836 digital sources from 1992 to 2024 ranked by THD

[audio_tony]

Thanks your for your interest in this data table and congratulations for your effort to put your site together .

However, I see that harmonics and THD which are clearly distinguished by the metrics (dBV or dBr) in the original tables are mixed up on your site and the quoted THD in dBV is directly converted to percentage. To my mind, this is wrong.

For instance, the translation of the quoted -104 dBV THD of the Sony CDP-XA7ES on your site is 0,00063 %, as if the 100% level from which the percentage are calculated is 0 dBV, which is 1 V RMS. But the nominal maximum output level of the Sony CDP-XA7ES when 0 dBFS is played is not 1 V RMS. It is 2 V RMS (about 6 dBV). Actually it is well documented (Stereophile or Audio reviews) that the output level is a bit higher, about 2.5 V RMS (which is about 8 dBV). Thus, the actual THD is not 104 dB under 1 V RMS output level, but 112 dB under 2.5 V RMS, which corresponds to 0.00025 % THD relative to 8 dBV.

Problem is : I was not able to compute harmonics relative to full scale output neither calculate THD in percentage for every source for which the measurements have been showed in dBV by the magazines, because the analogue output level with 0 dBFS digital input of each and every source were scarcely documented.

In conclusion, I recommend that measurements in dBV remain clearly separated from the others and stated in dBV. Otherwise, readers would be mislead.

Also, the nature of the test signal become undocumented on your site.

Perhaps the "Note" column (in which I originally only put casual information about each device) could be replace by a "Test signal" column.

May I suggest a relabeling of the title of each column on your site :

View attachment 323062

Yes, I did have a few concerns about some of the THD figures. I had seemingly thought that I could convert the dB figure into % directly?

I do however need to get them into % because I use that to determine the row colour - so to have a mix of dB and % messes things up somewhat.

I am looking at a way to work the test signal into the data - although it's a bit tricky to match the entries, however your suggestion looks like it might be plausible.

I'll be putting more work into it over the next few days, as there are a few others things I'm not happy with.

 

[Scytales]

Well, there are not that much devices for which test results were showed in dBV in the printed issues. Perhaps it would be better to exclude them if it is inconvenient to show them.

I have just crosschecked a few test reports, and I have to this moment only found two devices for which the output level is known, thus enabling a translation of the THD from dBV to dBr (conservatively calculated) :

1) July 95 issue, Sony CDP-XA7ES (0 dBr = 8.3 dBV output at 0 dBFS) : thus H2 at -107 dBV becomes -115 dBr and H3 at -107 dBV becomes -115 dBr. No H4 and no H5 neither higher order harmonic. THD in percentage ref. 0 dBr : 0,00025 %. Test signal : 1 kHz, 16/44.1 PCM.

2) July 95 issue, Denon DCD-S10 (0 dBr = 7.4 dBV output at 0 dBFS) : thus H2 at -105 dBV becomes -112 dBr, H3 at -110 dBV becomes -117 dBr, H4 at -106 dBV becomes -113 dBr and H5 at -107 dBV becomes -114 dBr. Harmonic(s) greater than H5 : yes. THD in percentage ref. 0 dBr : 0,00042 %. Test signal 1 kHz, 16/44.1 PCM.

That's the best I can do with my documentary resources.

 

[audio_tony]

1) July 95 issue, Sony CDP-XA7ES (0 dBr = 8.3 dBV output at 0 dBFS) : thus H2 at -107 dBV becomes -115 dBr and H3 at -107 dBV becomes -115 dBr. No H4 and no H5 neither higher order harmonic. THD in percentage ref. 0 dBr : 0,00025 %. Test signal : 1 kHz, 16/44.1 PCM.

I've never seen THD expressed with no actual reference in this way.

For example, it might be 0.01% at 0dB and drop to 0.0001% at -3dB - or expressed in dB it would be -80dB and -120dB respectively.

I don't see how (for a given input level) the THD can change when expressed in dBr or dBv or dBFS.

It will always be relative to input level.

I am open to correction on this however.

For absolute accuracy, we would need to specify the reference points for dBr and dBFS, and the relationship between them.

For instance, if the reference for 0 dBr is -107 dBV, then a signal at -107 dBV would be 0 dBr. However, if the reference for 0 dBr is different, you'd need to calculate the difference accordingly.

I suspect it's probably best to remove any ambiguous listings though, as you say.

I do intend to provide a link to the PDF file for absolute clarity anyway, and as I say - there is some work to be done re adding levels and so on.

Maybe I should take the site down for now...

 

[Scytales]

I've never seen THD expressed with no actual reference in this way.

