100 Mhz sq wave w/ >5 pSec jitter?

[[email protected]]

I have a request to measure the edge-to-edge jitter from some F100314
diff ECL line drivers. The customer wants it done with 100 MHz square
waves, and the jitter spec is > +/- 5 pSec. Does that mean I need a sq
wave with better than that jitter? Can I get one?
Or: could I trigger off the sq wave source while inputting it into the
line driver, then look at the output of the driver on the scope and
measure the time from trigger to output edge rising and get the jitter
of that time?

 

[John Larkin]

I have a request to measure the edge-to-edge jitter from some F100314
diff ECL line drivers. The customer wants it done with 100 MHz square
waves, and the jitter spec is > +/- 5 pSec. Does that mean I need a sq
wave with better than that jitter?

Can I get one?

Yup. A medium-good 100 MHz ECL xtal osc will have maybe 1 ps RMS
jitter. Vectrons are good.

Or: could I trigger off the sq wave source while inputting it into the
line driver, then look at the output of the driver on the scope and
measure the time from trigger to output edge rising and get the jitter
of that time?

Do that too. The problem there is being sure that the edge you're
observing is the same one that caused the trigger; if not, you're back
to depending on the oscillator's cycle-to-cycle jitter being low. Most
sampling scopes have a 40-70 ns trigger delay, several clock cycles. I
guess you just prowl the edges for the one with the least jitter!

You can also measure jitter with and without the chip in the path, and
deconvolve the device jitter. That takes out both the oscillator and
the scope jitter.

What kind of scope do you have? Samplers run roughly 1-3 ps RMS jitter
all by themselves, when they're in a good mood.

I sure hope you're talking 5 ps RMS, and not peak-to-peak.

John

 

[mike]

I have a request to measure the edge-to-edge jitter from some F100314
diff ECL line drivers. The customer wants it done with 100 MHz square
waves, and the jitter spec is > +/- 5 pSec. Does that mean I need a sq
wave with better than that jitter? Can I get one?
Or: could I trigger off the sq wave source while inputting it into the
line driver, then look at the output of the driver on the scope and
measure the time from trigger to output edge rising and get the jitter
of that time?

Sometimes people requesting measurements don't fully grasp what they're
asking and don't have any idea what they're gonna do with the data when
they get it. I'm gonna go way out on a limb and state that measuring
5pS in a circuit environment other than the actual application is
meaningless.

Look at the slew rate of the input and figger out how much noise, power
supply rejection, reflection influence etc. it's gonna take to swamp
your measurement. Even if you make the perfect circuit, how will that
translate to the actual environment?

Having said that...Can you make a ring oscillator and look at the FM
with a spectrum analyzer?
mike

--
Return address is VALID.
500MHz Tek DSOscilloscope TDS540 $2200
http://nm7u.tripod.com/homepage/te.html
Wanted, 12.1" LCD for Gateway Solo 5300. Samsung LT121SU-121
Bunch of stuff For Sale and Wanted at the link below.
http://www.geocities.com/SiliconValley/Monitor/4710/

 

[[email protected]]

Thanks for the response.
Oooh, good idea on the XTAL osc, I'll probably do that.
I've got a TDS5104 ( I think its a 4) with the jitter package
installed, but I'm stuck on a source.
What Harris suggested is a splitter/delay method, trigger off the
undelayed, display and find the delayed trigger edge and meas the
jitter of the time till the next edge. But right now I'm measuring
"period jitter" using the option in the scope. I'm getting ~50 pSec
using an old HP pulse gen.

 

[[email protected]]

An outside company came to us and asked "can you verify that the
edge-to-edge jitter of these ECL F100314 parts is < 5 pSec? ". They
sent a drawing of a suggested setup: splitter/delay line. From what
I've found searching, it's a recommended method, but I'm not sure of
the impact of the source on the reading. I know jitter is related to
phase noise by integrating and there's a nice spreadsheet to calculate
it out there somewhere (Wenchel website?), I could do some thing FM,
but... of course they don't want to spend any money or time, but want
it done fast, like tommorow, and cheap, so don't waste time figuring
anything out, hook somethings up, push a button on the scope and write
down the right answer!.

 

[mike]

[email protected] wrote:
snip

don't waste time figuring

anything out, hook somethings up, push a button on the scope and write
down the right answer!.

Where can I get a job like that. I'm there in a New York Minute ;-)

mike

 

[Fred Bloggs]

The customer wants it done with 100 MHz square
waves, and the jitter spec is > +/- 5 pSec. Does that mean I need a sq
wave with better than that jitter? Can I get one?
Or: could I trigger off the sq wave source while inputting it into the
line driver, then look at the output of the driver on the scope and
measure the time from trigger to output edge rising and get the jitter
of that time?

