Maker Pro
Maker Pro

External clock for Analog to Digital Converter

E

East Hunk

Hi Everyone,
I am working on Jitter in Analog to Digital Converters(ADCs). I am
trying to setup an experiment to see the effects of jitter in real
time ADCs. I have an ADC evaluation board with external clock input
for sampling (i.e Sampling clock).
Now I want to produce a self created jittery signal to see the effects
of jitter in ADC. Can anyone out there have any idea(s) how I can
produce a real time clock signal with "variable jitter" for the input
of external clock? Producing a simple clock signal (i.e. without
jitter) is straight forward with the help of any signal generator but
a clock signal with "variable jitter" is a problem.

Thanks for your help.

Cheers

Bilal
 
E

Eeyore

East said:
Hi Everyone,
I am working on Jitter in Analog to Digital Converters(ADCs). I am
trying to setup an experiment to see the effects of jitter in real
time ADCs. I have an ADC evaluation board with external clock input
for sampling (i.e Sampling clock).
Now I want to produce a self created jittery signal to see the effects
of jitter in ADC. Can anyone out there have any idea(s) how I can
produce a real time clock signal with "variable jitter" for the input
of external clock? Producing a simple clock signal (i.e. without
jitter) is straight forward with the help of any signal generator but
a clock signal with "variable jitter" is a problem.

Add noise. It's an analogue issue effectively.

Graham
 
J

John Larkin

Hi Everyone,
I am working on Jitter in Analog to Digital Converters(ADCs). I am
trying to setup an experiment to see the effects of jitter in real
time ADCs. I have an ADC evaluation board with external clock input
for sampling (i.e Sampling clock).
Now I want to produce a self created jittery signal to see the effects
of jitter in ADC. Can anyone out there have any idea(s) how I can
produce a real time clock signal with "variable jitter" for the input
of external clock? Producing a simple clock signal (i.e. without
jitter) is straight forward with the help of any signal generator but
a clock signal with "variable jitter" is a problem.

Thanks for your help.

Cheers

Bilal

Get a fast analog comparator or a differential-LVDS-to-CMOS converter
chip.

Apply your low-jitter logic-level signal generator to the + input of
the comparator. Bias the - input to about logic mid-swing and AC
couple a noise source into that. The output clocks your adc, and the
jitter will depend on the slew rate of the edge at the + input and the
ac noise level at the - input; the math here is direct. The noise
source can be any waveform you like and will determine the probability
distribution of the jitter; random, gaussian noise would be the
typical choice, or use a triangle wave for a flat probability
distribution.

To get lots of jitter, you may have to reduce the slew rate of the
clock generator. Some pulse generators have adjustable slew rate, but
if yours doesn't, a simple r-l-c lowpass filter can control edge rate
pretty accurately.

Some of the adc's around these days need femtosecond RMS clock jitter
to meet their accuracy specs.

John
 
E

Eeyore

John said:
Get a fast analog comparator or a differential-LVDS-to-CMOS converter
chip.

Apply your low-jitter logic-level signal generator to the + input of
the comparator. Bias the - input to about logic mid-swing and AC
couple a noise source into that. The output clocks your adc, and the
jitter will depend on the slew rate of the edge at the + input and the
ac noise level at the - input; the math here is direct. The noise
source can be any waveform you like and will determine the probability
distribution of the jitter; random, gaussian noise would be the
typical choice, or use a triangle wave for a flat probability
distribution.

To get lots of jitter, you may have to reduce the slew rate of the
clock generator. Some pulse generators have adjustable slew rate, but
if yours doesn't, a simple r-l-c lowpass filter can control edge rate
pretty accurately.

Some of the adc's around these days need femtosecond RMS clock jitter
to meet their accuracy specs.

Nice detailed explanation there John.

It's useful to note that most audio converters now simply require wordclock and
generate the bit clock from an internal PPL.

As such there's not much you can do to influence their performance.

Graham
 
L

Lionel

Hi Everyone,
I am working on Jitter in Analog to Digital Converters(ADCs). I am
trying to setup an experiment to see the effects of jitter in real
time ADCs. I have an ADC evaluation board with external clock input
for sampling (i.e Sampling clock).
Now I want to produce a self created jittery signal to see the effects
of jitter in ADC. Can anyone out there have any idea(s) how I can
produce a real time clock signal with "variable jitter" for the input
of external clock? Producing a simple clock signal (i.e. without
jitter) is straight forward with the help of any signal generator but
a clock signal with "variable jitter" is a problem.

