Narrow-Bandwidth Television Association

I'm surprised to find there isn't an easy supply of crystals (or oscillators) that are an integer multiple of 44.1kHz, the audio CD sample rate. On the assumption that a CD player is rock-bound, what crystal frequencies are common?

Perhaps it's synthesized from some lower frequency via a PLL? But that adds complexity and cost.

The two common frequencies I found closest were both 97ppm out, 9.2160MHz and 18.4320MHz, divisions of 209 and 418 respectively. This is an error of about one third of a second per hour. That error is comparable to the frequency error of a consumer quality crystal, typically ±100ppm though far better exist.

Without actually opening up the only audio CD player we have here, can anyone shed light on this?

Steve A.

P.S. Some time back Klaas did explain where the 44.1kHz frequency came from, I recall it was related to video. But I've done a search here and nothing thus far has emerged.

Probably not helpful to you, but I'm fairly sure that I read most computer soundcards these days use either a 48kHz or 96kHz clock internally. This maybe accounts for 44.1kHz crystals being harder than expected to obtain.

It's possible that there are monolithic integrated circuits for CD player circuitry that have an internal RC oscillator, or suchlike. (remembering that it's audio and most people won't even know if the clock isn't accurate)

Steve Anderson wrote:
P.S. Some time back Klaas did explain where the 44.1kHz frequency came from, I recall it was related to video. But I've done a search here and nothing thus far has emerged.

Wikipedia wrote:It also specifies the form of digital audio encoding: 2-channel signed 16-bit Linear PCM sampled at 44,100 Hz. This sample rate is adapted from that attained when recording digital audio on a PAL (or NTSC) videotape with a PCM adaptor, an earlier way of storing digital audio.[6]

An audio CD can represent frequencies up to 22.05 kHz, the Nyquist frequency of the 44.1 kHz sample rate.

http://stason.org/TULARC/pc/cd-recordable/2-35-Why-44-1KHz-Why-not-48KHz.html

The reason I posed this question is that it's quite easy to generate and use 48kHz or its multiples and get a micro to rum at some integer multiple of 48kHz via suitable selection of off-the-shelf crystal(s). But in anticipation of those that would want a 44.1kHz result I did hit a bit of a brick wall, hence my question.

As I mentioned before, if an error of 97ppm is acceptable, then it can be done (350ms in one hour). If not...well, what crystals do CD players use? Or is it a case of a few percent either way is good enough and we'll use a ceramic resonator? (read cheap).

For so many reasons I prefer 48kHz sampling, DVD's use it, broadcaster use it or some multiple, 44.1kHz is insane, but the product of history. Sony and Philips rushed out the audio CD to understandably get ahead of others developing similar formats. Sadly they didn't quite 'tidy it up'.

But anyway, there is a mathematical relationship between 625/50i FSTV and 44.1kHz....I just can't recall what it is...

Steve A.

Well, a few ppm is not critical for audio, but they do become a serious snag for NBTV.

When Vic Brown asked me to help out with some software for his vestigial side band scheme for transmitting NBTV over the amateur radio bands he told me they were using crystal locking with trim (neswletter V32 No. 4) to negate the differences between source and receiver, particularly when the source is a cd player.

I was surprised the small ppm variation in crystals was causing a problem and said so, however on doing a quick test I verified it is a problem.

Note that for NBTV over the air it is problematical to use syncs and so having accurate clocking becomes important if not essential.

I attach a video is what happens with the ppm error you describe Steve.

PS: I have developed software to trim these slight clocking errors out for people using computers to display the NBTV video.

Thanks for your input Gary. Although Vic and partners frame drift isn't what I was quite driving at, it is related. I presume that due to the difficulties they encountered in handling syncs in a VSB RF TX/RX circuit they went sync-less. This requires a rock-solid common frequency and/or time reference.

Three sources of this spring to mind. The simplest is the 50/60Hz mains, it may not be the most stable (even though long-term it is) but if the two ends of the link are powered from the same power generating station then that should work. The advantage is that it's quite easy and cheap to do.

