A modest hypothesis on the 16 bit question

[Mallette]

Having slept on it and rested my ears, my astonishment that 5 varied pairs of ears detected such minimal differences in the Rega Planet and MBS-3 remains. Chris Purcell, Shaunn's 17 year old (who should DEFINITELY meet Justin sometime) is, of course, in possession of that undiminished range of frequency response (enjoy, Chris...it won't always be that way) only the young possess and according to his father, slept between K'horns as a baby, was especially careful to match levels and syncing between the selections. I would also say I was really blown away with the appearance of the Rega in addition to it's sound. Solid as a brick, all the money in the right places. The simple top loading manual cover makes great sense as opposed to the flimsy automatic trays usually served.

The only significant difference seemed to be connected to error correction. This was most pronounced on my location recordings, which were the only ones I knew for sure exactly how the music had been treated from source to ear. It was agreed by all that there were places where the MBS-3 had the edge in lacking a certain bit of what we decided was digital artifacting "riding" on top of the music. This has led me to postulate the following:

With equal DACS employed, no real-time CD player can equal the sound of PC-based playback due to the lack of time and resources to restore the archaic Redbook format files to accuracy.

That should be good for some discussion!

Perhaps someone should build a Rega that can play wav files?

The epiphany that put the above in my mind was when I was explaining to Mike and the Purcells that neither of my ripping computers, in spite of having up to 48X drives, ever exceeds 2.7X extraction speed. Futher, if you watch the time remaining you will find that actual processing to disc time after read is even less, sometimes just over 1-1. This is due to Exact Audio Copy being set to the highest possible accuracy settings and the length of time before it is satisfied that the file it has built is as close it is going to get to the original Redbook.

Therefore that slight "edge" in quality most obvious on very pure acoustic sources.

I might add that, compared to my location recordings, several very highly regarded CD's appeared to have just a thin "veil" between us and the source. I suspect this is a dithering from hi-res and/or processsing issue.

All in all, an immensely pleasant, instructive, and brain food evening with great company! Thanks to Mike and the Purcells (as well as the PAW) for participating! Also, Mike seems to think my 'horns could use the P traps Bob Lemker prescribed for his Cornies. But that's another story...

Dave

[NOSValves]

Mallet,

I have to say I find this all very interesting and makes perfect sence. I know many seem to think that a PC is full of stray electrical noises and what not. My music server seems dead silent. As it is now I will have to buy some HD space what do you think about a raid array ? a better soundcard is a must but that will have to wait for money to free up. I just have to many things going on. I do have one question that will have to be addressed. What is the quietest PC Power supply available the fan is going to drive me nuts on the one in there now. Keep in mind I also plan to upgrade the Processor and motherboard to the P-4 platfrom someday so I want a healthing power source for future upgrades.

Craig

[Mallette]

See http://www.pcpowercooling.com/prices/index.htm for some excellent quiet power supplies and fans. I used a "Silencer" PS in the MBS-2 and it is very much quieter than the MBS-3. Also, the prices have dropped a bit since then.

A raid array is a great idea, as we know HDD's WILL die. OTOH, HDD's used exclusively for music should last indefinitely longer than system drives as they are accessed much less and the heads should be reading pretty linearly most of the time. Also, a new sound card would be cheaper.

As to a P-4, I wouldn't bother unless you are going to make it also a video/HTPC. Anything over 400mhz or so is a waste for audio purposes. I may have to move that up a bit when I get 24/196 ability, as that is over 8 times that data of CD! However, the above holds for 24/96 and below.

Dave

[Mallette]

OK, so I'm talking to myself here. I sez "Self, why doesn't somebody build a CD Player that:"

1. Has a 10 mbyte or so buffer with a ROM routine for correcting Redbook extraction errors to the level that EAC does, and

2. Can play wav files as well.

Personally, I think option item 2 would make Redbook a "legacy" format since it would solve many ills.

Option 1 suggests CD player builders have been throwing money at the wrong issues.

Dave

[Deang]

"The only significant difference seemed to be connected to error correction. This was most pronounced on my location recordings, which were the only ones I knew for sure exactly how the music had been treated from source to ear. It was agreed by all that there were places where the MBS-3 had the edge in lacking a certain bit of what we decided was digital artifacting "riding" on top of the music. This has led me to postulate the following:"

"With equal DACS employed, no real-time CD player can equal the sound of PC-based playback due to the lack of time and resources to restore the archaic Redbook format files to accuracy."

