Jitter Spectrum Matters...
Hi Guido and Jocko,
Isn't the jitter that passes through according to the capture and track characteristics of the recieve PLL and its control of the FIFO data buffer stage ? - ie PLL feedback characteristics.
Or are you saying that incoming data timing variation modulates the recieve PLL, and it is this modulation that is audible ?.
I contend that jitter can be sonically allowable, so long as the nature (spectrum) of the jitter at the input of the DAC stage is sonically acceptable, and the amplitude of the timing variation is second in importance if it cannot be eliminated.
I agree that less jitter = better reproduced sound, but even if less, the spectral nature must be sonically acceptable too.
I have tried experiments that reduce the jitter on SPDIF transmission systems somewhat, but more importantly alter the jitter nature to one that is entirely acceptable - a bit like second order/third order dominant amplifiers sounding different, only in this case changing the IMD characteristics.
This has worked very nicely on some high-end Jap AV gear.
Eric.
Hi Guido and Jocko,
Isn't the jitter that passes through according to the capture and track characteristics of the recieve PLL and its control of the FIFO data buffer stage ? - ie PLL feedback characteristics.
Or are you saying that incoming data timing variation modulates the recieve PLL, and it is this modulation that is audible ?.
I contend that jitter can be sonically allowable, so long as the nature (spectrum) of the jitter at the input of the DAC stage is sonically acceptable, and the amplitude of the timing variation is second in importance if it cannot be eliminated.
I agree that less jitter = better reproduced sound, but even if less, the spectral nature must be sonically acceptable too.
I have tried experiments that reduce the jitter on SPDIF transmission systems somewhat, but more importantly alter the jitter nature to one that is entirely acceptable - a bit like second order/third order dominant amplifiers sounding different, only in this case changing the IMD characteristics.
This has worked very nicely on some high-end Jap AV gear.
Eric.
Re: Jitter Spectrum Matters...
Hi Eric,
Within (below) the PLL BW all jitter comes through, unattenuated. Above the PLL cutoff, all jitter is attenuated.
In my own DAC I have a 3rd order PLL, with dominant pole at about 1 Hz, as low as possible.
It is obvious that phase modulation of the clock used for the conversion is audible. We have caried out many tests to obtain audible threshold levels as function of amplitude and frequency.
The CD players that I modify ofcourse have lower jitter at their outputs than before.
Whatever best you do to attenuate the outgoing jitter, it still caries data dependancy.
Modern input receivers like the 8415 do recover the clock based on the preamble, to avoid data dependant jitter.
Do realise that levels far below 1 ps (rms, 3 sigma) are still audible, at least if fed to PCM63K.
We did test front ends of DAs equiped with fifo and free rnning oscillators to clock the data and the converter. Still, the difference nbetween drives feeding that DAC are audible, though to a somewhat lesser extend than with my own (PLL) DAC.
The jitter still ripples through, I guess by means of the data containing jitter. This still arives on the PCM63 chip, and deteroriates the on-board clock.
This was one of the reasons t apply reclocking of all signals entering the DACs, though still not enough.......
Our DAC design can be found at
http://members.chello.nl/~m.heijligers/DAChtml/dactop.htm
I agree with you on the spectral content, it is not only amplitude that counts. This is an area still to explore. The sonical impact ofcourse depends on the type of DAC.
Needless to say that SACD is sensitive to clock jitter as well.
best regards
mrfeedback said:
Hi Eric,
Within (below) the PLL BW all jitter comes through, unattenuated. Above the PLL cutoff, all jitter is attenuated.
In my own DAC I have a 3rd order PLL, with dominant pole at about 1 Hz, as low as possible.
It is obvious that phase modulation of the clock used for the conversion is audible. We have caried out many tests to obtain audible threshold levels as function of amplitude and frequency.
The CD players that I modify ofcourse have lower jitter at their outputs than before.
Whatever best you do to attenuate the outgoing jitter, it still caries data dependancy.
Modern input receivers like the 8415 do recover the clock based on the preamble, to avoid data dependant jitter.
Do realise that levels far below 1 ps (rms, 3 sigma) are still audible, at least if fed to PCM63K.
We did test front ends of DAs equiped with fifo and free rnning oscillators to clock the data and the converter. Still, the difference nbetween drives feeding that DAC are audible, though to a somewhat lesser extend than with my own (PLL) DAC.
The jitter still ripples through, I guess by means of the data containing jitter. This still arives on the PCM63 chip, and deteroriates the on-board clock.
This was one of the reasons t apply reclocking of all signals entering the DACs, though still not enough.......
Our DAC design can be found at
http://members.chello.nl/~m.heijligers/DAChtml/dactop.htm
I agree with you on the spectral content, it is not only amplitude that counts. This is an area still to explore. The sonical impact ofcourse depends on the type of DAC.
Needless to say that SACD is sensitive to clock jitter as well.
best regards
Hi Guido, thanks for your info.
Is what you saying indicating that the FIFO register is not long enough ?.
