Does LPCM over HDMI suffer from Jitter?
Posted 01-01-2009 at 6:17 PM by welwynnick
My long-held view is that digital audio consists of two information streams - amplitude info and timing info. These are often referred to as data and clock respectively, though I prefer to avoid the latter. Amplitude info is simply the ones and zeros, the number that gives you the sample value at a point in time. This info is digital, and I have always considered it to be robust (but I wait to hear from Andy). Timing info is what gives you those precise points in time. This is analogue information, and is not robust.
Timing info originates in the master audio clock, and is used to synchronise the transport and the DAC. The DAC needs the timing info to be as accurate as possible, while wet string will do for the transport. So in any digital audio system, its the quality of the audio clock, and how the timng info gets to the DAC, that affects timing info. And it is degradation of the accuracy of the timing info that is jitter. Specifically, the accuracy of the word clock, which sets the point in time where the DAC should re-create the new amplitude sample.
I have generally taken jitter to be the crucial second element of digital audio, but I think there may be other contributions to degradation. If the transport has a lot of noise or interference from power, control, display, decoding or video circuits, then these may be coupled onto the digital audio output. I expect noise and interference are unlikely to impair the amplitude info/data, but since the timing info is analogue, this is vulnerable. Where you have an spdif connection, noise and interference may corrupt the timing info directly, but it may also be carried on the ground line. This may import noise and interference into the DAC, amp or processor, which will presumably have some level of susceptibility. Toslink obviously doesn't suffer from ground noise (unless its coupled to the signal at the sender), and HDMI is well protected by differential screened twisted pairs.
Jitter may not be the only mechanism to degrade digital audio, but I suppose I group noise and interference in the same category, even if they don't come down the signal channel.
I think the reason why bitstream may sound better (processing faults notwithstanding) is down to the digital audio architecture, and how it gets the the timing info from the master audio clock to the DAC. Bitstream itself doesn't carry a clock, but if the timing info is generated in the amp, then it doesn't have to make its way over from the player, which is the best way to minimise jitter (or whatever corruption of the timing info).
HOWEVER, even when bitstream is carried over HDMI, the video clock is still carried from the player to the amp. And the digital audio architectures described and illustrated in the HDMI spec have the audio clock being recovered from the video clock. If this is a representative architecture, then it doesn't matter if the data channels are carrying LPCm or bistream, because NEITHER are carrying the clock anyway. That comes over the clock channel irrespective.
Therefore, if the amplifier uses the same process to generate the audio clock, then bitstream or LPCM won't make any difference to audio jitter. The bitstream process will therefore be a transparent processing layer in the data domain that simply results in the same data emerging at the receiving end, and the "bits are bits" boys will be right. Bitstream will sound the same (IF the data itself is the same of course, which is the biggest of "ifs").
However, that does assume that the crucial digital audio replay architecture is the same for bitstream as for LPCM (have I said that enough times yet?) . If the amplifier ignores the video clock, and uses a similar process to generate the clock as bitstream over spdif for example, then the architecture is different, and bitstreamed audio will not suffer from the same timing info degradation as LPCM would over HDMI. Therefore in this case, there IS a way in which bitstream could sound better than LPCM.
(This reasoning does suggest though, that both the lossless bistreamed codecs should sound the same, and any differences are likely to lie in the mastering rather than the processing.)
So, it all depends on how the amplifier implements the audio clock recovery.
Timing info originates in the master audio clock, and is used to synchronise the transport and the DAC. The DAC needs the timing info to be as accurate as possible, while wet string will do for the transport. So in any digital audio system, its the quality of the audio clock, and how the timng info gets to the DAC, that affects timing info. And it is degradation of the accuracy of the timing info that is jitter. Specifically, the accuracy of the word clock, which sets the point in time where the DAC should re-create the new amplitude sample.
I have generally taken jitter to be the crucial second element of digital audio, but I think there may be other contributions to degradation. If the transport has a lot of noise or interference from power, control, display, decoding or video circuits, then these may be coupled onto the digital audio output. I expect noise and interference are unlikely to impair the amplitude info/data, but since the timing info is analogue, this is vulnerable. Where you have an spdif connection, noise and interference may corrupt the timing info directly, but it may also be carried on the ground line. This may import noise and interference into the DAC, amp or processor, which will presumably have some level of susceptibility. Toslink obviously doesn't suffer from ground noise (unless its coupled to the signal at the sender), and HDMI is well protected by differential screened twisted pairs.
Jitter may not be the only mechanism to degrade digital audio, but I suppose I group noise and interference in the same category, even if they don't come down the signal channel.
I think the reason why bitstream may sound better (processing faults notwithstanding) is down to the digital audio architecture, and how it gets the the timing info from the master audio clock to the DAC. Bitstream itself doesn't carry a clock, but if the timing info is generated in the amp, then it doesn't have to make its way over from the player, which is the best way to minimise jitter (or whatever corruption of the timing info).
HOWEVER, even when bitstream is carried over HDMI, the video clock is still carried from the player to the amp. And the digital audio architectures described and illustrated in the HDMI spec have the audio clock being recovered from the video clock. If this is a representative architecture, then it doesn't matter if the data channels are carrying LPCm or bistream, because NEITHER are carrying the clock anyway. That comes over the clock channel irrespective.
Therefore, if the amplifier uses the same process to generate the audio clock, then bitstream or LPCM won't make any difference to audio jitter. The bitstream process will therefore be a transparent processing layer in the data domain that simply results in the same data emerging at the receiving end, and the "bits are bits" boys will be right. Bitstream will sound the same (IF the data itself is the same of course, which is the biggest of "ifs").
However, that does assume that the crucial digital audio replay architecture is the same for bitstream as for LPCM (have I said that enough times yet?) . If the amplifier ignores the video clock, and uses a similar process to generate the clock as bitstream over spdif for example, then the architecture is different, and bitstreamed audio will not suffer from the same timing info degradation as LPCM would over HDMI. Therefore in this case, there IS a way in which bitstream could sound better than LPCM.
(This reasoning does suggest though, that both the lossless bistreamed codecs should sound the same, and any differences are likely to lie in the mastering rather than the processing.)
So, it all depends on how the amplifier implements the audio clock recovery.
Total Comments 0
