What's up with SRC's

A 8 · 9th April 2003, 06:06 PM

Built a 24 bit dac using the CS8420 and there is something strange going on......
On some cd's something happens in the upper mid range/lower high range, it gets kind of grainy or perhaps modulated.

Lower midrange, midrange and high range are always extremely soft, detailed and very clean.
Low level resolution on any material (including the ones that I have a problem with) is as good as I ever heard but still this issue on probably 5% of the material I have available.......
I assume this is some form of jitter and consulting the CS 8420 data sheet it don't go into any detail as to how it handles the bit and rate conversion???
Is it just a FIFO that changes the word length to accommodate a 24 bit word length...how does it then go about calculating the change in time base i.e. whatever ->96kHz???

Anyone know??

Jocko Homo · 10th April 2003, 01:18 AM

That's exactly what happens if you do not eliminate it.

Jocko

A 8 · 10th April 2003, 08:27 AM

Well...that's what I figured, still need to figure out where and how to kill it

I have read Bob Katz article on jitter but it excludes src's in the dsp context so I am getting worried that any jitter present in the transmission will be part of the data going out of the src?

To figure it out I would really appreciate if anyone could explain or point me to articles etc. on how the src handles the word integrity during resampling and word-length change.

Thx

BTW I run the dac chip directly from the clock source and even though I could get a clock with lower jitter this one is probably many times better then the cs8412 recovered clock in the dac I use to compare sonic performance.

Norman Tracy · 10th April 2003, 09:12 PM

The fact you wrote:

"On some cd's something happens in the upper mid range/lower high range, it gets kind of grainy or perhaps modulated."

Makes me wonder if your CS8420 is getting into its hardware glitch mode.

At

http://www-test.cirrus.com/en/pubs/errata/er245c2.pdf

in the "Errata: CS8420-CS Revision D Perfromance Update" Crystal engineers talk about the part getting into an "invalid mode". I have wittnesed the CS8420 in my XD0 and X-DAC 3.24 designs get into this mode both on my bench and in the listening room. In this "invalid mode" the CS8420 DSP has bad coefficients and generates notches in its output frequency response and spurious tones. On the bench one can see these as bursts of noise. They don't sound as bad as they look, but it is very much like your description of "it gets kind of grainy or perhaps modulated." A hardware reset gets the CS8420 back in correct operation. If you are running it in software mode (i.e. controlled from a microcontroller) there is a software workaround. There is no fix in hardware mode.

No device is perfect, despite this glitch I am still a huge fan of the CS8420. The upsampling + jitter reduction it offers has a big impact on getting our CD and DADs to SOTA hi-fi sound.

As to how SRCs work check the Analog Devices web site. Its WAY beyond the scope of a discussion board posting and AD has some good articles on how they work. The theory of operation section of:

AD1890 Data Sheet

is a good start.

A 8 · 11th April 2003, 07:11 AM

Thx,

Good thinking, I have been wondering about the invalid mode but expected it to be VERY audible if it happened.

I'll put in a manual reset and do a bit more testing.

My issue aside I agree it makes for a better listening experience, even crappy cd's from the mid 80's sound good!

Reg

A 8 · 11th April 2003, 10:09 AM

Problem gone!

I think you were right, I added a manual reset and the artifact is gone---thanks again.

BTW Have anyone tried to "overrun" the 8420/DF1704/PCM1704K on a 27-30MHz clock?
It is above max ratings on all chips but I seem to remember that MSB Tech did something like this and thought it improved sonic performace?

OliverD · 11th April 2003, 07:20 PM

Others report the PCM1704 to sound better at lower bit clocks... Guess you have to try and decide yourself

A 8 · 11th April 2003, 07:36 PM

The 24.576Mhz clock = 96kHz sampling sounds great so I was thinking that 108kHz and up would be even better

Anyone else have any experiance?

Jocko Homo · 11th April 2003, 08:57 PM

Lower clock rate would probably be less sensitive to jitter. Running all mine at 11,289.6 MHz.

Jocko

OliverD · 11th April 2003, 09:24 PM

Speaking of jitter...

In a AD8561 -> CS8420 -> DF1704 -> PCM1704 chipset, we can use a low jitter oscillator such as the Kwak-clock (as it provides both inverted and non-inverted outputs) to feed both the bit clock input of the PCM1704 and the master clock input of the CS8420... Oscillator, DAC and ASCR could be powered by three completely separated power supplies (using ISO150's for the signals)... I think this way (at least in theory) jitter in the source signal has no way to get through to the DAC, but I might be proven wrong

Until now I have not tried ASCR's and I'm not exactly excited about upsampling, but I like the "A" in ASCR and will soon give it a try.

Does anyone know: Will the CS8420 output an exact clone of the input data words for an 1:1 sampling rate ratio?

9th April 2003, 06:06 PM	#1
A 8 diyAudio Member Join Date: Mar 2003 Location: Gothenburg	What's up with SRC's Built a 24 bit dac using the CS8420 and there is something strange going on...... On some cd's something happens in the upper mid range/lower high range, it gets kind of grainy or perhaps modulated. Lower midrange, midrange and high range are always extremely soft, detailed and very clean. Low level resolution on any material (including the ones that I have a problem with) is as good as I ever heard but still this issue on probably 5% of the material I have available....... I assume this is some form of jitter and consulting the CS 8420 data sheet it don't go into any detail as to how it handles the bit and rate conversion??? Is it just a FIFO that changes the word length to accommodate a 24 bit word length...how does it then go about calculating the change in time base i.e. whatever ->96kHz??? Anyone know??

