• 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!

Sony CDP-337ESD Review (CD Player)

NTTY

Addicted to Fun and Learning
Forum Donor
Joined
Apr 27, 2018
Messages
537
Likes
2,115
Location
Switzerland
This is a review and detailed measurements of the Sony CDP-337ESD CD stereo player and transport.

SonyCDP-337ESD-000.jpg


I said multiple times that I am into old CD Players, and this is one more.

When I was much younger, I was absolutely fond of Sony players but they were so expensive! I love the look and especially the pad on the right, that looked futuristic at the time.

I decided to go for this one because it hosts two DACs from... Philips, the famous TDA1541A. And so, a Sony with Philips chips while Sony had theirs, I could not resist to give it a try.


Sony CDP-337ESD - Presentation

Released in 1988, it was a member of a big family, nearly at the top of it, just behind the 557ESD (US designation was the 707ESD), and above the 227ESD which has reviewed by @audio_tony.

With 25lbs (11.5kg) the Sony CDP-337ESD plays in the heavyweight category.

The elements of interest are:
  • Philips conversion: Like I said the conversion is performed by the Philips TDA1541 chip, in its improved "A" version which accepts a faster sampling rate and therefore allows the use of 8x oversampling, instead of 4x previously. It's a stereo chip and the Sony 337ESD has two of them, and they work in parallel mode.
  • 8x oversampling: This was kind of new at the time and here it is a Sony chip (CXD1144) which takes care of the oversampling. Yamaha and NPC had released theirs too, so it was a bit of a battle of components at the time. The performances given by Sony actually showed good technical progress. We'll measure that.
  • Staggered conversion: While the two TDA1541A DACs are not crowned versions (S1 or S2), the front of the Sony sports a proud "STAGGERED 4 D/A CONVERTER SYSTEM". In reality, the two stereo chips convert a “staggered” digital signal, coming from the 8x oversampling filter (see detailed explanation below). From an engineering point of view, it's very elegant, and it's mostly identical to the monstrous Sony DAS-R1 DAC.
  • KSS-190 mech: This mech is all metal (BU-10 optical unit). The construction, the silence of operation (including when opening/closing the drawer) and the incredible speed of execution (never seen for me, even on the Philips CD-Pro) contribute to the pleasure of use.
  • Gibraltar Chassis: The KSS-190 is suspended and rests on a chassis made of calcium carbonate (major component of marble and other limestones) in resin reinforced with fiberglass, which Sony called "Gibraltar". It itself rests on a double metal bottom which contributes greatly to the weight of the CD Player.
The back of the player is relatively classic with fixed and variable outputs together with Tolink and Coax digital outputs:

SonyCDP-337ESD-017.jpg


Note the selector which forces to chose RCA or Digital outputs, that is a bit weird.

Here are the specifications (from the user manual):
  • Frequency characteristics: 2 Hz to 20 kHz ± 0.5dB
  • Total harmonic distortion factor : Less than 0.0025%
  • Dynamic range : More than 98dB
  • Signal to Noise ratio : More than 110dB
  • Channel Separation : More than 100dB
  • Output level : 2 Vrms (Fixed)
  • Output level : 0 Vrms - 2 Vrms (Variable)
  • Headphone output level : 28 mw (32 Ω)
  • Power consumption : 22W
  • External dimensions : Width 430mm, Height 125mm, Depth 375 mm
  • Weight : 11.5kg
  • Remote Wireless Remote Control : RM-D650

Staggered conversion

Allow me to spend some time talking about this because, as opposed to what some said, it's not just marketing BS printed on the front face of this beautiful CP Player:

SonyCDP-337ESD-001.jpg


The staggered conversion is described on page 164 of the Digital Audio and Compact Disc Technology, when talking about a newly developped then Sony 18bits oversampling running at 8x the base frequency, and in these terms:

In order to cope with 18 bit and such a high conversion rate great care must be taken in the designing of the D/A converter. To reduce the load imposed onthe D/A converter Sony developed an 'Overlapped Staggered D/A Conversion System' (Figure 12.15). The basic idea is to use a digital filter circuit at8 Fs output, combined with two D/A converters for each channel.
SonyCDP-337ESD-022.jpg


The datasheet of the filter CDX1144 tells us more about the concept, and is all summarized by the below drawing:

SonyCDP-337ESD-020.jpg


In fact, this filter offers several outputs. One of them, called I²S mode, outputs data of the left and right channels in a multiplex mode. That is, each channel contains data for left and right channels, but they are staggered as you can see above: L1 is output from the left channel (DATAL) and L2 is output from right channel (DATAR) one sequence later.