For example, it might be 0.01% at 0dB and drop to 0.0001% at -3dB - or expressed in dB it would be -80dB and -120dB respectively.

I don't see how (for a given input level) the THD can change when expressed in dBr or dBv or dBFS.

It will always be relative to input level.

I am open to correction on this however.

Problem is, for a few devices, the published FFT is graduated in dBV with the fundamental notched out, but the actual output level is not stated (I have to look elsewhere to know them for the above mentioned Sony and Denon). All that is known is that the measures were taken at the output of a device when 0 dBFS digital test signal is put at the input (or read on a test disc).

Actually, I think that it is my table that is misleading about that. The column "THD" for the report of the actual measurement results should have been named something like "Absolute Sum of H2 to H5" instead of "THD". I think it's my mistake to have labelled it "THD". Perhaps I should have discarded this few entries instead of following my zeal to list as many devices as possible.

 

[audio_tony]

I have removed the following entries:

Feb-10​	T+A	MP 1260R	USB in	Burr Brown PCM1795	-123​	-117​	-128​	-129​	yes	-115.5​
Jul-95​	TEAC	VRDS 20		Philips TDA1547	-119​	-118​	-110​	-117​	yes	-108.3​
Sep-08​	Benchmark	DAC 1 Pre	Pre-Proc.	Analog Dev. AD1853	-109​	-123​	-114​	-131​	yes	-107.6​
Jul-95​	Rotel	RHDC 10		Burr Brown PCM63	-113​	-115​	-109​		yes	-106.8​
Sep-08​	Apogee	Rosetta 200	DAC	Analog Dev. AD1852	-106​	-121​	-119​	-129​	yes	-105.6​
Dec-95​	Linear Acoustic	CD 1		Philips TDA1547	-113​	-105​				-104.3​
Apr-11​	Marantz	NA 7004	Network	Cirrus Logic CS4398	-107​	-115​		-108​	yes	-104​
Jul-95​	Sony	CDP-XA7ES		Sony CXD2562	-107​	-107​				-104​
Dec-95​	Marantz	CD 17		Philips TDA1547	-107​	-107​				-104​
May-09​	Cambridge	DacMagic	DAC	Wolfson WM8740	-106​	-109​	-115​	-113​	yes	-103.3​
Jan-12​	Naim	ND5 XS	+ Wifi	Burr Brown PCM1791	-108​	-105​	-116​	-117​	yes	-102.8​
May-11​	Naim	NDX	+ Wifi	Burr Brown PCM1791	-103​	-107​	-116​	-119​		-101.3​
Jan-13​	Pioneer	N-50	Network	AKM AK4480	-106​	-103​	-123​	-115​	yes	-101​
Oct-12​	Naim	Converter NDS Power supply XPS	+ Wifi	Burr Brown PCM1704	-103​	-118​	-107​	-111​	yes	-100.9​
Jan-12​	Linn	Klimax DS/1	Network	Wolfson WM8741	-122​	-102​	-130​	-107​	yes	-100.7​
Jul-95​	Denon	DCD-S10	Digital in	Burr Brown PCM1702	-105​	-110​	-106​	-107​	yes	-100.6​
Apr-11​	Olive	06 HD	Network	Burr Brown PCM1792	-101​	-113​	-111​	-129​	yes	-100.3​
Sep-08​	M-Audio	Fast Track Ultra	DAC		-101​	-105​	-116​	-107​	yes	-98.7​
Dec-08​	Accustic Arts	Tube DAC II SE	Tubes out		-109​	-110​	-99​	-112​	yes	-98.1​
Sep-08​	Terratec	DMX 6Fire USB	Sound c.	Cirrus Logic CS42426	-106​	-99​	-121​	-113​	yes	-98​
May-09​	Lavry	DA 11	DAC	Analog Dev. AD1955	-98​	-117​	-128​	-124​	yes	-97.9​
Sep-95​	CEC	TL 3			-100​	-103​		-107​	yes	-97.6​
Apr-11​	Musical Fidelity	M1 Clic	Network	Burr Brown DSD1796	-110​	-97​	-120​	-124​	yes	-96.7​
Jan-12​	Linn	Klimax DSM	Network	Wolfson WM8741	-117​	-95​	-126​	-105​	yes	-94.5​
Nov-09​	Linn	Majik DS Dynamik	Network	Wolfson WM8740	-114​	-94​	-123​	-106​	yes	-93.6​
Nov-09​	Linn	Akurate DS Dynamik	Network	Wolfson WM8740	-112​	-94​	-122​	-106​	yes	-93.6​
Jan-13​	Denon	DNP-F109	Network	AKM AK4424	-91​	-99​	-116​	-110​	yes	-90.3​
May-09​	Musical Fidelity	V-DAC	DAC	Burr Brown DSD1792	-96​	-92​	-105​	-101​	yes	-90​
Apr-11​	Rotel	RT-09	Network	Wolfson WM8740	-98​	-95​	-93​	-95​	yes	-88.8​
Jan-13​	Yamaha	CD-N500	Network	Burr Brown PCM5121	-91​	-96​	-92​	-114​	yes	-87.7​
Nov-09​	Linn	Klimax DS Dynamik	Network	Wolfson WM8740	-90​	-92​	-109​	-97​	yes	-87.3​
Jan-13​	Rotel	RT-12	Network	Wolfson WM8740	-94​	-88​	-113​	-104​	yes	-86.9​
Apr-11​	Meridian	Sooloos MediaCore 200	Network		-88​	-87​	-118​	-98​	yes	-84.2​
Feb-10​	PS Audio	PerfectWave	USB in	Wolfson WM8742	-81​	-96​	-119​	-107​	yes	-80.8​
May-09​	Advance Acoustic	MDA 503	DAC	Analog Dev. AD1955	-77​	-92​	-102​	-117​	yes	-76.8​
Feb-10​	Ayre	QB 9	USB in	Burr Brown DSD1796	-77​	-84​	-103​	-112​	yes	-76.2​
Dec-07​	PS Audio	DL III	Conv.	Burr Brown PCM1798	-76​	-86​	-106​	-104​	yes	-75.6​
Sep-08​	Aqvox	USB 2 D/A	DAC	Burr Brown PCM1796	-77​	-76​	-104​	-93​	yes	-73.4​
Nov-11​	Ayon	S 3	Network	Burr Brown PCM1792	-72​	-90​	-115​	-116​	yes	-71.9​
Jun-11​	Audio Research	DAC 8	USB in	Burr Brown PCM1792	-79​	-73​	-106​	-88​	yes	-71.9​
Feb-10​	Audio Research	DAC 7	USB in	Burr Brown PCM1792	-88​	-71​	-110​	-102​	yes	-70.9​
Feb-13​	Atoll	ST 100	Network	Burr Brown PCM1796	-40​	-54​	-68​	-81​	yes	-39.8​