You will never do it:
http://www.evaluationengineering.com/archive/articles/0102scope.htm

 

[John Larkin]

You will never do it:
http://www.evaluationengineering.com/archive/articles/0102scope.htm

I do measurements like that all the time. My 11801A scope has
trigger-to-vertical jitter of about 1.5 ps RMS, so it can make useful
device jitter measurements to about 1 ps. (Just got a CSA803C off
ebay, lower jitter still, but haven't fired it up yet.)

The guy in the article chose not to mention equivalent-time sampling
scopes; the newer ones have jitters way below 1 ps.

John

 

[Rich Grise]

[email protected] wrote:
snip

don't waste time figuring

Where can I get a job like that. I'm there in a New York Minute ;-)

Depends what kind of money you want to make. I've got a guy right now
who will let you do that.

For seven-fifty an hour. ;-)

Cheers!
Rich

 

[[email protected]]

You guys are scaring me.

 

[[email protected]]

Thanks for the article. I'll add it to my growing collection. At least
I'm learning alot. My scope is a TSD5104 with 8 pSec trigger jitter.

 

[[email protected]]

Congrats on the new CSA !! How much was it? I'm using a TDS5104. They
say it ha ~8pSec trigger jitter. With the best source I have, I'm
getting 50-60 pSec jitter.

 

[John Larkin]

Congrats on the new CSA !! How much was it? I'm using a TDS5104. They
say it ha ~8pSec trigger jitter. With the best source I have, I'm
getting 50-60 pSec jitter.

About $1200. The older boxes, 11801 or 11802's, go for a few hundred.
It's the sampling heads that are expensive; $500 for an SD14 dual
probe (3 ghz), $800 maybe for an SD22 (12 ghz dual-channel), kilobucks
for an SD24 (20 GHz TDR). Still cheap compared to the new stuff!

Your cycle-cycle jitter is probably the pulse generator. Even a cheap
20 MHz TTL xtal osc can do a lot better than 50 ps. There are lots of
cheapish ecl xtal oscillators that will guarantee 1 ps RMS.

John

 

[Rene Tschaggelar]

I have a request to measure the edge-to-edge jitter from some F100314
diff ECL line drivers. The customer wants it done with 100 MHz square
waves, and the jitter spec is > +/- 5 pSec. Does that mean I need a sq
wave with better than that jitter? Can I get one?
Or: could I trigger off the sq wave source while inputting it into the
line driver, then look at the output of the driver on the scope and
measure the time from trigger to output edge rising and get the jitter
of that time?

The jitter is measured as close sideband noise. Just integrate the
spectrum found from 10 to 1MHz beside the carrier on either or both sides.

Rene

 

[[email protected]]

Thanks to all for chiming in. I really appreciate the info, you can
tell I'm out of my realm here.
Here's what I'm getting: Using a TDS5104, trigger jitter ~ 8pSec, to
measure the jitter of just the source, a Tektronix AWG2021, I get ~ 45
pSec. Measuring one of the outputs of the ECL part, F100314, with the
unused outputs terminated in 50 ohms, I get ~ 30 pSec. Why less, and
not more?
Digikey carries an Epson 106 MHz XTAL osc, 3 pSec RMS jitter. I
ordered a couple. I'll use one for a source and post what I get.
Stay tuned...

 

[mike]

Thanks to all for chiming in. I really appreciate the info, you can
tell I'm out of my realm here.
Here's what I'm getting: Using a TDS5104, trigger jitter ~ 8pSec, to
measure the jitter of just the source, a Tektronix AWG2021, I get ~ 45
pSec. Measuring one of the outputs of the ECL part, F100314, with the
unused outputs terminated in 50 ohms, I get ~ 30 pSec. Why less, and
not more?
Digikey carries an Epson 106 MHz XTAL osc, 3 pSec RMS jitter. I
ordered a couple. I'll use one for a source and post what I get.
Stay tuned...

What are you using for a test breadboard?
If you think of the the input slew rate as millivolts of noise per
picosecond of jitter, you quickly determine that you can't have any...
mike

--
Return address is VALID.
500MHz Tek DSOscilloscope TDS540 $2200
http://nm7u.tripod.com/homepage/te.html
Wanted, 12.1" LCD for Gateway Solo 5300. Samsung LT121SU-121
Bunch of stuff For Sale and Wanted at the link below.
http://www.geocities.com/SiliconValley/Monitor/4710/

 

[sdy]

I have a very nice board I made from copper clad. ( The device the 5
ECL line drivers, F100314) One set of inputs is tied to device's
internal refernce voltage, the other inputs are connected to SMAs by
50 ohm coax, shields soldered to copper PCB. Outputs are connected
likewise, un-used ones terminate in 50 ohm sma loads. The device is in
a good socket and power suppies bypassed at the socket.I took care
knowing I needed fast, highh freq signals. At 30 MHz, theres no
ringing at all and rising/falling edges are very nice.