I must be a scope guy. I realize that a VNA (or TDR) will give you the
difference in electrical length of traces on your board, but I'd like to
suggest that if you actually want to know what the skew of the two signals
arriving at the receiving end of those lines will be, you also need to consider
the skew from the transmitting chip and its package.
In PCIe land, we do use the signal out of the DUT to deskew:
Device + --> cable1 ---> C1
Device - --> cable2 --> C2
Invert C2 and measure skew. That's device skew + cable skew + scope channel
skew
Then switch the cables at the device:
Device + ---> cable 2 --> C2
Device - ---> Cable 1 --> C1
Now the skew you measure is -Device skew + cable skew + scope channel skew.
Each skew measurements is averaged from multiple sweeps.
Add these skew values together, you get 2 * (cable skew + scope channel skew).
The +Device skew and -Device skew cancel.
Subtract them, and you get 2 * Device skew.
We compensate for the cable + scope channel skew, not for the device skew.
My point in saying all this is that in my experience the device (DUT)'s skew is
usually not zero, despite people's best efforts to match the trace lengths from
the chip to the connector that plugs into the fixture, and from the connector
to SMP connectors on the fixture. It is very small usually, but not zero.
--- Joe S.
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On ;
Behalf Of Iliya Zamek
Sent: Thursday, July 13, 2017 12:27 AM
To: buenoshun@xxxxxxxxx; Chris.Scholz@xxxxxxxxxxxxxxxxxxxxx
Cc: tom@xxxxxxxxxxxxxxxxx; dmarc-noreply@xxxxxxxxxxxxx; si-list
<si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: measuring low levels of skew
Hi Istvan,
It would be interesting to see your experimental results. These are some more
thoughts about measurements: two TDR triggered simultaneously from the same
oscilloscope Timebase, working in parallel, should remove timebase low
frequency noise and jitter; some additional noise due to added second TDR might
be removed by averaging; Flicker noise and jitter between two TDRs hope will
cancel each other if TDRs are identical; keeping reference trace on the same
board helps compensate for the temperature variations.
Good luck at your measurements.
Thank you.
Iliya
From: Istvan Nagy <buenoshun@xxxxxxxxx>
To: Chris.Scholz@xxxxxxxxxxxxxxxxxxxxx
Cc: tom@xxxxxxxxxxxxxxxxx; dmarc-noreply@xxxxxxxxxxxxx; si-list
<si-list@xxxxxxxxxxxxx>
Sent: Tuesday, July 11, 2017 9:40 AM
Subject: [SI-LIST] Re: measuring low levels of skew
Hi,
The IF bandwidth was 1kHz on the VNA. Agilent PNA 10M-40G.
So to determine the delay, I measured the frequency of the 10th harmonic
resonance on gain-S11. Would it be more accurate to look at phase-S11 instead?
As Iliya Zamek suggested, I will look into doing differential measurements,
although my test board does not have diffpairs but lots of parallel
single-ended traces nearby. This seems to be a good idea to eliminate time base
issues, for the TDR-based test. This would require 2 separate probes with
separate cables. I use single-ended to eliminate design-specific detail while
analyzing a material.
Some papers mentioned measuring 2-3ps skew as the end result. I wonder whether
they just did lots of measurements and did averaging? The problem with
averaging is it hides the deviation from the average on each board. Instead of
averaging many boards, I would measure many boards and write down the worst
numbers. In engineering we design for worst case scenario.
Regards,
Istvan Nagy
From: Chris.Scholz@xxxxxxxxxxxxxxxxxxxxx
Sent: Saturday, July 08, 2017 3:50 PM
To: buenoshun@xxxxxxxxx
Cc: tom@xxxxxxxxxxxxxxxxx ; dmarc-noreply@xxxxxxxxxxxxx ; si-list
Subject: Re: [SI-LIST] Re: measuring low levels of skew
István,
Thanks for the additional information! Can you tell us what bandwidth (also
called IF bandwidth) you are using on your VNA?
Chris
Sent from my iPhone
On Jul 8, 2017, at 16:29, Istvan Nagy <buenoshun@xxxxxxxxx> wrote:
Hi,
I sent this email yesterday, but it might have been scrambled. It
looks scrambled in the si-list archive page. resending...
https://www.freelists.org/post/si-list/measuring-low-levels-of-skew,14
Its not really periodic, its random wander and schizophrenic shaking.
In 1 sec it drifts by several ps on the screen.
The problem is, I need to measure single ended traces, 12" each, to
find min and max tpd to calculate max worst case skew. With diffpairs
I would have to measure too many boards and the glass pitch and its
variation would influence it too.
So I might be seeing 2-3 effects:
- Tdr measurement real time wander and random noise, while probe steady.
- vna micro probe lift up, release down on same trace (viewing in
microscope) the the resonant freq changes.
- Even without lifting up, the next vna measurement changes.
The tdr aas hand held, the vna was solid heavy fixed microprobe.
vna Settings. First I checked where the 10 th harmonic was, around 6 ghz.
Then zoom into 5.8...6.1 ghz and scan with 30...100k points. To get
accurate distinction of frequency measurement. That takes some time.
10...30 sec
Regards, Istvan Nagy
-----Original Message-----
From: Tom Dagostino
Sent: Friday, July 07, 2017 3:33 PM
To: dmarc-noreply@xxxxxxxxxxxxx ; 'Istvan Nagy' ; 'si-list'
Subject: RE: [SI-LIST] Re: measuring low levels of skew
Jeff
I've been beating my head against this issue for years and a lot lately.