 

[Scytales]

I suggest to explain the table with a short text and a brief summary, such as :

"The data are taken and calculated from FFTs obtained with an Audio Precision System One or an Audio Precision System Two at the laboratory run by the German publisher WEKA MEDIA PUBLISHING Gmbh. The measurements have all been published in the German magazines Stereoplay, Audio or Audiophile.
Links to the laboratory : https://www.connect-testlab.com/#testlab "

An explanation of the color scale you chose to use could be added somewhere.

 

[audio_tony]

EDIT: After some consideration, I decided to leave the site as is, and provide a note at the top of the page to explain that replicating the various comments and annotations in the PDF was proving to be somewhat difficult.

It would be extremely difficult and time consuming to add these to the site. (if someone wants to rise to this challenge and is familiar with PHP/MySQL then I'll be happy to host it...)

Therefore I have recommended that for further reading, people should download the PDF file (I have provided a download link).

I am working on some ideas for the table formatting.

One problem is that there are too many columns (with some too wide) to fit all screens properly, so I'm looking at alternative ways to present the data.

My current thinking is to simply provide a hyperlink (or possibly 'speech bubble') that can describe the test parameters (+ other descriptions that need to be fitted in).

Said speech bubble can either be activate by a mouse click or rollover type action. I can't think of any other better ways at this stage.

I'm not a fully fledged web developer, this is just a sideline hobby carried over from when I used to work in IT (3rd line enterprise server support).

The project has suffered some significant 'creep' since conception lol.

 

[Scytales]

I update the files this January 15th, 2024 with 5 new tests (see first post) :

Test in 24 bits :
Denon DVD-5000

Tests in 16 bits :
Arcam Alpha 8 SE (2nd test of another sample)
Marantz CD 7
Musical Fidelity X-Ray
Wadia 850i

I have also corrected some typos and given some more informations (D/A converter chips identification, technical notes) whenever I can.

The new files have been uploaded in the first post.

Thanks for reading.