In this I²S mode, each channel contains data output from the oversampling filter and so each runs at 4x output, instead of having an output at 8x speed. What's the advantage of that? Simply to benefit from an 8x oversampling using DACs that runs at 4x only. The only drawback being that you need two DACs instead of one, of course.

Fact is the first iteration of the Philips TDA1541 was running at 4x only, while the subsequent 1541A version was compatible with 8x speed.

The service manual of the Sony CDP-337ESD shows the implementation as per Philips and Sony recommendations for their respective chips:

SonyCDP-337ESD-018.jpg


You can see that the Pin4 of the TDA1541A is unused, only the Pin3 receives data. A particular setup of the TDA1541A allows to de-mux the input and feeds the two internal DACs with left and right data. The output of respective left and right channels are paralleled.

The trick is that, because each TDA runs at half the speed of the oversampling filter, there will be a 50% overlap in time of their outputs (they are paralleled). Let me try to explain:
A DAC like the TDA1541A has only one function. It is to convert a digital word into a voltage, and it will hold that voltage until the next word arrives. Since the data is staggered at the output of the Sony digital filter and sent to two different Philips TDA, it means one will be still holding the previous word when the other one receives the subsequent word. In the time domain, that means two different digital words will overlap by 50%. This is why Sony was talking about an “Overlapped Staggered D/A” in their documentation.

It's not easy to think about what the output looks like with that technique, But after my investigations, I discovered that @bramjacobse had a nice representation of it on his website, where he talks and tests the staggered mode of the CXD1144 (drawing probably from some brochure of the time):

SonyCDP-337ESD-021.jpg

As you can see, the outputs overlap by half a cycle rendering the signal smoother. You see the digital words at the bottom of the drawing and how they overlap in the respective DACs.

In addition, running oversampling at 8x presented the below advantages (from Sony Digital Audio and Compact Disc Technology) which we are familiar with (see Analog Devices Oversampling Interpolating DACs):

Since the outputs of the two DACs are added continuously a maximum improvement of 3dB is realised in quantization noise with reduced distortion. The output current is also doubled, improving the S/N ratio of the analog noise by a maximum of 6dB.

So no, "Staggered 4 D/A Converter" was not just another marketing name :)


Let's have a quick look inside before moving to measurements:

SonyCDP-337ESD-003.jpg


Nothing fancy, but everything is serious. The two Philips TDA:

SonyCDP-337ESD-005.jpg


As opposed to BurrBrown, there were no available external trimming. Also, they were not benefitting from laser triming to improve linearity. Rather, Phillips had a process to stamp the chips per their tested linearity after production per the below:

Selections have been made with respect to the maximum differential linearity error (EdL):
  • TDA1541A/N2
    • bit 1-16 EdL < 1 LSB
  • TDA1541A/N2/R1
    • bit 1-16 EdL < 2 LSB
  • TDA1541A/N2/S1
    • bit 1-7 EdL < 0.5 LSB
    • bit 8-15 EdL < 1 LSB
    • bit 16 EdL < 0.75 LSB

The TDA1541A of the Sony CDP-337ESD are of category N2.

This player is really an absolute delight to use, completely silent and that mech is soooooo fast, I love it!


Sony CDP-337ESD - Measurements (Fixed Analog out)

From now on, I will be consistent with my measurements as I described them in the post "More than we hear". It will help comparing the devices I reviewed. All measurements were performed using the latest Test CD.

The Sony CDP-337ESD outputs a little more (0.4dB) than the standard 2Vrsm from its RCA outputs. There was a slight channel imbalance of 0.13dB. The RCA outputs are noninverting.

