Orpheus Zero Review (CD Player)

[Multicore]

Rack mount and top load. Interesting combination.

 

[Berwhale]

Nice measurements, but the Industrial design is very strange. It's as if they were halfway through designing a rack mount chassis and then someone pointed out it was a top loader...

 

[NTTY]

Thanks for the review. Can you please keep that £400 or so Denon machine to hand for future testing should your techniques evolve?

I’ll keep it

 

[Scytales]

Okay, but the modern gear that @amirm does normally test (with often 120 dB + SINAD!)……
Amir’s measurements without any A-weighting or so are simply more strictly. And the weighting suggestions of the DIN guys or the AES guys may change from decade to decade. So let’s keep the unweighted measurements because these are the RAW stuff. Weightend - let’s see it as an addition!

I think it's a bit more complicated than that.

The CCIR 468 standard mandates a weighting curve that, contrary to A-weighting, actually provides a gain of up to 12+ dB in a bandwidth which have been found to be critical for the subjective annoyance of noise:

And the standard also mandates the use of a special detector circuit (the electronic circuit which convert the instantaneous noise voltage to average values to be used for computing the displayed noise voltage) which has characteristics designed to give a certain weight to noise spikes inside very high crest factor average signal contents.

As a result, if "simple" A-weighting applied to RMS noise level measurements usually produces more impressive figures (say a 95 dB unweighted signal to noise ratio over a 20 Hz - 20 kHz bandwidth becomes 101 dBA), CCIR 468 actually produces less high figures (say the same 95 dB becomes something as 87 dBqps, dBqps being the metrics used to show CCIR weighted noise measurements).

For specific purposes, the CCIR 468 standard also specified an unweighted measurement, using the same special detector circuit, which has a flat frequency response from 22.4 Hz to 22.4 kHz.

This standard has not been amended since several decades : it is a very well-established method of noise measurement.

I have at my disposal a Rohde & Schwarz psophometer which incorporates both CCIR compliant circuits and flat band-limited RMS noise measurement circuits. It is very enlightening to measure and compare the noise levels of various devices obtained with both methods.

By the way, I think the CCIR noise measurement standard shows in itself why noise-shaped dither can be interesting to improve the subjective performances of a digital chain in certain conditions.

 

[Multicore]

The talk about the Swiss Maid got me thinking how much credit Phillips should get here. And then I realized that I don't understand the architecture of these things at all. OP mentions a "large servo control card". But what does it do? Control the position of the laser and diode? Control the main drive motor? Or does Phillips handle that? How does control of those mechs relate to system performance? Is there a re-clocking and error correction buffer like on a diskman? I have no idea.

 

[Rja4000]

The CCIR 468 standard mandates a weighting curve that, contrary to A-weighting, actually provides a gain of up to 12+ dB

Amir is using CCIR-2k for his Dynamic range measurements. It's the same, but zeroed at 2kHz instead of 1kHz. That means a "gain" approx. 5.6dB lower compared to CCIR-1k.

 

[totti1965]

I think it's a bit more complicated than that.

The CCIR 468 standard mandates a weighting curve that, contrary to A-weighting, actually provides a gain of up to 12+ dB in a bandwidth which have been found to be critical for the subjective annoyance of noise:

View attachment 396712
And the standard also mandates the use of a special detector circuit (the electronic circuit which convert the instantaneous noise voltage to average values to be used for computing the displayed noise voltage) which has characteristics designed to give a certain weight to noise spikes inside very high crest factor average signal contents.

As a result, if "simple" A-weighting applied to RMS noise level measurements usually produces more impressive figures (say a 95 dB unweighted signal to noise ratio over a 20 Hz - 20 kHz bandwidth becomes 101 dBA), CCIR 468 actually produces less high figures (say the same 95 dB becomes something as 87 dBqps, dBqps being the metrics used to show CCIR weighted noise measurements).

For specific purposes, the CCIR 468 standard also specified an unweighted measurement, using the same special detector circuit, which has a flat frequency response from 22.4 Hz to 22.4 kHz.

This standard has not been amended since several decades : it is a very well-established method of noise measurement.

I have at my disposal a Rohde & Schwarz psophometer which incorporates both CCIR compliant circuits and flat band-limited RMS noise measurement circuits. It is very enlightening to measure and compare the noise levels of various devices obtained with both methods.

By the way, I think the CCIR noise measurement standard shows in itself why noise-shaped dither can be interesting to improve the subjective performances of a digital chain in certain conditions.

Is this the reason why the Parasound DAC-1100 HD of my friend Peter (from 1999 or so) wins any (nonscientific) blind shoutout against the modern Chinese DACs?
(We did it in Cologne with bended eyes and 5 min between each session AND it was not double blind)

 

[NTTY]

You may easily create this kind of signal with Sox.

I'll post how here later today.

First read:
Following post from user @miero on ASR explains how to play a J-Test signal with sox

Thank you! I saw Multitone from @pkane offers the possibility to save a WAV file of a 16bits/44.1kHz JTest. So if it’s better than the one of REW, it will be it.
Thanks again!

 

[Multicore]

Nice measurements, but the Industrial design is very strange. It's as if they were halfway through designing a rack mount chassis and then someone pointed out it was a top loader...

Right. If they had ordered the tray-load drive then they could have put the holes in the lugs and it would all make sense.

 

[Deleted member 77012]

Right. It may make some sense to buy a really nice looking device even though it measures well. But this one does really not look nice at all.

 

[Canuck57]

@NTTY thank you for the very detailed review of this CD player.
I think it's a cool looking unit!

 

[kencreten]

Thank you very much for your effort! This test shows, that perfection was available at least two decades ago. We could have stopped at this point, but the show must go on. It reminds me of the ongoing "evolution" of shavers: they are beating a dead horse for decades now, but still the crowd jumps on the new 5-blade xy-shaver for a smoother shaving experience....

I hear you. Personally I won't use any less than 7 blades on my shavers (joking) heh

 

[AndreaT]

I understand your trepidation in testing a CD player that shaped your listening session for the past two decades. I have an old Esoteric X-05, I purchased in 2008. It is my main system first line of back-up when the Weiss MAN-301 is re-reading the digital collection on the SSDs and therefore not available. It was expensive, it does not have AES/EBU digital out, plays SACD and it still works as a CD transport through its S/PDIF digital output. I wonder how will it measure if tested today. Thank you for a well done review of a state of the art player from two decades ago.

 

[AnalogSteph]

Is this the reason why the Parasound DAC-1100 HD of my friend Peter (from 1999 or so) wins any (nonscientific) blind shoutout against the modern Chinese DACs?
(We did it in Cologne with bended eyes and 5 min between each session AND it was not double blind)

At the very least that's a PMD100 based DAC, which may mean an extra 6 dB of digital headroom on CD material if Parasound chose to use the chip's internal gain scaling, in addition to the standard 1 dB attenuation. I would rather look for any advantages there, especially if fed straight from a CD player (any ESS-based DAC with DPLL on basically hard-clips at 0 dBFS, although you can turn down the volume a few dB and things should straighten out). The extra dither seems to be beneficial in straightening out small-scale differential nonlinearity issues but no replacement for proper MSB trimming, apparently.

 

[ichonderoga]

Excellent work. All of a sudden I want a CD-player again

 

[Scytales]

Now to be fair, DSD64 is about the equivalent of 18/88.

I think that the effective resolution of DSD (even at the original 64FS, with FS=44.1 ksps) cannot be summarized as unequivocally as that.

One of the fascinating properties of sigma-delta modulation is that with the very same bit-rate (single bit and a known sampling rate), different modulator designs can produce very different output resolutions.

The Italian magazine Audioreview I referenced above once published two articles in 2014 where they reported studies made with the help of an Italian university, where a special software written to digitally analyze single bit modulations was developed. Various sigma-delta modulator designs of that time were studied : all the original from Philips, plus a novel design based on Trellis algorithm, Korg's Audiogate, Jriver and three different modulators available in the Weiss Saracon software. Sadly, none of the Sony Feed-Forward delta-sigma modulators were part of the study.

This two articles are available online for free :

DSD Un pianeta tutto da esplorare (part 1)

DSD Un pianeta tutto da esplorare (part 2)

The first table in part 1 summarizes the computed effective resolution in a 400 Hz to 20 kHz bandwidth of all the modulators that were part of the study. It shows that the effective resolutions of these modulators spread over a wide 11.7 bits span, from 13.2 bits to 24.9 bits.

The remaining graphs and tables in this two articles are very interesting. Unfortunately, they are only in low resolution on the website.

 

[ObjectiveSubjectivist]

In Poland we have this saying:
"Ugly as a night in November"

 

[NTTY]

The next interesting part would be how the analog output fares with this. This player is using a WM8740 DAC, a relatively high-performance part in its day (DR 117 dB(A), THD+N -104 dB). You may need to resort to the -20 dBFS tone for an accurate picture as that is getting close to the Ultralite's capabilities.

I already tested but did not publish.
As you suggest, I’m hitting the noise floor of the Ultralite, because the Orpheus outputs only 2Vrms and so It’s too low to benefit from the full resolution of the Motu.
From memory it stopped at 17.6bit. But I can report and compare directly with an SACD. If you could do me a -16dBFS sine with shape dither, it could be a direct comparison.
In fact I tried to replicate following your instructions to create a shape dithered file but I always end up with very low level of distorsion in the file, so it’s not as good as yours.
Thanks in advance!

That tells me it's a purely impedance-balanced output, i.e. one leg probably just has a resistor and coupling cap going to ground. If it's good enough for Behringer and countless other manufacturers, it was good enough for Orpheus, apparently.

You’re correct.

It's a good thing it has those outputs, too, as the device is firmly IEC Class I. Hence also the absence of any mains leakage to speak of. What's a good thing in (galvanic) isolation could turn into a ground loop in a real setup though.

Would there be a simple way for me to test that?

That's peculiar, it behaves more like a NOS player. I wonder whether there is any upsampling following digital attenuation + limiting going on in one of the socketed PLCC package ICs. (If you ever need to service this unit, I'd almost be tempted to buy a PLCC chip puller. A bit of contact cleaner may eventually be appreciated.)

I was surprised by the results too. I supposed it was headroom given to the OS filter. The white noise analysis shows there’s decent filtering, so it’s not NOS. But I’m not sure of what you meant about contact cleaner need.

Build appears neat and tidy inside with good-quality parts, very Swiss that. It's somewhat reminiscent of some entry-level test gear I've seen. Still it's not exactly impressive what you get in terms of electronics for what must have been north of $5000 at the time... I guess being made in small numbers in Switzerland did it no favors in that regard.

Yes cost of living and salaries are very high in Switzerland, so it makes anything built here very expensive outside Switzerland.

 

[restorer-john]

@NTTY

There's one simple test that's always overlooked by reviewers (except JA). Polarity.

Either use a clipped sine or a single 0dBFS impulse.

 

[NTTY]

You're right @restorer-john. John always starts his reviews with this information and I'm happy to include it. In reality there's no reason why I did not do it earlier. I'll update accordingly.

EDIT: The balanced outputs of The Orpheus Zero invert absolute polarity. That said, the interface allows to invert the polarity, and it shows as is (below) not to forget about it:

Impulse responses in the two cases below:

Cheers