 

[Scytales]

I update the files this February 21th, 2024 with 17 new tests (see first post) :

Test in 24 bits :

Linn Akurate DS
SPL Diamond
TEAC UD-505-X

Tests in 16 bits :

Accuphase DP-77
Audiomeca Keops
Audionet ART V2
Ayre D-1
Cambridge Audio AXC35
Jadis JS1 MkIII
Kondo M100 DAC
Krell SACD Standard
Meridian 800+861
Micromega Minium
Reavon UBR-X200
Rotel CD11 Tribute
Sony SCD-555ES (in addition to a test in DSD already mentionned)
Yamaha CD-S303

I have corrected a few errors: Eternal Arts Dp DAC MkII and McIntosh MCD85 AC were tested in 24 bits and not 16, and Lumin T2 and NAD C658 were tested at 1 kHz and not 330 Hz. I have also corrected some typos and given some informations about some peculiar digital signal processing whenever I can.

The new files have been uploaded in the first post.

Thanks for reading.

 

[Scytales]

I update the files this March 2nd, 2024 with 4 new tests (see first post) from old issues.

Tests in 16 bits :

Accuphase DC-330
Accuphase DP-75V
dCS Verdi+Elgar plus
Sony DTC-ZA5ES

More importantly, I have finally found in an old issue a statement about the nominal level at which the test are conducted : 2 V RMS.

I have also identified some DAC chips or signal processing peculiarities previously undocumented and simplified the color grading for clarity.

The files have been uploaded in the first post.

Thanks for reading.

 

[Scytales]

03/21/2024 update : correction of an error about the Denon UDP 2012UD (2nd harmonic at -99 dB instead of -89 dB, hence better THD number) and news informations about some digital signal processes.

 

[Scytales]

03/22/2024 update : new informations about some signal processes.

 

[Scytales]

03/24/2024 update : new informations about some signal processes.

 

[Scytales]

I update the files this May 26th, 2024, with 8 new tested devices (see first post).

Tests in 16 bits :

Eve Systems Play
Pier Audio CD-880 SE

Tests in 24 bits :

Cambridge CXN V2
Cocktail Audio X-45
MacIntosh D1100
NAD CS1
Pro-Ject Pre Box DS2 Digital
Technics SL-G700M2

I have also identified some signal processing peculiarities previously undocumented.

I removed the few mobile DAPs from the list, because it make little sense to mix them with fixed installation devices and because they are measured at to much different signal condition.

The files have been uploaded in the first post.

Thanks for reading.

 

[Scytales]

I have updated version 16 to 16a.

After having read detailed measurement on the Miller Audio Research website and reexamined service manuals and DAC datasheet, I find I have mistakenly stated that Marantz DV9500 and its clone Lexicon RT20 and DV9600 systematically convert DSD to PCM whereas it is only true when bass management is engaged. This error was corrected.

Some new details about signal processing added for some devices.

Having also realized that some devices do not bear the same designation in Europe than in the America, I also mentioned some cross-references between some European and US/Canadian devices.

 

[Scytales]

I update the files this June 23rd, 2024, with 2 new tested devices (see first post).

Tests in 16 bits :

Pioneer UDP-LX800

Tests in 24 bits :

RME Fireface UCX

I have also identified some signal processing peculiarities previously undocumented.

The files have been uploaded in the first post.

Thanks for reading.

 

[Scytales]

Error in the THD ranking (2x Onkyo DV-SP502 in place of missing Sharp DX-SX1) has been corrected.

 

[Scytales]

I update the files this August 3rd, 2024, with 16 new tested devices and 2 more measurements from already tested devices (see first post).

Tests in 16 bits :

Accuphase DC-300
Audio Analogue Paganini
Berendsen CDP 1
Burmester 970
Copland CDA 266
Creek CD43 MK2
Denon DCD-100
Einstein 'The Last Record Player'
Ensemble Dichrono DAC/PRE
Exposure XM CD-Player
Luxman D-380
Marantz SA-1 (already tested in SA-CD mode)
Phonosophie Impuls 2
Rega Apollo CDP
Wadia 27i (2nd sample)

Tests in 24 bits :

Audiolab 7000N Play
Mark Levinson N° 360S (Amirm's former DAC )
NuPrime DAC-9


The files have been uploaded in the first post.

Thanks for reading.

 

[NTTY]

Bonjour Scytales,

This is an amazing and impressive exercise! Thanks for posting and not keeping it only for yourself.

I see I own three of the top ten list, so I’ll give them a second chance to measure as I did not recall they were doing so great, especially the Denon DCD-S10.

I ordered the Denon SACD disk you mentioned and in use here for many SACD players. I’ll give it a try.

I was using a Denon test CDA in the past but since it was not dithered, and because I was able to burn a test CD of the same quality, I stopped using it. I never directly used SACD test tones, I should have.

