So who is going to be first with trying to mod their cd/dvd player to use a clock like this?
http://news.harmony-central.com/Newp/2007/Antelope-Audio-Isochrone-10M.html
http://news.harmony-central.com/Newp/2007/Antelope-Audio-Isochrone-10M.html
rfbrw said:
Not hard at all. Accuracy and precission are two entirely different things.
Accuracy is a measure of how close a result (or average of a series of results) is to the true value or desired value.
Precission is a measure of how close a series of results are grouped together.
Imagine throwing darts at a dart board while aiming for the Bulls Eye. If all of the darts end up around the circumference of the dart board with none of them landing close to together........this is not very precise.....but the average position might be quite close to the Bulls Eye and be deemed to be accurate. In a second attempt at throwing the darts at the board - all of the darts might land in one neat group in the doubble 20 position. This would be very precise...but not accurate.
So - with clocks - you could have one that has a very accurate average output being only a few ppm or ppb from the specified value. However, the timing of the output from one instant to the next could still be varying considerably and therefore not very consistant. The accuracy would be good but the precision would be bad.
It is my understanding that the accuracy is not so important as long as it is within Red Book specs.....but the precission (low jitter)is most important.
Quote from Guido Tent:
Techtalk
I’d like to share with you some basics on jitter. Jitter
are short term deviations from something that is supposed to show constant performance. Consider a normal watch that ticks every second. If the ticks are not evenly spaced we have a jittery clock. Usually this is no problem, as long as the long-term stability is OK you’ll be in time for your next appointment. However,
there are occasions that clock-jitter is of utmost importance, for example a 1 00m race. Differences are in the order of 0.01 s so each second should precisely be second!.
Redbook
The same holds for audio! It is of no importance whether the clock is slightly off its’ absolute frequency, specified in ppm’s (parts per million). According to the redbook CD standard, limits are +/-100ppm. A single 440Hz tone then could only deviate +/- 0.044Hz- inaudible.
It now becomes clear that specifications like 2ppm accuracy do not add to sound quality. A TCXO is such clock that achieves low ppm values, as it is very constant over long time. Fine to have on board of an accurate frequency counter, but no use for audio.
Short term
The short-term stability is important! Cycle to cycle deviations, or cycle to 100.000th cycle. If you look at the spectral content of such signal, it should look like a single frequency, with the noise floor as low as possible
(as all periods T are equally long, this translates in a single frequency 1/T). Practical clocks do not show that very narrow single frequency, but a bit wider spectrum, rapidly decreasing at 10 to 100Hz from the central frequency. A well-known clock supplier tried to convince us by showing this graph:
What we actually see is the performance of the analysers’ video filter, namely 1 kHz. Good clock spectra are well withing this window. The only clock performance we can deduce from this measurement was the one that didn’t count for audio: It meets the redbook spec. It does not meet the manufacturers own spec of 5ppm though .........
Read the whole article here.
http://www.tentlabs.com/News/assets/Tentlabs newsletter September 2006.pdf
Techtalk
I’d like to share with you some basics on jitter. Jitter
are short term deviations from something that is supposed to show constant performance. Consider a normal watch that ticks every second. If the ticks are not evenly spaced we have a jittery clock. Usually this is no problem, as long as the long-term stability is OK you’ll be in time for your next appointment. However,
there are occasions that clock-jitter is of utmost importance, for example a 1 00m race. Differences are in the order of 0.01 s so each second should precisely be second!.
Redbook
The same holds for audio! It is of no importance whether the clock is slightly off its’ absolute frequency, specified in ppm’s (parts per million). According to the redbook CD standard, limits are +/-100ppm. A single 440Hz tone then could only deviate +/- 0.044Hz- inaudible.
It now becomes clear that specifications like 2ppm accuracy do not add to sound quality. A TCXO is such clock that achieves low ppm values, as it is very constant over long time. Fine to have on board of an accurate frequency counter, but no use for audio.
Short term
The short-term stability is important! Cycle to cycle deviations, or cycle to 100.000th cycle. If you look at the spectral content of such signal, it should look like a single frequency, with the noise floor as low as possible
(as all periods T are equally long, this translates in a single frequency 1/T). Practical clocks do not show that very narrow single frequency, but a bit wider spectrum, rapidly decreasing at 10 to 100Hz from the central frequency. A well-known clock supplier tried to convince us by showing this graph:
What we actually see is the performance of the analysers’ video filter, namely 1 kHz. Good clock spectra are well withing this window. The only clock performance we can deduce from this measurement was the one that didn’t count for audio: It meets the redbook spec. It does not meet the manufacturers own spec of 5ppm though .........
Read the whole article here.
http://www.tentlabs.com/News/assets/Tentlabs newsletter September 2006.pdf
Thank you Menno. This is the point I was trying to make - but Guido does it better.
Accuracy is not relevant to jitter.
Precision is most important.
From a scientific, analytical and statistical point of view - it is obviously desirable to have good accuracy with a high level of precision. However, Accuracy and Precision are not intrinsically linked. As mentioned previously, a series of measurements (or the time interval between pulses from a clock) can be very tightly grouped with almost identical values and a low standard deviation. This indicates a high level of precision and good reproducibility. This is what the best clocks are trying to achieve and results in lower jitter.