With all due respect -- this is quite a leap! I can only agree with the very last sentence in the context of older recordings.

It may be that I am just not completely clear on what you guys did. At what point were the same DAC's employed? Seems to me you may be comparing more of the actual sourcing medium (PC HD vs. CD), than the complete hardware package if you are using the same DAC's.

At any rate -- at times, and in various places -- the system bested a 5 year old machine using outdated DAC's.

[Mallette]

See http://www.pcpowercooling.com/prices/index.htm for some excellent quiet power supplies and fans. I used a "Silencer" PS in the MBS-2 and it is very much quieter than the MBS-3. Also, the prices have dropped a bit since then.

A raid array is a great idea, as we know HDD's WILL die. OTOH, HDD's used exclusively for music should last indefinitely longer than system drives as they are accessed much less and the heads should be reading pretty linearly most of the time. Also, a new sound card would be cheaper.

As to a P-4, I wouldn't bother unless you are going to make it also a video/HTPC. Anything over 400mhz or so is a waste for audio purposes. I may have to move that up a bit when I get 24/196 ability, as that is over 8 times that data of CD! However, the above holds for 24/96 and below.

Dave

[artto]

Aren't RAID arrays usually (or always?) SCSI based? SCSI uses substantially more system resources than standard ATA drives. Sound Forge recommended that I not use SCSI drives for that reason.

[Mallette]

No. Mostly IDE these days. Most mid to upper MB's support it on board.

Dave

[Marvel]

SCSI actually has less overhead than IDE. However, there are IDE RAID controllers out now, and controllers and drives that are 133Mbs.

Craig,

Most fan noise is from the CPU fans and not the PS fan. I have an Enermax in my dual Athlon, but the two CPU fans are too loud. I can't even tell there are two fans in the PS unit. Folks I know who use a computer for a DAW (digital audio workstation) end up putting them on the other side of a wall, with just the monito mouse and keyboard where they work. They can hear spiders on the walls that way. My younger son's new PIV (single cpu) is pretty quiet. I am in the process of building balanced i/o for the Audiophile 2496, and getting a video/mouse/keyboard extender to move my system out of my editing suite (read -- small corner off the bedroom). That's why I still record on my 8 track ADAT, as it is much quieter than the pc. The KVM works over Cat5, so the system can be a few hundred feet away.

You could do water cooling if you wanted to, but the easiest way is to move it.

Marvel

[Mallette]

While I have not personally tried them, the PC Power and Cooling site sells supposedly quiet CPU coolers as well as power supplies.

Frankly, I am a bit surprised my hypothesis has kicked up no more comment than it has. Can't tell if the silence is assent or that it's complete nonsense.

Dave

[Mallette]

While I have not personally tried them, the PC Power and Cooling site sells supposedly quiet CPU coolers as well as power supplies.

Frankly, I am a bit surprised my hypothesis has kicked up no more comment than it has. Can't tell if the silence is assent or that it's complete nonsense.

Dave

[Tom Mobley]

SCSI is not hard on the processor unless you're using an software-based RAID like was available in NT4 when it first came out. I don't think anybody much is doing that any more. Modern SCSI RAID is hardware based, doesn't impact the processor much if any. I've been looking at some used SCSI cabinets on ebay, these could be loaded up with modern large drives and used pretty cheap.

One caveat to those getting interested in RAID: THere's several flavors of RAID, all involve treating several hardware devices as one larger logical drive. The issue is that if one of the drives fails you lose the info on all of them since the data is striped across them. RAID5 bypasses this by using some of the space for parity data. If one device fails the array can continue to function rebuilding the data on the fly from the parity info stored in the striping. There's usually some sort of alerting function that tips you off that there's problems, it's assumed you will replace the bum drive soonly.

An issue here is that you lose the capacity of one of the physical devices. In other words, 4 100GB physical drives become a 300GB RAID5 array. The more physical devices the smaller the percentage hit there is.

Another issue that crops up soon is backup, what do you use to backup a 400GB drive? there's nothing cheap laying around, let me assure you.

IDE RAID? lots of current mobo's support this, but I haven't fooled with it at all, don't know if RAID5 is supported.