10th April 2003, 01:18 AM	#2
Jocko Homo Banned Join Date: Feb 2002 Location: As far from the NOSsers as possible	Jitter, jitter, and more jitter........ That's exactly what happens if you do not eliminate it. Jocko

10th April 2003, 09:12 PM	#4
Norman Tracy diyAudio Member Join Date: Apr 2003 Location: Tulsa OK USA	The fact you wrote: "On some cd's something happens in the upper mid range/lower high range, it gets kind of grainy or perhaps modulated." Makes me wonder if your CS8420 is getting into its hardware glitch mode. At http://www-test.cirrus.com/en/pubs/errata/er245c2.pdf in the "Errata: CS8420-CS Revision D Perfromance Update" Crystal engineers talk about the part getting into an "invalid mode". I have wittnesed the CS8420 in my XD0 and X-DAC 3.24 designs get into this mode both on my bench and in the listening room. In this "invalid mode" the CS8420 DSP has bad coefficients and generates notches in its output frequency response and spurious tones. On the bench one can see these as bursts of noise. They don't sound as bad as they look, but it is very much like your description of "it gets kind of grainy or perhaps modulated." A hardware reset gets the CS8420 back in correct operation. If you are running it in software mode (i.e. controlled from a microcontroller) there is a software workaround. There is no fix in hardware mode. No device is perfect, despite this glitch I am still a huge fan of the CS8420. The upsampling + jitter reduction it offers has a big impact on getting our CD and DADs to SOTA hi-fi sound. As to how SRCs work check the Analog Devices web site. Its WAY beyond the scope of a discussion board posting and AD has some good articles on how they work. The theory of operation section of: AD1890 Data Sheet is a good start. __________________ Norman Tracy Audio Crafters Guild

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

New To Site?	Need Help?
Register to Participate Search Privacy Statement Contact Us	Frequently Asked Questions Did you forget your password? Mark Forums Read


	Please consider donating to help us continue to serve you. Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving

10th April 2003, 08:27 AM	#3
A 8 diyAudio Member Join Date: Mar 2003 Location: Gothenburg	Well...that's what I figured, still need to figure out where and how to kill it I have read Bob Katz article on jitter but it excludes src's in the dsp context so I am getting worried that any jitter present in the transmission will be part of the data going out of the src? To figure it out I would really appreciate if anyone could explain or point me to articles etc. on how the src handles the word integrity during resampling and word-length change. Thx BTW I run the dac chip directly from the clock source and even though I could get a clock with lower jitter this one is probably many times better then the cs8412 recovered clock in the dac I use to compare sonic performance.

11th April 2003, 07:11 AM	#5
A 8 diyAudio Member Join Date: Mar 2003 Location: Gothenburg	Thx, Good thinking, I have been wondering about the invalid mode but expected it to be VERY audible if it happened. I'll put in a manual reset and do a bit more testing. My issue aside I agree it makes for a better listening experience, even crappy cd's from the mid 80's sound good! Reg

11th April 2003, 10:09 AM	#6
A 8 diyAudio Member Join Date: Mar 2003 Location: Gothenburg	Problem gone! I think you were right, I added a manual reset and the artifact is gone---thanks again. BTW Have anyone tried to "overrun" the 8420/DF1704/PCM1704K on a 27-30MHz clock? It is above max ratings on all chips but I seem to remember that MSB Tech did something like this and thought it improved sonic performace?

11th April 2003, 07:20 PM	#7
OliverD diyAudio Member Join Date: Nov 2002 Location: Germany	Others report the PCM1704 to sound better at lower bit clocks... Guess you have to try and decide yourself

11th April 2003, 07:36 PM	#8
A 8 diyAudio Member Join Date: Mar 2003 Location: Gothenburg	The 24.576Mhz clock = 96kHz sampling sounds great so I was thinking that 108kHz and up would be even better Anyone else have any experiance?

11th April 2003, 08:57 PM	#9
Jocko Homo Banned Join Date: Feb 2002 Location: As far from the NOSsers as possible	Lower clock rate would probably be less sensitive to jitter. Running all mine at 11,289.6 MHz. Jocko

11th April 2003, 09:24 PM	#10
OliverD diyAudio Member Join Date: Nov 2002 Location: Germany	Speaking of jitter... In a AD8561 -> CS8420 -> DF1704 -> PCM1704 chipset, we can use a low jitter oscillator such as the Kwak-clock (as it provides both inverted and non-inverted outputs) to feed both the bit clock input of the PCM1704 and the master clock input of the CS8420... Oscillator, DAC and ASCR could be powered by three completely separated power supplies (using ISO150's for the signals)... I think this way (at least in theory) jitter in the source signal has no way to get through to the DAC, but I might be proven wrong Until now I have not tried ASCR's and I'm not exactly excited about upsampling, but I like the "A" in ASCR and will soon give it a try. Does anyone know: Will the CS8420 output an exact clone of the input data words for an 1:1 sampling rate ratio?

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)