I mean if the FIFO is much longer, then the corner PLL can have a much longer settling time.
Have you heard CDP with anti-shock memory ?.
The ones that I have heard are rather good, in that the data rate (off the disc) modulation is pretty much eliminated, and any instability is more so only coming from the master clock.
A long FIFO I feel is required for really stable playback speed.
Do you agree that some older mastering ADC were very bad for introducing jitter into the digital recording ?.
Eric.
Is what you saying indicating that the FIFO register is not long enough ?.
I mean if the FIFO is much longer, then the corner PLL can have a much longer settling time.
Have you heard CDP with anti-shock memory ?.
The ones that I have heard are rather good, in that the data rate (off the disc) modulation is pretty much eliminated, and any instability is more so only coming from the master clock.
A long FIFO I feel is required for really stable playback speed.
Do you agree that some older mastering ADC were very bad for introducing jitter into the digital recording ?.
Eric.
mrfeedback said:
Hi Eric
As long as the Fifo stays somehwere between just empty, and nearly full, it is big enough. A decently designed spindle motor speed algoritm will take care of that.
The problem lies within the crosstalk, not within Fifo size.
Any CDP has a memory, in order to integrate data rate differences down to zero. If you suffer from data rate differences due to disc speed modulation, a serious design error is within the product: Bring it back to the shop and friendly ask your money back.
Jitter at the process of AD conversion can never be corrected for: Many early digital stuff sounds crappy for that reason.
best regards,
Guido Tent said:
Hi all,
I measured today. I took a normal wire, twisted it quite tightly, and came to 100 ohms.....
Then i lossened the twist, up to a level that say 2 or 3 mm hllow space appeared within the (still) twisted wires, I then arrived a some 200 ohms.
I guess I can get down to lower impedances if I use wire with thinner isolation.
best regards
transmission lines
Mr. Johnson uses a natural log function but the actual relationship is the inverse hyperbolic cosine of the (wire with insulation diameter divide by the conductor diameter) I don't know how accurate the ln function is as s approaches d. The permittivity is also simplified. The impedance is a function of the square root of the effective permittivity. This is a function of angle of the twist and the dielectric constant of the wire insulation. The impedance value changes very quickly as the insulation becomes thin compared to the wire diameter. Small variations in insulation thickness and how tight the wire is wound will give fairly large variations in characteristic impedance a long the wire's length. You are going to wind up with a poor transmission line. Getting any consistency for it's impedance is very difficult. I would not recommend twisted pair for even a short run of the SPDIF interface.
http://mathworld.wolfram.com/InverseHyperbolicCosine.html
http://www.webcalc.net/calc/0066.php
Transmission Line Design Handbook (Artech House Microwave Library)
by Brian C. Wadell
http://www.amazon.com/exec/obidos/t...f=sr_1_2/002-0728767-7597654?v=glance&s=books
Transmission Line Design and Analysis website
http://www.circuitsage.com/tline.html
Mr. Johnson uses a natural log function but the actual relationship is the inverse hyperbolic cosine of the (wire with insulation diameter divide by the conductor diameter) I don't know how accurate the ln function is as s approaches d. The permittivity is also simplified. The impedance is a function of the square root of the effective permittivity. This is a function of angle of the twist and the dielectric constant of the wire insulation. The impedance value changes very quickly as the insulation becomes thin compared to the wire diameter. Small variations in insulation thickness and how tight the wire is wound will give fairly large variations in characteristic impedance a long the wire's length. You are going to wind up with a poor transmission line. Getting any consistency for it's impedance is very difficult. I would not recommend twisted pair for even a short run of the SPDIF interface.
http://mathworld.wolfram.com/InverseHyperbolicCosine.html
http://www.webcalc.net/calc/0066.php
Transmission Line Design Handbook (Artech House Microwave Library)
by Brian C. Wadell
http://www.amazon.com/exec/obidos/t...f=sr_1_2/002-0728767-7597654?v=glance&s=books
Transmission Line Design and Analysis website
http://www.circuitsage.com/tline.html
Fred,
Great math links. Ive book marked them. They should occasionally help an armchair engineer like myself be a little smarter.
Are you a Mathmatica user? Ive dabbled w/ Matlab, but can’t seem to come up with a compelling need for it.
Guido,
I'm curious about how you are making these interesting tests. Do you have an AP? I was wondering if you could share how you made these impedance and jitter measurements?
Great math links. Ive book marked them. They should occasionally help an armchair engineer like myself be a little smarter.
Are you a Mathmatica user? Ive dabbled w/ Matlab, but can’t seem to come up with a compelling need for it.
Guido,
I'm curious about how you are making these interesting tests. Do you have an AP? I was wondering if you could share how you made these impedance and jitter measurements?
Da5id4Vz said:
Hi,
I use twisted pair within one cabinet only, outside world is too cruel.
I have acces to a Wavecrest 3000, which is in my eyes the only tool able to do proper jitter measurements.
I do have an AP one at home, have yet to set it up and automate the whole. Time is too limitting now.
best regards
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.