Despite this high level of precision, it tells us nothing about the accuracy of the measurement or result (how close this tightly grouped series of measurements is to the "true" or specified value). Fortunately - for audio - this accuracy is not too important.
So - to restate my original point.....it is possible to have:
1. Good accuracy with good precision.
2. Good accuracy with poor precision.
3. Poor accuracy with good precision.
4. Poor accuracy with poor precision.
I hope this helps to clear up any confusion between Accuracy and Precision.
.........................................
Menno Keizer said:
Accuracy is not relevant to jitter.
Menno Keizer said:
Precision is most important.
From a scientific, analytical and statistical point of view - it is obviously desirable to have good accuracy with a high level of precision. However, Accuracy and Precision are not intrinsically linked. As mentioned previously, a series of measurements (or the time interval between pulses from a clock) can be very tightly grouped with almost identical values and a low standard deviation. This indicates a high level of precision and good reproducibility. This is what the best clocks are trying to achieve and results in lower jitter.
Despite this high level of precision, it tells us nothing about the accuracy of the measurement or result (how close this tightly grouped series of measurements is to the "true" or specified value). Fortunately - for audio - this accuracy is not too important.
So - to restate my original point.....it is possible to have:
1. Good accuracy with good precision.
2. Good accuracy with poor precision.
3. Poor accuracy with good precision.
4. Poor accuracy with poor precision.
I hope this helps to clear up any confusion between Accuracy and Precision.
.........................................Jitter in Playback systems
Just a few additional comments to the already good explanations provided.
First: Jitter consists of two components: deterministic Dj and random Rj. When we quote a Dj term it defines the worst case jitter measured over an arbitrarily long time interval. Rj is different. It behaves according to something called Gaussian statistics. What this means is that, there is no maximum limit to the jitter if we measured it for a sufficientlly long enough interval. In real systems, however, it is convenient to assign a bit error rate (BER) and use this value (typically 1E-12) as a limit. At 1E-12, there is a multiplicative factor for random jitter of +/-7. For example, if we quoted an RMS Rj of 1.0 ps, then we would expect that the Rj could reach + or - 7 ps approximately once in every 1E+12 samples. therefore, a complete jitter specification needs to include both Rj and Dj components.
Second: We only care about those jitter components that contribute to audible distortion. That usually means that they reference down to the 20 Hz - 20 KHz range. Consequently, when evaluating clock jitter, it is necessary to know how the clock is divided down and used by the various elements on the D/A playback chain. Some jitter may not contribute to audible distortion, and it can be ignored. Since DACs, digital filters, etc. do not all use the same division ratios or clock ratios, it is important to evaluate each architecture on a case by case basis.
In most CD playback architectures, the incoming biphase data from the laser detector is re-timed by a PLL circuit, and the clock derived from the PLL is used in subsequent states.
Therefore, there are two key components in the clock chain: the reference clock and the PLL. Both components need to be optimized for low jitter. It is quite possible to buy a very low jitter reference clock (TCXO, for example) and still have unacceptable distortion if the PLL is poor. Most PLL manufacturers are careful to minimize jitter, but it still pays to check and compare the specified jitter values.
Just a few additional comments to the already good explanations provided.
First: Jitter consists of two components: deterministic Dj and random Rj. When we quote a Dj term it defines the worst case jitter measured over an arbitrarily long time interval. Rj is different. It behaves according to something called Gaussian statistics. What this means is that, there is no maximum limit to the jitter if we measured it for a sufficientlly long enough interval. In real systems, however, it is convenient to assign a bit error rate (BER) and use this value (typically 1E-12) as a limit. At 1E-12, there is a multiplicative factor for random jitter of +/-7. For example, if we quoted an RMS Rj of 1.0 ps, then we would expect that the Rj could reach + or - 7 ps approximately once in every 1E+12 samples. therefore, a complete jitter specification needs to include both Rj and Dj components.
Second: We only care about those jitter components that contribute to audible distortion. That usually means that they reference down to the 20 Hz - 20 KHz range. Consequently, when evaluating clock jitter, it is necessary to know how the clock is divided down and used by the various elements on the D/A playback chain. Some jitter may not contribute to audible distortion, and it can be ignored. Since DACs, digital filters, etc. do not all use the same division ratios or clock ratios, it is important to evaluate each architecture on a case by case basis.
In most CD playback architectures, the incoming biphase data from the laser detector is re-timed by a PLL circuit, and the clock derived from the PLL is used in subsequent states.
Therefore, there are two key components in the clock chain: the reference clock and the PLL. Both components need to be optimized for low jitter. It is quite possible to buy a very low jitter reference clock (TCXO, for example) and still have unacceptable distortion if the PLL is poor. Most PLL manufacturers are careful to minimize jitter, but it still pays to check and compare the specified jitter values.
rfbrw said:
It is amazing how many "supposedly" intelligent and educated self proclaimed experts do not understand and cannot grasp simple concepts. Your attempt at humor simply puts you in that category...and I suppose that is funny....hahaha.
Surely it would be more correct to say:
"Cue po-faced lecture on geographical accuracy"
You still don't understand - do you?
If you do understand something (or even if you don't - but are just pretending to), why have a go at those who are trying to help others to understand. Your comments are not particularly constructive and only cause confusion for those who are only learning about these concepts. Please remember that not all members of this forum have a scientific background and that many are not native English speakers. Your claimed "humor" would easily be lost on them.
Guys, grow up.- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.