower level than full scale, we see that THD remains below -100dBr, which is very good.
The calculated noise floor is around -105dB which I’m sure is the noise floor of my measurement interface. Indeed, because the Denon output “only” 2Vrms, I have to increase the input gain of my interface by 15db! This is to respect my own protocol of measurements to get the ADC of the interface close to 0dBFS when capturing data. This allows me to get dashboard measurements from REW software that are correctly calculated, especially the ENOB. Since I’m almost only measuring CD Players, this limitation is not a problem, as a 16bits test tones with dither (which I use) have a noise floor of -93dB. Of course, here the SACD (DSD) lower bit depth puts my interface in an uncomfortable zone. On one side, that is good news for the Denon DCD-SA1.
If I go more into SACD testing, I will need to figure out a way to increase the voltage before feeding my interface. I suppose I could use an extremely transparent Preamplifier to do that. I’ll report here if I get there.
All that said, and since the Denon DCD-SA1 offers the possibility to convert DSD to PCM on the fly, and therefore benefit from the AL24 filtering, let’s see if that changes something;
View attachment 391608
Well, no. But that’s not all.
Let’s play a nasty game. I remember at the time of SACD release, I’ve been told that the extended bandwidth of DSD was a major improvement, along with the increased bit depth. If I can already testify on the later, not yet on the former.
Because of its nature (1bit stream) the DSD generates a lot of mathematical errors, say imprecision. Noise shaping comes into play to reduce these errors and to reject them out of audio band (beyond 20kHz).
So, let’s have a look at a measurement of the same 1kHz sine (still at -16dBFS) but using a wider bandwidth, up to 48kHz, allowing the software to calculate noise and distorsion up to 48kHz:
View attachment 391611
Hmmm, what we see here is a lot of noise created as soon as we go over the 20kHz theoretical limit of our ears. This is an effect of Noise Shaping. So what’s preventing a lot of noise in audio band is creating a lot of noise out of audio band. And that’s really a lot, because if you look at the dashboard, and I left it calculate the noise and distortion up to 48kHz, this is a huge difference. The calculated ENOB is now 10.5bits, ouch...
This typical noise accumulation, out of band, can be found on all tracks of the Denon Audio Check SACD.
Now, let’s try when converting DSD to PCM before feeding the DACs, which means the AL24 filter kicks in too. Here you go:
View attachment 391612
That view demonstrates that the noise generated by the noise shaping technique is part of the test Denon test SACD, it’s not generated by the player. When converting DSD to PCM, the AL24 improves the filtering by removing this out of band noise present on the track (not all, but some of it). ENOB improves by 2bits.
Last and not least, let’s feed the same test tone, but this time from my interface (24bits/96kHz), directly into the digital input of the Denon (and therefore in PCM, of course):
View attachment 391618
And voilà, despite noise included up to 45’600Hz, we keep very close performances to what we initially saw when limiting the calculation at 20kHz (again I’m nearly sure that Noise is limited by my interface).
Is that a fair test? Yes and I no. Again, I remember the promise to get more than the CD limitation at 20kHz from SACD. But if it is to get that amount of noise,
I wonder.
Because again, when playing pink noise or music from the Denon Audio Check SACD, I see the same noise on all tracks. This is below an analysis of musical content (1700+ FFT averages over more than 5min) with Pink Noise:
View attachment 391620
The track (#27) in the Denon audio check SACD is
Mahler: Symphony No. 2 in C minor "Resurrection" - 5th movement closing part (F.!) (Conducted by Wenceslav Neumann, Czech Philharmonic Orchestra, the choir, Gabriela Benachikova-Chapova (soprano), Eva Landova (alto))
I overlaid the SACD layer (green), CD Layer (red) and conversion from DSD to PCM (blue) of the same track to show the differences. This is in Linear Frequency Scale for a better view. Here we see that the only potential advantage of SACD is between 20kHz and 25kHz where noise is still below musical content, From 25kHz on, the shaped noise of the SACD track takes over.
Of course, that is from Denon test SACD. But isn't it supposed to be a disc containing “reference” material? If yes, can we get better than that?
Conclusion
That was a long review! Thanks for those of you who made it!
How to conclude?
- Used as a CD player, in audio band, these are very good results for a 20 year old player, nearly best in class. Distortion could have been better, yes, for those who can hear -100dB distortion relative to 0dB signal.
- Used as CD transport, it is flawless, no surprise (only CDA).
- Best performances are obtained from digital inputs, thanks to Denon to let us benefit from the DAC’s full performances as we want. No DSD input though, but I guess the DSD to PCM convert will make you think about the real need for that feature.
- The behaviour of the AL24 filtering is funny, designed to shine under measurements. It's been ongoing for a long time. The Denon DCD-S10 was already including some tricks and the very new Denon DCD-900NE carries that heritage.
- This is an incredibly well-built CD/SACD player, I love it, one we see once in a decade. It’s still alive and kicking!
- SACD measurements left me wondering what was into this move from Sony and Philips. I did not see anything tangibly good for me, especially knowing that I used a referenced SACD Test disc for the measurements, which I suppose has been created with great care(?).
- Noise shaping techniques have improved, allowing the theoretical 16bits of CD Audio to be actually much lower. What's left to SACD when compared to CDA is that extended bandwidth which I did not find here, at least from the Denon Audio Check SACD.
I hope you enjoyed the long review and, as usual, let me know how to improve and if you have questions. I have recorded all the
44 measurements for both XLR and RCA outputs (and much much more). And if you want me to publish others or run one of your choice, feel free to ask.
--------
Flo