What sort of clock frequency does your ADC need, how much jitter do
you want, & how precisely to you need to control it?
 
J

John Larkin

Nice detailed explanation there John.

It's useful to note that most audio converters now simply require wordclock and
generate the bit clock from an internal PPL.

As such there's not much you can do to influence their performance.

For audio, I wouldn't think that even a microsecond of RMS jitter
would be audible.

John
 
M

martin griffith

For audio, I wouldn't think that even a microsecond of RMS jitter
would be audible.

John
It does, it alters the noise, AFAIR, from a paper from Crystal/Cirrus,
many years ago, which I cant find, and a AES paper by Julian Dunn,
1992. Basicallyt, it depends on modulation frequency, and at about 1K,
needs to be less than 1nS to be inaudible, and less than 100pS as you
go higher.

But in these days after a nice linear 24Bit signal has been mangled by
bean counting software, I suspected 1uS may not be far from the truth



martin
 
J

John Larkin

It does, it alters the noise, AFAIR, from a paper from Crystal/Cirrus,
many years ago, which I cant find, and a AES paper by Julian Dunn,
1992. Basicallyt, it depends on modulation frequency, and at about 1K,
needs to be less than 1nS to be inaudible, and less than 100pS as you
go higher.

I don't understand that. Sound moves about 340 m/s, or 3 milliseconds
per meter. 3 microseconds per mm. 3 ns per micron. Think about a
microphone on a stand, sitting on a stage with a thousand people
walking around, on a street with cars and trucks, in a building with
people and vending machines and hvac, various musical instruments
making enough noise to damage your hearing... what's the jitter
introduced by all that?

And listening: is your speaker cabinet solid granite, sitting on
bedrock? Are you sure your woofer has no effect on the position of the
cone of your midrange? Is your head in a clamp? Do you stop your heart
and breathing to listen? What do you do about air currents and
thermals in the room, modulating sound velocity all over the place?

Picoseconds are absurd here.

John
 
M

martin griffith

On Tue, 27 Feb 2007 11:07:13 -0800, in sci.electronics.design John
Larkin said:
I don't understand that. Sound moves about 340 m/s, or 3 milliseconds
per meter. 3 microseconds per mm. 3 ns per micron. Think about a
microphone on a stand, sitting on a stage with a thousand people
walking around, on a street with cars and trucks, in a building with
people and vending machines and hvac, various musical instruments
making enough noise to damage your hearing... what's the jitter
introduced by all that?

And listening: is your speaker cabinet solid granite, sitting on
bedrock? Are you sure your woofer has no effect on the position of the
cone of your midrange? Is your head in a clamp? Do you stop your heart
and breathing to listen? What do you do about air currents and
thermals in the room, modulating sound velocity all over the place?

Picoseconds are absurd here.

John
Not really, I think that you are just looking at it from the wrong
direction
I've dumped a couple of papers:
http://es.geocities.com/mart_in_medina/JITTER.PDF
http://es.geocities.com/mart_in_medina/JITTER92.PDF
the Steven Harris, should be on the cirrus site, but it's down for
repairs


martin
 
J

John Larkin

On Tue, 27 Feb 2007 11:07:13 -0800, in sci.electronics.design John

Not really, I think that you are just looking at it from the wrong
direction
I've dumped a couple of papers:

Silly. One of the things he doesn't note is that adc or dac clock
jitter doesn't create signal energy, it just splatters signal around
the spectrum. So 8 ns of jitter may raise the noise floor to, say, 84
dB below a 0 dB signal, but the jitter adds no noise when there's no
signal. Who can hear -84 dB of noise *when*the*main*signal*is*0*dB ?

Jitter doesn't add to the "quiet" background, which is the one that
might conceivably be audible. As long as nobody in your listening room
is, say, breathing.

I can't load that one for some reason. I'm guessing it must be equally
silly.

John
 
M

martin griffith

Silly. One of the things he doesn't note is that adc or dac clock
jitter doesn't create signal energy, it just splatters signal around
the spectrum. So 8 ns of jitter may raise the noise floor to, say, 84
dB below a 0 dB signal, but the jitter adds no noise when there's no
signal. Who can hear -84 dB of noise *when*the*main*signal*is*0*dB ?

Jitter doesn't add to the "quiet" background, which is the one that
might conceivably be audible. As long as nobody in your listening room
is, say, breathing.


I can't load that one for some reason. I'm guessing it must be equally
silly.