The second is to use the off-air TV signal from the same network, again using the same transmitter so the two ends of the link would have to be able to receive the same TV transmitter. Not so easy as using the mains, but everyone has a TV. There is a potential problem though if using the same network, but different transmitters, e.g. London and Glasgow on BBC1. Often the incoming feed is re-synchronized to a local station reference via a frame store. This can result in a very slight drift, but far, far less than in your avi. However I don't think you could use a digital system and/or satellite downlnks. With the shut-down of the analogue terrestrial service this option is vanishing.

The obvious one is GPS using the 1PPS from the receiver. Not cheap but there's no other better world-wide. Some GPS receivers do also output frequency-locked waveforms, 10kHz is the most common.

But returning to 44.1kHz, I guess I'm gonna have to open up the CD player and see what I can find!

Steve A.

P.S. Having actually read the item in Vol. 32 No. 4 I realize the first two options are a no-go as it was trans-Atlantic! That means GPS or nothing.

Steve Anderson wrote:Thanks for your input Gary. Although Vic and partners frame drift isn't what I was quite driving at.

Aah? so not related to NBTV? I guess then, the ppm error may or may not be a problem.

gary wrote:...Aah? so not related to NBTV?...

It is actually for NBTV, but even if it were not, where/how do I derive 44.1kHz? Somehow the millions of CD players and PCs do it...it seems to be either synthesis or custom xtals...or the small error of 97 ppm is ignored...or given the math I should be able to do it. Klaas, where are you?

Steve A.

Oh, I realised that generating 44.1kHz was your aim here, I was merely indicating that a drift of 97 ppm is common in CD players, and at the same time I thought it interesting to show the effect that has on the video signal.

Isn't a 11.2896 Mhz Crystal used to get 44.1? i.e. binary division 256×44.1

or am I off the page again?

http://uk.farnell.com/abracon/abm7-11-2 ... dp/1812369

gary wrote:...Isn't a 11.2896 Mhz Crystal used to get 44.1? i.e. binary division 256×44.1....or am I off the page again?

Nope, absolutely correct! I did a simple spreadsheet that listed all the integer multiples of 44.1kHz and sure enough 11.2896MHz is there. But no local supplier stocks that value, or multiples (e.g. 22.5792MHz).

So my question is answered, I'll just have to order from the UK or elsewhere. Thanks Gary, that saves me cracking open the CD player.

Steve A.

just some others but I don't think they're any easier to obtain (although you have probably calculated them by now):

16.9344 (384 integer division)
22.5792 (512 binary division)
33.8688 (768 integer division)
45.1584 (1024 binary division)

gary wrote:...just some others but I don't think they're any easier to obtain...

Quite correct, well not here...it is a contingency plan anyway. I intend using 48kHz sampling, but I know that someone will ask can it be done at 44.1kHz so I can burn an audio CD?

48.0kHz x 256 = 12.288MHz which local stockists have and I have in stock here at home as well, ditto 48kHz x 512 = 24.576MHz.

Steve A.

Steve, i was curious after reading this thread and decided to pop open one of my cd players from many a year ago (manufactured August 1999) to see what was in it. sure enough, there is a crystal that reads "ZTA, 16.93MX".

Lawnboy wrote:....sure enough, there is a crystal that reads "ZTA, 16.93MX".

That looks awfully like a ceramic resonator rather than a crystal where the frequency tolerance is likely to be ±0.5% rather than ±0.01% of even the cheapest crystal...gotta keep those costs down! In practice, typically it will probably be much closer than 0.5%, but there's no guarantee of that. Evidence that supports my previous statement and question...is an error of 97ppm in a CD player deemed acceptable?...it would appear so...at least to this manufacturer. If you're a musician with absolute pitch this should worry you.

Anyway, thank you for going to the trouble of pulling it apart and taking the photos, much appreciated.

Steve A.

Steve Anderson wrote:P.S. Having actually read the item in Vol. 32 No. 4 I realize the first two options are a no-go as it was trans-Atlantic! That means GPS or nothing.

For UK based things you could use the standards transmission from the national physics laboratory. In the US you could use one of the WWV signals.

As you say, if you want to synchronise your clocks globally then you need to use GPS. (you can receive WWV around the world but being shortwave the signals tend to come and go)