Here you go with the standard 999.91Hz sine @0dBFS (without dither) from the test CD:

SonyCDP337ESD_999.91Hz_0dBFS_LR.jpg


Both channels are represented but only one gets evaluated in that window. Right channel does a little better (by 2dB).
THD is at -101dB and so will clear CD Audio content. SINAD (94.1dB) and ENOB could have been better but 2 or 3dB, maybe that's what Sony left to the big brother ESD507.

As I mentioned so many times, the old R2R architectures always suffered when lowering the digital scale. As seen below, THD is at -96dB with a -6dBFS sine (without dither):

SonyCDP337ESD_999.91Hz_-6dBFS_LR.jpg


The "grass" you can see is really typical of R2R converters and especially the Philips TDA1541A. But if you go check the Denon DCD-3560 which used BurrBrown DACs or the much more recent Topping Centaurus, it's the same distorsion profile.

As I did with the big Denon DCD-3560, let's try to push the staggered TDA conversion to its limits with a shaped noise signal. This type of noise shaping technique allows to lower the noise floor in an area of interest, where our ears are most sensitive. In a bit depth limited format as we have here (16bits only), it allows us to go above and beyond the 16bits, which is a very nice software/mathematic trick.

So, here we go with the same 999.91Hz test tone @0dBFS with shaped dither:

SonyCDP337ESD_999.91Hz_-6dBFS_Shape.jpg


I did not changed the scale of the view so you can see the noise floor actually really going down. It is limited, yes, but that means this conversion is clean enough to reach a "real" SINAD of 100dB (where it matters, here calculated from 20Hz to 6kHz), which is 17bits resolution. And yes, that is with an audio CD and a WAV file at 16bits/44.1kHz, thanks to the trick of noise shaping which is largely used in modern CD Audio Masters.

As you can see above, the noise shaping reduces noise floor up to 6kHz after which the noise is rejected in a less sensitive zone or our ears, and that's what creates the two waves you see, from 6kHz to 20kHz.

One interesting thing is that we get a much cleaner trace here than with the Philips TDA1541A S1 (crown version) as implemented in the Revox B 226-S. That is a big thanks to 8x oversampling of Sony, as opposed to the much weaker 4x processing of Philips.

You probably noticed, this is a very quiet CD Player (999.91Hz @0dBFS, left and right channels):

SonyCDP337ESD_PS_LR.jpg


There are no visible power supply related components, this is very good, what we always want to see.

Now, let's have a look at the bandwidth (from periodic white noise analysis):

SonyCDP337ESD_Bandwidth.jpg


We see a gentle roll-off at both ends and we are in the specs of Sony (-0.2dB).

Let's have a look at the 8x oversampling filter, which was the new offer of Sony at the time (Overlay of White Noise and 18k+20Hz dual tones):

SonyCDP337ESD_Filter.jpg


The attenuation is 90dB and was fully active as soon as 23.5kHz, not bad at all!

Do you want to know how it compares with the standard Philips 4x oversampling filter (SAA7220)? Here below is comparison between the two:

SonyCDP337ESD_Filter02.jpg


Yes, you can see a massive improvement from the Sony filter which attenuates much more and is way faster. And I'm sure @AnalogSteph will like the below view, in Linear Frequency Scale, that shows the Sony filter prevents the visible ringing in audio band that the SAA7220 was generating:

SonyCDP337ESD_Filter03.jpg


Again, massive improvement even if we see the low decrease by 0.2dB at 20kHz.


Multitone (1/10 decade) shows a happy CD player, not having much issue to clear 16bits of data:

SonyCDP337ESD_MT_LR.jpg


Left channel is a bit weaker, generating more noise and distortion, but nothing of a major concern.

Jitter shows again a little room for improvement for the big brother:

SonyCDP337ESD_JTest.jpg


The red trace is what is recorded on the test CD (From the digital outputs), it can't be better. The blue trace is from the RCA outputs. That said the side bands here are so low that it is impossible to hear.

Let’s have a look at an undithered 997Hz sine at -90.31dBFS. With 16bits data, the signal should appear (on a scope) as the 3DC levels of the sign magnitude smallest digital signal:

Sony CDP-337ESD_3DC.jpg


This is again a very good result with the TDA1541A as we can very clearly see the 3DC level correctly represented and symmetrical. That is again a big thanks to the 8x oversampling of Sony.

On your request and support (more information here), I am adding an "intersample-overs" test. It intends to identify if the oversampling filter has sufficient headroom to process near clipping signals. Indeed, and because of the oversampling, there might be interpolated data that go above 0dBFS (see the values in the table) and would saturate (clip) the interpolator and therefore the output. This effect is highlighted with the measurements below, and revealed through THD+N measurements up to 96kHz:

Intersample-overs tests
Bandwidth of the THD+N measurements is 20Hz - 96kHz
5512.5 Hz sine,
Peak = +0.69dBFS
7350 Hz sine,
Peak = +1.25dBFS
11025 Hz sine,
Peak = +3.0dBFS
Teac VRDS-20-30.7dB-26.6dB-17.6dB
Yamaha CD-1-84.6dB-84.9dB-78.1dB
Denon DCD-900NE-34.2dB-27.1dB-19.1dB
Denon DCD-SA1-33.6dB-27.6dB-18.3dB
Onkyo C-733-88.3dB-40.4dB-21.2dB
Denon DCD-3560-30.2dB-24.7dB-17.4dB
SMSL PS200 (16bits/44.1kHz)-92.8dB-35.5dB-21.0dB
Sony CDP-337ESD-30dB-24.3dB-16.2dB

I left several other references for you to compare with. The oversampling filter of the Sony CDP-337ESD does not have headroom to process Inter-sample overs. The Yamaha CD-1 shines in that test because it does not use an oversampling filter.

Other measurements (not shown):
  • IMD AES-17 DFD "Analog" (18kHz & 20kHz 1:1) : -96dB
  • IMD AES-17 DFD "Digital" (17'987Hz & 19'997Hz 1:1) : -90.2dB
  • IMD AES-17 MD (41Hz & 7993Hz 4:1): -95.7dB
  • IMD CCIF (19kHz & 20kHz 1:1) : -90.2dB
  • IMD DIN (250Hz & 8kHz 4:1) : -87.0dB
  • IMD TDFD (13'58Hz & 19841Hz 1:1) : -118.8dB
  • IMD TDFD Bass (41Hz & 89Hz 1:1) : -130.8dB
  • IMD SMPTE (60Hz & 7kHz 1:4) : -98.5dB
  • Dynamic Range : 96.9dB (1kHz @-60dBFS no dither)
  • Crosstalk: -140dB (1kHz), -115dB (10kHz)
  • Pitch Error : 19'996.39Hz (19'997Hz requested) ie -0.003%
These are very good measurements, again for a Philips TDA. Special note for the cross talk, which is extremally low for the time.

Last but not least, I like to have a look a the THD vs Frequency when using a -12dBFS signal. This has proven to me to be a key differentiator, especially when I'm reviewing an old CD Player using R2R conversion. Here are the results with the Sony (Left and Right channels shown):

SonyCDP337ESD_THDvsFreq_LR.jpg


This is not so bad, but BurrBrown was doing better at the time. I overalaid the Onkyo C-733 trace for reference, 1 bit conversion nails this test, and that's why I like to use it.

Now, what about a little battle with a TDA1541A in its S1 (crowned) version? Here you go compared to a Revox B 226-S (green traces):

SonyCDP337ESD_THDvsFreq_vsTDA1541AS1.jpg


Yeah, Sony managed to push the TDA1541A beyond a single crowned version. Well done Sony, and thanks again to the 8x oversampling and staggered mode.

What about a single TDA1541A and about the promises of Sony in terms of noise and distortion reduction? Here you compare to a Revox B 126:

SonyCDP337ESD_THDvsFreq_vsTDA1541A.jpg


I left the plot at 1khz and you can in the legend that the THD is 6dB lower. So when Sony said there was a benefit of "up to 3dB" in reduction of quantization noise, well they might have been conservative.