John
Who can hear -84 dB of noise *when*the*main*signal*is*0*dB

Not many people record at FS, normally there is 12dB headroom, thats
standard practice, so that's now down to -72dB, and if you are mixing
many sources the noise just tends to add up, so 8nS is not
unreasonable.

Sorry about the second file, try this
http://es.geocities.com/mart_in_medina/Jitter92.PDF
I've been trying to find ref(8), but its's not freely available on
line


martin
 
E

Eeyore

martin said:
Not many people record at FS, normally there is 12dB headroom, thats
standard practice, so that's now down to -72dB, and if you are mixing
many sources the noise just tends to add up, so 8nS is not
unreasonable.

Sorry about the second file, try this
http://es.geocities.com/mart_in_medina/Jitter92.PDF
I've been trying to find ref(8), but its's not freely available on
line

Fancy that. I've worked with Julian in the past.

Graham
 
E

East Hunk

Get a fast analog comparator or a differential-LVDS-to-CMOS converter
chip.

Apply your low-jitter logic-level signal generator to the + input of
the comparator. Bias the - input to about logic mid-swing and AC
couple a noise source into that. The output clocks your adc, and the
jitter will depend on the slew rate of the edge at the + input and the
ac noise level at the - input; the math here is direct. The noise
source can be any waveform you like and will determine the probability
distribution of the jitter; random, gaussian noise would be the
typical choice, or use a triangle wave for a flat probability
distribution.

To get lots of jitter, you may have to reduce the slew rate of the
clock generator. Some pulse generators have adjustable slew rate, but
if yours doesn't, a simple r-l-c lowpass filter can control edge rate
pretty accurately.

Some of the adc's around these days need femtosecond RMS clock jitter
to meet their accuracy specs.

John.

Hi John,

Thank you for your reply. This is indeed an excellent idea. Let me try
this and then I will get back to you.

Cheers

Bilal
 
I

Ian

martin griffith said:
On Tue, 27 Feb 2007 19:50:25 -0800, in sci.electronics.design John


Not many people record at FS, normally there is 12dB headroom, thats
standard practice, so that's now down to -72dB, and if you are mixing
many sources the noise just tends to add up, so 8nS is not
unreasonable.

Sorry about the second file, try this
http://es.geocities.com/mart_in_medina/Jitter92.PDF
I've been trying to find ref(8), but its's not freely available on
line


martin

I see that's a preprint, did the paper ever get into the Journal?

Regards
Ian
 
J

John Larkin

Not many people record at FS, normally there is 12dB headroom, thats
standard practice, so that's now down to -72dB, and if you are mixing
many sources the noise just tends to add up, so 8nS is not
unreasonable.


No. This noise is a spectral scattering of the main signal. As the
program level declines, the jitter-induced noise floor tracks it down.
If the jitter-caused noise is 84 dB below 0 dBm, it's also 84 dB below
-40 dBm, namely -124. No signal, no noise.

This isn't like hum or thermal noise; they are constant adders to the
signal. This is more like distortion, a *multiplier*.

The only time moderate sampling jitter introduces noise is when you
can't hear it.

John
 
M

martin griffith

No. This noise is a spectral scattering of the main signal. As the
program level declines, the jitter-induced noise floor tracks it down.
If the jitter-caused noise is 84 dB below 0 dBm, it's also 84 dB below
-40 dBm, namely -124. No signal, no noise.

This isn't like hum or thermal noise; they are constant adders to the
signal. This is more like distortion, a *multiplier*.

The only time moderate sampling jitter introduces noise is when you
can't hear it.

John

Thanks for that explanation, I see what you mean now


martin
 
M

martin griffith

I see that's a preprint, did the paper ever get into the Journal?

Regards
Ian
No idea, gave up the AES mag years ago, it was much was too esoteric,
I spent all my time keeping 6 sound studios + all the other bits
running. Might be worth a search through AES site


martin
 
J

joseph2k

martin said:
It does, it alters the noise, AFAIR, from a paper from Crystal/Cirrus,
many years ago, which I cant find, and a AES paper by Julian Dunn,
1992. Basicallyt, it depends on modulation frequency, and at about 1K,
needs to be less than 1nS to be inaudible, and less than 100pS as you
go higher.

But in these days after a nice linear 24Bit signal has been mangled by
bean counting software, I suspected 1uS may not be far from the truth



martin

I really doubt the jitter is that bad, 10% of (clock) conversion rate in
jitter is a really atrocious clock. Few reasonably appropriate clock
generetors have over 0.01% jitter, which would be about 1 ns for audio
work. Soounds like one of the differences between home audio and pro
audio.
 
Top