THe key here is that if you use RAID that's not RAID5 you have be willing to lose the data in the event of a hardware failure. This implies the existance of an acceptable back-up strategy or the ackknowledgement on the part of the data owner that he can stand to lose it, i.e., can be rebuilt from other sources or whatever.

I work in server support at a large data center, we have around 3200 servers running on the production floor. That's a lot of servers and even more drives, a lot of these are hooked up to large RAID arrays. hardware failure is not really common, and there's high-speed backup across a gigabit network to really large storage on mainframes.

Tom

[Tom Mobley]

Hmpf, multi-post paradise here.

Tom

[artto]

Ok. I guess I'll throw my two cents in. I've suspected for a very long time that error correction had a lot do with it (as Dave mentioned his initial post). I would even go so far to say that error correction is responsible for at least some, if not most of the sound differences we hear between CD players. And even certain discs. Not every company uses the same error correction scheme.

I've noticed in mastering my own digital recordings, that if I "do too many things", such as excessive edits, resampling, acoustic mirrors, time compression, lots of indexing, etc., the more problems I have burning CD's that play trouble free on nearly any equipment.

One particular situation a couple of years ago was so bad it prompted me to buy a new CD player because I thought my existing CD player (Revox B225) finally went bad. The CD played fine on the new CD player (Denon 600F pro audio). The Revox (circa 1984) had nothing more than standard CIRC (cross interleave code) error correction. No oversampling. It obviously couldnt deal with the amount of error correction that was required with this particular disc.

I recently asked Dave about some problems I was having on my most recent recording since were both using Sound Forge. Dave, it turns out the is problem similar to what I mentioned above. Im quite sure at this point that resampling from 48Khz down to 44.1Khz for the CD burn, AFTER editing, inserting fades, etc., had something to do with it. When I re-recorded the DAT down to the hard drive at 44.1 instead of 48KHz & started the editing from there, the tracking problems disappeared. Before doing that, I backed thru each step of the mastering process, one by one (I save each stage of the mastering process as I go thru it so I can backtrack if anything goes wrong) & was still having problems after the final indexing was put in until I dumped the initial file that I edited AND THEN later converted to 44.1 KHz.

Sound Forge calculates the waveform envelopes, cross fades, & any other effects in real time when burning a CD. All of this can add up to a CD that has more errors on it. And at some point, the frequency & magnitude of the errors can become more than what the CD player can deal with.

[jazman]

Dave,

I have not tried the PC approach for CD playback. I do own a Rega Planet. Your opening comments about it's build quality and ergo's are on the mark. As for the veil, how does that apply if you're using the same off the shelf CD for playback? I would expect any massaged recording to sound better via the PC and sound card. Looking at the dollars and cents of off the shelf CD playback, a used Rega at approximately $350-400 is tough to beat, period. Give the Rega some first class interconnects and you've really got something as an audio component. The PC as a music server and editing device cannot be denied it's utility vs any CD player. I just kind of think both have different purposes and directions. Your early on premise, some many weeks ago, was that your PC playback was untouchable vs any CD player for the price. Did the Rega change your mind with respect to quality CD playback vs cost, is the question I would like answered.

Klipsch out.

[Mallette]

It's my opinion you are correct in your analysis. Transcoding appears to be the root of all digital evil not accounted for by Redbook approximation errors. Stick to either recording at the target sample rate, or exact mathmatical multiples.

I'm still amazed at how little interest their appears to be in this topic. We are working at the very roots of digital issues, and it is important to anyone who has any disatisfaction with their digital sonics and only irrelevant to those who listen to nothing but analog. In spite of my thrust in all these posts, I still spend most of my time listening to analog sources, and probably always will due to the fact that I can get extraordinary music for 50 cents that no Redbook CD will ever even approximate, much less surpass. But when I record, I've no real choice but the digital domain, so I want it to be as good as it can.

Dave

[anarchist]

IDE Raid only provides Raid 0 and Raid 1. Its purpose is largely performance not data integrity.

Mallet, I have several questions from a theoretical standpoint and will play Devil's Advocate here in the interest of thoughtful analysis.

The benefit of your audio server with respect to CD's; you believe the sound is better because you are able to read the CD as it was meant to be read without error and postulate a CD player such as the Rega never plays it the same way. Is that an adequate understanding?