Sony CDP-337ESD - Measurements (Digital out)

The reasonable -30ppm clock deviation, which I measured previously, generates some windowing errors when I perform FFT of the digital output. Nothing that would impact the PLL of a DAC, though.

So I had to take the below without averaging, and this is the standard 999.91Hz @0dBFS (without dither), which is perfect:

SonyCDP337ESD_999.91Hz_0dBFS_Opti.jpg


Distorsion and noise are the ones of the WAV file that was used to create the Test CD. It won't be better.

When feeding an external DAC, like the SMSL PS200, I get the below (same 999.91Hz @0dBFS, this time with 32 averages, since no windowing errors as the PLL of the DAC stabilizes the input):

1734954236544.png


We find the -0.003% clock deviation again (-30ppm) which demonstrates that the PLL of the DAC has no issue with the incoming digital stream.
We are extremely close to what's on the test CD, by the way. So in case you think the above is better than the internal DAC of the Sony CDP-337ESD, then you know what to do.


Sony CDP-337ESD - Testing the drive

What would be good measurements if the drive would not properly read a slightly scratched CD, or one that was created at the limits of the norm? The below tests reply to these questions.

Test typeTechnical testResults
Variation of linear cutting velocityFrom 1.20m/s to 1.40m/sPass
Variation of track pitchFrom 1.5µm to 1.7µmPass
Combined variations of track pitch and velocityFrom 1.20m/s & 1.5µm to 1.40m/s & 1.7µmPass
HF detection (asymmetry pitch/flat ratio)Variation from 2% to 18%Pass
Dropouts resistanceFrom 0.05mm (0.038ms) to 4mm (3.080ms)Up to 0.3mm without interpolation, up to 1.5mm with interpolation.
Combined dropouts and smallest pitchFrom 1.5µm & 1mm to 1.5µm & 2.4mmUp to 1.5mm with interpolation.
Successive dropoutsFrom 2x0.1mm to 2x3mmUp to 0.2mm without interpolation, up to 1.5mm with interpolation.

The drive was able to consistently continue playing, without generating typical digital clicks, with dropouts of up to 1.5mm. The interpolation effect remained hidden to my ears when testing although I could clearly see its effect in the live FFT measurements. Results don't show this, but as opposed to other drives, this one will never give up even with large dropouts. In this case, interpolation will be heavy and clicks can be heard up to 4mm dropouts, but the player will not stop reading as opposed to many others.
The Sony had no issue with variable linear velocity and/or track pitch, as well as with HF detection.


Conclusion

As always with Sony, we get the best of the best. They took a couple of Philips DACs and pushed them were they never went, and thanks to the oversampling filter CXD1144. Sony released a subsequent oversampling filter containing an additional noise shaping technique. I'd like to test one.

Anyways, there is a reason why Sony Players are so praised, and you have a nice demonstration of their "savoir faire" here.

Needless to say that this player is a delight to my ears, as so many others, but this one looks so cool with a nice inside tech, and is such a pleasure to use that it makes me thrill every time I'm about to spin a disc. Does it sound better? No, but it's definitively a great pleasure to listen to it.

I hope you enjoyed this review, and I wish you a happy end of the year!
 
Last edited:
"It's not easy to think about what the output looks like with that technique, But after my investigations, I discovered that @bramjacobse had a nice representation of it on his websitewhere he talks and tests the staggered mode of the CXD1144"

I also measured CXD 1144 vs CXD 1244 in staggered mode :)

 
Last edited:
Hi Bram, hey nice, I see you added on your site some additional test from the Test CD with rectangular dither and shaped noise. It's cool!
Noise shaping, as well as dither, are more efficient below full scale to reduce the quantization errors, as they grow relative to the small amplitude of the signal.

At -60dBFS this is the difference without dither (green trace) and with noise shaping (red):

1734857663844.png


There's a lot less noise, but harmonic distortion itself did not change much (we see the plot on H3 shows a reduction of 6db but that's basically it).

The CXD1244 adds a noise shaper to the CXD1144 and that's why your tests are really interesting, allowing a comparison between the two with the exact same following D/A conversion.

Cheers
 
Make sure you replace the Elna Duorex caps - they are notorous leakers!

PS. I have 557ESD, which used to be their TOTL model (it has been sold as 707ESD in the US and Japan). The whole range of those players makes terrific CD transports. The CD drive is almost undestructible - when they stop reading CDs, it is ussually a dried up cap in the servo or just a dirty pickup.
 
"The CXD1244 adds a noise shaper to the CXD1144 and that's why your tests are really interesting, allowing a comparison between the two with the exact same following D/A conversion."

CD 160 with CXD1244 and Staggerd TDA1541(a) with Grundig Dem and C0G SMD caps

IMG_2959.jpeg


The CXD1244 low level

FFT Spectrum _ -60 dB Shaped.png


FFT Spectrum _ -110 dB Shaped vs Rectangle.PNG
 
-110dBFS nailed with 16bits, that's what is called linearity :cool:
Remind me, is it 1bit D/A that saves us from non-linearity, or noise shaping :p

@bramjacobse : you have a PM ;)
 
The whole range of those players makes terrific CD transports. The CD drive is almost undestructible - when they stop reading CDs, it is ussually a dried up cap in the servo or just a dirty pickup.
I bought a CDP-338ESD new in the early 1990s and used it for many years until the transport got dodgy and tended to skip on Track 1, then play the rest of the disc with no problem. Cleaning the pickup didn't help. Fine tuning the adjustment pots on the servo board didn't help.

Maybe I should pull it out of the shed and check those capacitors!
 
Yep, do that, it should solve the issue.
 
I bought a CDP-338ESD new in the early 1990s and used it for many years until the transport got dodgy and tended to skip on Track 1, then play the rest of the disc with no problem. Cleaning the pickup didn't help. Fine tuning the adjustment pots on the servo board didn't help.

Maybe I should pull it out of the shed and check those capacitors!

Most tech don't know / care how to properly clean a pickup. You need a complete pickup disassembly.

I have a friend who specialises in these old Sony players here in Poland - he says that 95% of those ES models that are being sent over for not reading CDs have lasers in perfect shape (and all pots completely out of whack, "tuned" by techs that have no idea what thay are doing).
 
Most tech don't know / care how to properly clean a pickup. You need a complete pickup disassembly.

I have a friend who specialises in these old Sony players here in Poland - he says that 95% of those ES models that are being sent over for not reading CDs have lasers in perfect shape (and all pots completely out of whack, "tuned" by techs that have no idea what thay are doing).
Considering that I did it all myself, I think you are right! :cool:
 
I have Sony CDP-X707ES and Yamaha CDX-993, both bought about ten years ago. They work fine and I don't know if they have been "serviced"
 
Most tech don't know / care how to properly clean a pickup. You need a complete pickup disassembly.

I have a friend who specialises in these old Sony players here in Poland - he says that 95% of those ES models that are being sent over for not reading CDs have lasers in perfect shape (and all pots completely out of whack, "tuned" by techs that have no idea what thay are doing).
Skills is the most valuable thing a customer should care for when in search for a repair technician.

May we all benefit for the name of this man or a link towards a website of him ?
 
The staggered conversion is described on page 164 of the Digital Audio and Compact Disc Technology[...]
Thanks for that source, a nice reading, especially due to the nostalgia at the beginning. Guess I still need such a digital processor combined with some slouchy VHS recorder as a proof of concept.

Strictly speaking off topic, but since I came across that in the book, I'm successfully confused when it comes to the author's explanation of dither:

Screenshot_20250104-142402_Samsung Notes.png

My understanding and also what other sources claim, was that without dither, the distortion and introduced side tones are in fact very related to the input signal which is part of the problem and usually undesired.

On the other hand, dither doesn't mask the issue but - in the best case - entirely removes it. At least, he put the 'mask' in quotation marks in the first sentence but not so in the other, plus, for my taste, he could have pointed out the brilliance of dithering in general better and more elaboratly. One doesn't want to miss it out, virtually (n)ever, including the video domain, where it is ludicrously omitted more than often, introducing ugly banding artifacts.
 
Back
Top Bottom