From the same TDR head I see the two channels move together with very
little drift if any between them. If I put all 4 of my TDR channels
up, again they all move in unison. I think this in in the timebase,
not the receivers or TDR pulses. And it seems to be related to ambient
temperature. Could this
1 sec variation Istvan is seeing related to air conditioners cycling?
Tom Dagostino
971-279-5325
tom@xxxxxxxxxxxxxxxxx
Teraspeed Labs
9999 SW Wilshire Street
Suite 102
Portland, OR 97225
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx] On Behalf Of Loyer, Jeff ;
(Redacted sender "jwloyer" for DMARC)
Sent: Friday, July 07, 2017 3:05 PM
To: tom@xxxxxxxxxxxxxxxxx; 'Istvan Nagy'; 'si-list'
Subject: [SI-LIST] Re: measuring low levels of skew
Hi Tom - good to hear from you. I'm remembering this from a few years
ago, but as I recall I was looking at the two waveforms from adjacent
channels of the same head excited differentially, one flipped
mathematically and the skew calculated automatically. I lined them up
exactly but as I watched, they would drift apart and the skew would
become non-zero, even using a lot of averaging. Of course, you have
access to those same scopes and can check, maybe I'm remembering
incorrectly. But, it appears Istvan is seeing something similar.
Jeff Loyer
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx] On Behalf Of Tom Dagostino
Sent: Friday, July 07, 2017 2:54 PM
To: dmarc-noreply@xxxxxxxxxxxxx; 'Istvan Nagy' <buenoshun@xxxxxxxxx>;
'si-list' <si-list@xxxxxxxxxxxxx>
Subject: [SI-LIST] Re: measuring low levels of skew
There tends to be some timebase wander in sampling scopes. If you look
at an open or some other feature over long periods of time you will
see its location drift back and forth on the screen. But generally all
of the scope channels drift at the same rate so making a skew
measurement between two channels the drift does not affect the skew.
The drift is random in my experience, I've not seen anything with periodic
drift as you describe.
Regards
Tom Dagostino
971-279-5325
tom@xxxxxxxxxxxxxxxxx
Teraspeed Labs
9999 SW Wilshire Street
Suite 102
Portland, OR 97225
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx] On Behalf Of Loyer, Jeff ;
(Redacted sender "jwloyer" for DMARC)
Sent: Friday, July 07, 2017 2:17 PM
To: Istvan Nagy; si-list
Subject: [SI-LIST] Re: measuring low levels of skew
Ah - that makes sense. And that agrees with my experience when using TDR.
It had nothing to do with probes since I also saw it when using only cables.
As I recall, it was less than a ps, but not by much.
Jeff Loyer
-----Original Message-----
From: Istvan Nagy [mailto:buenoshun@xxxxxxxxx]
Sent: Friday, July 07, 2017 1:20 PM
To: si-list <si-list@xxxxxxxxxxxxx>; Loyer, Jeff <jwloyer@xxxxxxxxxx>
Subject: RE: [SI-LIST] measuring low levels of skew
Actually by 1 sec I meant 1 sec.
with the tdr I can view the results real time, but it seems to be
modulated by some noise. 1254ps... 1259ps... 1251ps... varies at about
1 second rate, on the same trace as I hold the probe.
The vna measurement takes about 30 sec to complete ( not hand held),
and each completion has a different result. At a variance of 8 ps roughly.
Regards, Istvan Nagy
Sent from my BLU smartphone device
On Jul 7, 2017 11:45 AM, "Loyer, Jeff" <jwloyer@xxxxxxxxxx> wrote:------------------------------------------------------------------
As I recall, this could be a DesignCon paper all by itself. When we
did our work on the Fiberweave effect (DesignCon 2007 paper)
measuring TDR skew was a challenge. Eventually we were able to
construct a TDR de-skew method that got us close so we could do a lot
of measurements quickly, but I found TDT more accurate for final results
since a reliable "standard"
was easy to construct (a very short "thru" between transmit and receive).
Since we were interested in differential signaling skew, we only
looked at differential excitation, inverted one of the signals using
the math function for skew measurements.
Like Scott said, measure them at some convenient voltage (I would use
a lower voltage, 1/4 of full swing, especially for microstrip). If
you've looked at microstrip waveforms, you've noticed distortion
between the positive and negative waveforms when there is skew. The
slower waveform has a "dip" in it (FEXT) before it rises, making time
comparisons a bit ambiguous. This makes me think that group delay
might be a better method of deducing skew.
If you're looking for sub-ps (I assume by "~1sec" you mean "~1ps")
resolution, that's also going to take a lot of averaging and no
changes/movement between calibration and measurement. It's not hard
to induce a femtosecond or two (or several hundred) of skew.
Jeff Loyer
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx
[mailto:si-list-bounce@xxxxxxxxxxxxx]
On Behalf Of Istvan Nagy
Sent: Friday, July 07, 2017 10:08 AM
To: si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] measuring low levels of skew
Hi,
How do you guys measure skew on PCBs in the 2ps range?
We just made a test board, tried to measure skew but the results
varied randomly by +/-8ps. The test board had very good spread glass.
So I'm worried that the measurement inaccuracy was larger than the
actual value to be measured.
Actually the measurement was single-ended delay measurement with TDR
(20GHz TDR scope), then it was redone with a VNA as well. Both had
similar results.
With the 65GHz VNA (65GHz probes under microscope) the measurement
was for frequency of dips on S11, then converted to delay. Different
traces had different delay, even at the same length, due to skew, as
expected. But the actual number varied over time (~1sec) too.
Is there a better setup to get lets say 1ps accuracy? What instrument
to use, or what's the secret?
Do you measure delay on each leg, or measure diffpairs for skew directly?
Regards,
Istvan Nagy
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu