Given some recent misrepresentations about PCM, I was reminded that, at some point in the past, most CD players did in fact use sample-and-hold DAC technology. Which, as we know now, is rather inadequate for numerically accurate audio reproduction (despite the whole NOS crowd saying otherwise). RockFan's criticisms of PCM and its stairsteps would be correct if everybody did in fact use S&H, but we're pretty sure that nobody sane uses it nowadays.

However, I'm not really aware of any objective proof of that statement. In order to actually assert its truth, you need to verify oversampling on a large number of players. If you just want to make sure your own player oversamples, that is a lot easier - but very few people do that, or few manufacturers document their oversampling modes. I believe that it's taken as a point of faith, with little research backing it up, that most DACs used in audio equipment today use accurate oversampling and digital filtering.

I want to say that early sound cards - possibly up to the SB Live - did not, in fact, oversample. But I don't have any good evidence for that. So instead of proving that most/all DACs nowadays are oversampling properly, I'd like to take this from the opposite direction. Does anybody know of a mass market product, used for CD or DVD or MP3 playback, that uses sample-and-hold DACs, either now or in the past? (This can only be proven if the DAC only supports NOS, or it is an oversampling DAC that is known to be programmed into an NOS mode.)

If a large number of players produced today do in fact run NOS, then some arguments against PCM become slightly more compelling, as arguments against existing practices rather than against the state of the art.

With the low price of decent-enough oversampling DACs, it probably makes financial sense for a CD player designer targeting a particular sound quality to choose oversampling, because it will make the output filter much cheaper to implement.

That's just an assumption, though, and you are right that it would be interesting to test a large number of devices and see how they behave. Showing that a large number of real-world devices have poor DACs will make theoretical claims (that many of us make every time the topic comes up) about PCM quality pretty irrelevant. My feeling is that a few devices do have poor DACs, but those expecting an anti-PCM witch hunt will be somewhat dissapointed.

I'm sure others more qualified may come in and correct me here, but my understanding is that whether a DAC uses oversampling and whether it has a final analogue filter are different things. If a DAC is a NOS one, that doesn't mean it doesn't have a final filter. For example, I'm fairly sure that the very early Sony players from 1983 had non-oversampling DACs, but still had the final filter (to remove the reflected spectra). Similarly, the Cambridge CD2 from the late 1980s used 16x oversampling and didn't bother with a final filter (because the reflected spectra started up above 700kHz).

That said, I get the impression that today's trendy NOS DACs do tend to dispense with the final filter. I've seen a waveform posted somewhere else (I think it was on the Slim Devices forums) showing a 1kHz sine wave emerging from a NOS DAC, and it was indeed a stair-step shape. I presume the philosophy behind this strategy is that the rest of the system and/or the listener's ears/brain will naturally discard the ultrasonics. Perhaps this is less objectionable than the effect a final analogue filter might have. Or perhaps NOS DACs are just garbage. I don't know, never having heard one.


I think it's extremely unlikely that the SB Live has a NOS DAC. By the time of mainstream soundcards in PCs, DAC technology was well advanced to the point where any affordable chip was almost bound to use oversampling; indeed the majority of affordable chips are probably single-bit, which by definition have to use very fast oversampling to work at all.

IIRC (Google doesn't give a definitive reference) the _first_ CD player, the Sony CDP-101 had a 14-bit oversampling DAC. This (again, IIRC) was because they couldn't manage to get a linear 16-bit performance from the R2R ladder DACs of the time, so the 14-bit oversampled version gave better performance in terms of noise and distortion (never mind frequency response).

I don't know of any widely commercialised filterless DACs. Mind you, there was a trend for converters with switchable filter responses last decade - maybe some of those had no filter?

The reason you need some kind of filter is because lots of amps and speakers don't like bucket loads of ultrasonic junk being pushed through them. Who wants to play a 1kHz tone and get 43.1kHz and 45.1kHz tones on top!

Cheers,
David.

That's certainly true for the early Philips players - my first CD player was a Philips CD104, and that was a 14 bit 4x oversampler. But I thought that Sony used 16 bit DACs from the start. IIRC, Philips originally wanted the CD format to be 14 bit PCM (presumably because that's the best they could do at the time), and it was Sony who held out for 16 bit. So we have at least one thing to thank Sony for. (I'm struggling to think of another :-).

For anyone interested in the theoretical response of such a system (playing a white noise for example):

the rejection at 60 kHz is only 13.5 dB

So, from the very beginning of Redbook, all quality CD players have used oversampling, and if they didn't, it would be extremely audible. That sounds like pretty clear proof that the stair-step argument never had any basis in reality, right?

I'm not sure that is true. Philips certainly used oversampling, but at the time I think it was more to compensate for their inability to build a 16 bit DAC with good linearity. As I said in a previous post, I think Sony's early CD players were 16bit, non-oversampling. Philips certainly made a big marketing hoo-har when they introduced their 16bit 4x oversampling chip - which implies that at the time Sony were not able to do that.


Those that didn't oversample needed a brick-wall post-DAC filter to remove the aliasing. The received wisdom is that the phase errors this introduces into the passband are audible. I've never done any blind tests of such filters myself, and it would not surprise me if the critics might have overstated their audibility.

It's technically imaging, not aliasing (sorry, I couldn't resist being pedantic)

Phase error is one thing. I don't know if it will be audible, but as you say, it's likely that it's effect has been understated. The other problem with using non-oversampling then filtering is the steepness of the filter. If you saved a bit of cost and went with a 3dB per octave RC filter, then there would still be plenty of energy in bands above 20kHz. I read a test a while ago which suggested that IMD caused by inaudible high frequency material in tweeters is measurable (again, I don't know about audible) and is well within the hearing band. Google is not co-operating for the moment, so I'll post it when I find it again.

You're almost certainly right cliveb about the Sony vs Philips - I'd "heard" it was the first CD player, but it could equally well be "one of the first".



I found a Sony CDP-101 at university in the late 1990s. We also had Meridian's first CD player. Neither worked particularly well (functionally) - in fact the Sony was clearly faulty by then. All the same, it didn't sound "horrendous" as people suggest in retrospect. Maybe for the money at the time it wasn't good value sound quality-wise - I don't know - but as a CD player I found lying on a shelf it didn't jump out at me as inferior to everything else I'd heard in the preceding 15 years (which, if you believe all you read, it should have done).

We also had the first Sony CD player with a digital output, and the corresponding external DAC. I took the time to measure that. It wasn't linear around -90dB. You could hear that clearly, if you listened very loudly to quiet parts of correctly dithered recordings - they didn't sound correctly dithered, but quite distorted!

There was also an early music fidelity DAC with virtually no anti-image filter. It sounded OK, but (as you might expect) if you tried reproducing full scale ~20kHz sine wave, you could generate some horrible intermodulation distortion in some amplifiers and speakers.


I can just about imagine that at least some of this equipment was clearly inferior to good LP playback under some circumstances, and certainly audibly different from the master tape to those familiar with recorded music before it hit the LP. However, I can't imagine anyone listening to classical music with a wide dynamic range picking LP over CD, even in 1983. I have a Mozart Piano concerto on a CD from 1983. Contrary to what people say, it sounds lovely. I have heard people describe the recording as cold and clinical, but to me it sounds wonderful. So dramatically better than even the best LP, which will always have some background noise, the occasional louder sound that you don't want, and often some wow. Who wants all that when listening to Mozart?

Cheers,
David.

I stand corrected about early Sony players.

Given the preponderance of NOS DACs in audiophilia - at least compared to how many should exist if DACs were evaluated objectively, which means nobody would have one nowadays - it sounds like they should be pretty easy to tear apart. Either feed it a 22.1khz 0dbFS tone and watch the IMD sparks fly, or feed it very low level signals and observe distortion. Or, I suppose, if there's no analog filter, the frequency response artifacts (-1db at 10khz, roughly, from Seb's graph) would be audible?

hmm... reading this thread, something sounds odd to me. Few of the posts above say that garbage get's pumped into ultrasonic range under assumption that we can't hear it, but isn't the argument for "necessity" of higher bit depth and freqs for DVD-A and SACD is that we CAN hear ultrasonics or rather it influences audible sound?

That is the argument, but the general conviction of this forum is that the argument is snake oil.

Please don't take my hazy memories as gospel on this. If you wanted me to put a figure on it, I'd say that I was about 80% sure that early Sony CD players were 16 bit, non-oversampling. I could easily be wrong.


My recollection of early CD replay was that it sounded pretty horrid. Most people thought CD sounded "steely" and "brittle", but I reckon it's because in those days the systems we were listening on were optimised for LP's lack of top end. We were also probably used to how vinyl sounded, and so anything different seemed "wrong".


I didn't really much like classical until I heard it on CD. I bought my first CD player (a Philips 104) specifically for classical, continuing to buy rock on vinyl for a few years. Mind you, I still don't like Mozart. I know he was a genius, so It's my problem, and my loss. Maybe one day the penny will drop.

I think that's a huge amount to do with it.

AFAIK...

Some people thought radiograms sounded strange compared to gramophones (though they generally liked the transition).

Some people thought "hi-fi" sounded strange compared to radiograms etc (and they thought it sounded really "SSSSS"-ey and nasty).

Some people thought CD sounded strange compared to LP (and they thought it sounded really tinny).

I don't think many people can actually hear a difference between CD and DVD-A/SACD but most of the releases on the newer formats have good mastering, which sounds nice (though occasionally dull).


It is what you're used to. My stereo broke down some time in the mid 1990s, and took as age to repair (good old MasterCare!). I just listened to my gramophone in the mean time. When my stereo came back, it really did sound very strange. Suddenly there was lots of "top end" again, but it didn't sound right. If course you hear "top end" in everyday life in natural sounds, but that's accurate. From most stereos, it's quite inaccurate, and for a few days, I preferred the "no top end at all" from a gramophone rather than the "inaccurate" top end from a stereo.

The preference soon past, but it was quite a strange experience!

Cheers,
David.

The way oversampling moves images into the ultrasonic range is useful for one of two reasons. The first you got - it's probably inaudible. The second is that it's much easier (and cheaper) to filter it out there without mangling the signal you want. Making filter design easier means better sound quality for the same price.