Inaudable jitter?

[Sawzall]

As I continue to hack at the XM receiver boxes, I ran across the Application Data Sheets for the early second/ late first generation chipset as built by STMicro for XM satellite digital radio. They are fresh off the press – Sept. 2003. These are a first as they have never been public before. One chip is basically the demod and demux section and the other chip is a DSP with two cores that performs decoding and decryption – along with the kitchen sink thrown in. ST450A chip doc.

A great deal of information for potential improvements – and one sentence just jumped out at me.

An Audio PLL is embedded in STA450A.
The particularity of STA450A Audio PLL is the possibility to modify the Audio Sampling Frequency (LRCKT) in steps of a few p.p.m. to compensate dynamically the audio sampling frequency offset between the receiver and the broadcasting station; this compensation produces a jittering effect outside the audible range.
The STA450A receives from the STA400A (Channel Decoder) a dedicated signal every 432ms (PLL_SYNC) and uses this signal to perform the audio sampling rate compensation; the control is done by the DSP core updating the internal PLL registers.

So does the statement “this compensation produces a jittering effect outside the audible range” pass the sniff test? No? I would love to actually test this, but I don’t have the knowledge or the gear.

Then what would be an effective strategy to correct the problem? I have access to two digital outputs from the chip - I2S and S/PDIF. These came from pretty low bit rate signals so I am very hesitant to chop them up in anyway again.

Thoughts and /or suggestions? The first step of obtaining digital outputs significantly improved the sound (for example, cut noise in higher freq. by 6dB) and has proven to be a popular upgrade to the receivers, so much so that there is now a market for the modded sets. Running one from a Palm Pilot is being tested now

[GrahamnDodder]

Sawz-

This digital audio stuff is so out of my area of expertise that I hesitate to even throw this out, but something strikes me as familiar about it in my field, which is photojournalism, using hi-end digital photo equipment.

There are some digital (visual, not audio) output devices that stretch the files WAY up for outputting at large sizes - way larger than the original files should be stretched.

They insert a (relatively) random type of noise into the interpolation (or "stretching out," if you will) process of the pixels in the output process.

The result is ironic. The "noise" introduced into the rez'ing up makes the final stretched image look much better.

I may be way off base here, but they may be doing a similar process to make a lower-speed sampled, more compressed audio bitstream sound better than it actually should at that compression/transmission rate. Think of it as a quasi-random mask "overlaid" on what would be digital artifacting. The fact that the jitter is (supposedly) inaudible woud be immaterial to it's effectiveness.

The purposely induced jitter could be to the audio what the induced noise it to the visual image.

Am I crazy?

GnD

[Sawzall]

But for some reason our ears can hear, we "know" we hear it, but it is lost as we try to express it - I think the classic example is color vs sound. If I say red, you know what I mean in about 100 different languages. We don't have the ability to describe it in the sound world. We end up using other metaphores like warm and cold

Anyway, it does not look like I getting help on this one yet. My next test is to take two receivers and track the signal, one trace per radio. Ideally, there should be zero difference. Both radios are locked to the GPS clock at XM - and this should follow thru all the way to the DAC. Any difference should be the error. These is probably a better way, but its a start.

Now that XM has over a million subscribers, the interest in POOGEing the receivers should grow. That is a lot of radios that people will want to play with.. Ed tells me his digital output card is selling like hotcakes.

[A 8]

Controlled jitter -->>dither?