When it comes to establishing a ranking solely based on THD at 0dBFS, there’s a risk to miss the main difference between ancient R2R architectures and newer 1bit. Indeed, if R2R was competing at full scale, most of them suffered from -12dBFS, and that made a significant difference. Some high end R2R chips of the time offered an optional trim of the MSB (sometimes 4 MSB) which improved their THD levels at full scale, allowing them to shine on that particular measurement.

Also, I agree that most of the time the first 5 harmonics make the THD. But there are some players making this challenging such as the Teac VRDS 25x which shows its highest distorsion at the 4th, 6th, 8th and 10th harmonics (nearly equal levels). I understood you captured more than five harmonics for many players, and that’s when you have the “yes” in column Hn, correct?

I’ll pull out couple of players from their retirement storage when I receive the Denon SACD. Both the DCD-S10, DCD-SA1 have digital input too, so that’s also worth a test for the sake of comparison.

Anyways, impressive compilation of data

Merci encore!

 

[Scytales]

Bonjour Scytales,

This is an amazing and impressive exercise! Thanks for posting and not keeping it only for yourself.

I see I own three of the top ten list, so I’ll give them a second chance to measure as I did not recall they were doing so great, especially the Denon DCD-S10.

I ordered the Denon SACD disk you mentioned and in use here for many SACD players. I’ll give it a try.

I was using a Denon test CDA in the past but since it was not dithered, and because I was able to burn a test CD of the same quality, I stopped using it. I never directly used SACD test tones, I should have.

When it comes to establishing a ranking solely based on THD at 0dBFS, there’s a risk to miss the main difference between ancient R2R architectures and newer 1bit. Indeed, if R2R was competing at full scale, most of them suffered from -12dBFS, and that made a significant difference. Some high end R2R chips of the time offered an optional trim of the MSB (sometimes 4 MSB) which improved their THD levels at full scale, allowing them to shine on that particular measure.

Also, I agree that most of the time the first 5 harmonics make the THD. But there are some players making this challenging such as the Teac VRDS 25x which shows its highest distorsion at the 4th, 6th, 8th and 10th harmonics (nearly equal levels). I understood you captured more than five harmonics for many players, and that’s when you have the “yes” in column Hn, correct?

I’ll pull out couple of players from their retirement storage when I receive the Denon SACD. Both the DCD-S10, DCD-SA1 have digital input too, so that’s also worth a test for the sake of comparison.

Anyways, impressive compilation of data

Merci encore!

Hi!

Thank you for your kind words.

I think the Denon Audio Check SA-CD has sadly seldom use to make measurements of electronic because it only gets 100 Hz, 1 kHz, 10 kHz, 15 kHz and 30 kHz sinus @ -16 dB SA-CD (SA-CD layer) and 100 Hz, 1 kHz, 10 kHz, 15 kHz @ -16 dBFS (CD layer). GXAlan has analysed the signal on this disc and found there are not the purest. I think the most probable disc the German labs used to use is the Philips Super Audio CD DAC Test Disc (ref. 3122-783-00632). I have only ever spotted one on the used market in 20 years of Internet surfing. The only human I know of who owns that disc is here, on ASR: PMA.

Thank you for your insights about R2R type converters. I fully concur with them.

Concerning the harmonics above the 5th rank, your are right : as explained in the introductory text of each table, the “yes” in column "Hn" means that at least one harmonic above rank 5th is visible on the FFT. I reckon that it is a limited information, but I couldn't figure out a way of doing better. I once tried to compute THD up to the 10th harmonics, but the workload was unbearable. I then got the idea to give the number of visible harmonics above the 5th rank, but this added information seems not worth the effort, for without more details (level, exact order of the extra harmonics), it has little value. I had to accept that it was not possible to summarise valuable information on a table the kind I imagined at the time better that I have done. My intention was to ease the comparison of great many samples of digital devices, which was not an easy task to do from measurements whose results were represented graphically. A more thorough evaluation of each devices through their distortion residual can only be done effectively by looking at FFTs graphs, but comparing more that two graphs at a time is compelling and using graphs to rank hundreds of devices is impossible. Hence my idea of making a table by translating graphical information into few numbers. If anyone get a novel idea of how to overcome that issue, he will be very welcome !

The way I see things is that the devices that produce harmonics at most up to the 5th rank probably deserve more attention than the others, providing the THD remains acceptable. Of course, that does not means that a device don't produce other kinds of artefacts (noise modulation, intermodulation, switching artefacts,...). But generally speaking, the devices with the cleanest distortion spectrum and lowest THD are also the devices with the cleanest FFT overall.