A player reads a stream of bits continuously and in order. Should a bit be unreadable, error correction is used to prevent a skip. If the CD is not damaged in any way, playback - the audio stream - should be consistent not ever changing as you suggest.

You go on to state your program is reading the bitstream as accurately as possible. I would theorize it would be just as accurate as a CD player when the CD was undamaged and capable of recovering bad bits (due to scratching, poor pitting, etc.) to a better degree since it wasn't occuring in real time.

The real benefit regarding accuracy as I see it would be in the case of damaged CD's.

Playback on a PC. Here I think we ignore several factors which could hinder playback. A hard-drive is not thoroughly accurate and must often be re-read. It is not always at a consistent speed either. Disk I/O frequently suffers momentarily due to CPU interrupts and other tasks which suddenly 'interrupt.' Given the concern over a phonograph varying from 33RPM, I would expect the same slight variations in sound to occur using your computer based playback system.

Lastly, regarding CD playback. If the CD is 16/44, I question how it could possibly be improved by recording it to a 24/96 wave. No matter how good the copy, it can't be better than the source it was derived from - referring to accuracy, certainly you could manipulate the mix and apply effects to give a better sound but it certainly wouldn't be the same music as recorded or intended.

The floor is yours.

[Mallette]

Thoughtful analysis is always a good thing, and I am pleased you are providing some>

>The benefit of your audio server with respect to CD's; you believe the sound is better because you are able to read the CD as it was meant to be read without error and postulate a CD player such as the Rega never plays it the same way. Is that an adequate understanding?

Close. What I am postulating is that it is not possible for the Rega or any realtime CD player available to provide complete data accuracy on the fly. That MIGHT explain why all ears last night felt they were hearing digital artifacting on the most critical music we listened to. Of course, it might be that the Rega is simply inferior to the Card Deluxe. I can only repeat what we heard. Why remains to be sorted out.

>A player reads a stream of bits continuously and in order. Should a bit be unreadable, error correction is used to prevent a skip. If the CD is not damaged in any way, playback - the audio stream - should be consistent not ever changing as you suggest.

I am aware that a PC CDROM acts this way, but my understanding of Redbook is that it is less constant and the error correction routines are there to minimize this. Complete drop outs, which never occur with HDD playback, occur when the error exceeds the ability to recover. Inaccuraccy occurs between that point and perfect reading.

>A hard-drive is not thoroughly accurate and must often be re-read. It is not always at a consistent speed either. Disk I/O frequently suffers momentarily due to CPU interrupts and other tasks which suddenly 'interrupt.' Given the concern over a phonograph varying from 33RPM, I would expect the same slight variations in sound to occur using your computer based playback system.

Some of the above is my OH, this part I utterly reject. When an HDD reads inaccurately, your system promptly crashes. They MUST be 100% bit accurate. Further, at 5400 to 10,000 RPM and above, their speed accuracy compared to CD or 33.33 RPM is certainly no issue. Regardless of that, the CPU will process the bits and bytes at the clock rate, smoothing any inconsitencies that might occur. Several back up depths there.

>Lastly, regarding CD playback. If the CD is 16/44, I question how it could possibly be improved by recording it to a 24/96 wave.

Here we are 100% in agreement. Nothing good could come of this, and the signal would almost certainly sound worse due to the dithering that would be required to compensate for the indivisibility of 96 by 2.

Thanks for the thoughtful input. It's about time we heard some, and the floor is yours...

Dave

[anarchist]

Hi Mallett,

Some good thoughts there - the Redbook thing I need to do some research on as I don't want to stand on shaky ground.

Ok, some of your points I think merit further consideration:

I agree with the idea that a real-time player could not resolve bit errors as well (or at all) as a utility going to great lengths and long duration. That said, we still do not know we have complete accuracy. Depending on the problem, your utility is quite likely to skip bits as well - a better, more thorough attempt to read these troubled bits is certain, the result is questionable unless we can review the code and the assumptions made by the developer. It is almost a certainty that the program, upon finding a bit very difficult or impossible to read, assumes a value based on those surrounding it.

A disk I/O error does not always lead to an instant crash. In fact, most people aren't aware they have even suffered a disk I/O error. If we could orchestrate a mass simultaneous disk scan of all personal PC's in peoples homes, there would be enough check files generated by scandisk to fill hundreds of those fancy new 10,000 RPM drives. You are correct that there are layers of back-ups in place - the disk reattempts to read the data stream, the OS utilizes a different access method to read the data, the program utilizes error code to identify an issue and force another attempt, utility programs seek and correct errors, etc. My thought concerns what the application does when encountering such an error during playback and haven't we just encountered what we were trying to avoid?

We, however, are searching for the Holy Grail in accuracy in this discussion and its impact on the musicality of any piece appears to be relatively minor based on your listening test results.

To the playback issue and interrupts. My experience is much, much different. With both a 464 mhz processor and a 700 mhz processor, I frequently have momentary glitches during playback when the disk IO subsystem must give pause to some other process running on the system; not an application mind you but simply the operating system processes. How are you working around this?

[Mallette]

Quite a few folks to respond to.

Dean: Seem to have overlooked your post earlier.

>It may be that I am just not completely clear on what you guys did. At what point were the same DAC's employed? Seems to me you may be comparing more of the actual sourcing medium (PC HD vs. CD), than the complete hardware package if you are using the same DAC's.

>At any rate -- at times, and in various places -- the system bested a 5 year old machine using outdated DAC's.

Why would we want to use the same DAC's? Not quite sure what you mean there. As to your last statement, are you implying that a CD player has to be replaced every 5 years? Seems a bit expensive, and that's the first time I've heard someone suggest only the latest Rega's are any good. Finally, try to remember this was not a "PC's rule and CD players drool" contest but simply an attempt to break the prejudice barrier against PC's having to ride in the back of the bus. No one mentioned the "massive RFI" ghost and other reputed PC issues.

Jazman:

>As for the veil, how does that apply if you're using the same off the shelf CD for playback? I would expect any massaged recording to sound better via the PC and sound card.

We did use the same OTS CD's for playback. Everyone brought what they considered a top quality CD with which they were familiar. As to "massaged," if any of these were "massaged" it would be the OTS stuff...and, IMOH, that may be responsible for the "viel" which might well be evidence of transcoding from odd sample rates. My location recordings are as close to the old "direct to disk" analog tradition as digital allows. Straight from the mic preamp (not even a mixer) into the Card Deluxe to the CD with only sequencing and trimming of audience applause and such. No eq, no normalizing, no nothing. The CD is as close as bit for bit to the original as technology allows. While this was not a test of my recordings, comments were made that they were more natural sounding on both sources than the commercial materials.

>I just kind of think both have different purposes and directions.

True to some extent, but I have certainly come to believe that the PC approach is far more cost effective and useful to the audiophile than a stand alone player.

>Did the Rega change your mind with respect to quality CD playback vs cost, is the question I would like answered.

Absolutely! I'd have one of these in a minute if I needed or needed to recommend a low cost, top quality CD playback device. However, in my own two playback systems I much prefer the random access, 24/96, and other advantages of PC based playback at about the same cost and playback quality.

Anarchist:

>A disk I/O error does not always lead to an instant crash.

We are drifting semantically. When I say disk I/O error, I am implying uncorrected. Such an error will always lead to a system crash if that error is from the system disk or program code. Obviously a mis-read from a wav file wouldn't cause a crash. AAMOF, it would take a LOT of errors to even be audible, as ordinary Redbook playback proves.

>With both a 464 mhz processor and a 700 mhz processor, I frequently have momentary glitches during playback when the disk IO subsystem must give pause to some other process running on the system; not an application mind you but simply the operating system processes. How are you working around this?

Having built a LOT of machines for audio purpose over the years, I can definitely tell you you have a problem there. You have plenty of processor power for even hi res audio in both machines. I've successfully built DAW's all the way back to the mid-90's based on a lot less power. Can you imagine my chagrin if such dropouts happened during a location recording? I am not a PC engineer, so I cannot tell you exactly what is going on with your drives. If the issue is thermal recalibration which was a problem at one time with some drives, new AV rated drives will fix it. Otherwise, you have some much deeper problem in the motherboard/disk controller area. I can assure you I NEVER get even a hint of dropout or glitch from any source connected to the PC. If I did, I'd silently fold my tent and slink away.

OK, that gets me up to the moment, I think.

Dave