From: DAmbrosia, John F ([email protected])
Date: Tue Apr 25 2000 - 06:27:47 PDT
Francis,
Thanks for your response. I apologize for the lateness of my response, but
I have been traveling for the last week.
First, let me say that I understand the overall skepticism of your response.
The phenomenon of "nanosecond" discontinuities" has just been recently
discovered, and AMP has been working to resolve the issue, which has
resulted in the introduction of our Quiet MateTM Contact technology.
Nanosecond discontinuities should not be so quickly dismissed, however, as
only two of our customers have seen the phenomenon, which was discovered
while evaluating a system for high availability. Just because it is not
being seen in a controlled lab environment, does not mean that it will not
be seen in the field. It is unclear how rigorous the test environment was
used to test your devices. Furthermore, it is very difficult to measure
the resulting glitch. What equipment did you use to detect and measure the
glitch? Can you further describe the glitch? duration/width, frequency of
occurrence during hot-swaps, specific levels (20mV/50mV/100mV?), etc.
There are several comments that need discussed -
1. We know that there will be "discontinuities" during Live Insertion,
it is NOT "microsecond discontinuities" we are investigating BUT any data
errors in BLVDS during Live Insertion.
We agree that the area of concern is for any data errors
during any Live Insertion. High availability systems demand it! It has
been proven that in a CPCI environment nanosecond discontinuities can cause
data errors.
2. The occurence of abnormalities on the signals are COMMON on both
signals, thus data was not impacted.
Based on test data taken in our labs, we do not agree with
this statement. 'Glitches' occurring during live insertion events do not
and will not happen at EXACTLY the same points in time. If the resolution
is not accurate enough, they may appear to be occurring at the same time,
but will actually be offset from one another at some distinct point in time.
If the resolution is not that accurate enough, the measured glitch is
probably resulting from something other than nanosecond discontinuities.
They are extremely difficult to observe, even with the most advanced
equipment. We are not sure what was seen. Even the initial 'glitch',
resulting from the potentials on the card and backplane coming together
(assuming no nanosecond discontinuities occur) would never happen at EXACTLY
the same time. Can you measure any distinct separation in time of the
"glitch" on each leg of the differential pair?
3. The National parts were designed better than the industry
specifications which means the thresholds for high and the threshold for low
have a higher noise margin as specified in the data book. On top of
this,there are internal filters in the receiver that will further take care
of the glitches.
This very issue came up at PICMG last week, where it was realized by
committee members that higher margins / tighter thresholds can actually be
worse in a hot swap environment.
This issue is related to the system application. I suggest you contact me
directly, and we can discuss your logic and the hot-swap / nanosecond
discontinuity issue and the potential impact on systems. My contact info is
listed below.
Have a nice day!
John D'Ambrosia
AMP Circuits & Design
Program Manager, Strategic Accounts
Tel (717) 986-5692
Fax (717) 986-5095
Email: [email protected] <mailto:[email protected]>
URL: http://www.amp.com/simulation <http://www.amp.com/simulation>
-----Original Message-----
From: Francis Chiu [SMTP:[email protected]]
Sent: Friday, April 14, 2000 7:28 PM
To: si-list
Subject: Re: [SI-LIST] : BLVDS Hot Swap - Connectors
Hi, John,
Thank you for your information. I did go through the file you
attached. Actually, I am busy working on other projects, but I will spend
sometime on
the BLVDS Hot Swap issues.
First of all, I have to stay the emphasis for our Hot Swap tests
which we did one year ago - 0 data error in Live Insertion.
We know that there will be "discontinuities" during Live
Insertion, it is NOT "microsecond discontinuities" we are investigating BUT
any data errors in BLVDS
during Live Insertion.
The tests were performed on multichannel BLVDS (9 channels per
chip), the speed of theTektronix tester is 100Mb/sec, coded data.
The BLVDS common mode range is +/-1V. The signal swing is a
typical 300mV transition centered around +1.25V. Note the differential swing
is twice the
magnitude of the single-ended.
When a card is being hot-plugged into the system, that transceiver
would be disabled, the impedance should be around 300K Ohm (?)
Under these circumstances, the glitch we observed was around
several millivolts and this will not create any problem. The occurence of
abnormalities on the signals are COMMON on both signals, thus data
was not impacted.
The National parts were designed better than the industry
specifications which means the thresholds for high and the threshold for low
have a higher
noise margin as specified in the data book. On top of this,there
are internal filters in the receiver that will further take care of the
glitches.
Actually, I like to find out whether anybody did the Hot-Plug test
at a data rate higher than 100MBits/sec with BLVDS.
Up to this minute, I still have not hear any complaint from the
Telco guys who use the BLVDS in their 3G wireless basestation designs,
Hot-Plug
is a MUST requirement for the wireless basestations.
Regards.
Francis
[email protected] on 04/14/2000 01:30:00 PM
To: [email protected]@Internet
cc:
Subject: [SI-LIST] : BLVDS Hot Swap - Connectors
In the past few days there have been some questions regarding BLVDS
and HOT
SWAP applications. I am with AMP, and we have been very busy
investigating
glitches that have been seen in applications requireing HOT SWAP.
It was
questioned whether this was an issue in a BLVDS application, or
would the
differential nature make it a non-issue. A comment was made that
the
phenomenon would not be seen since it was seen on both signals of
the
differential pair simulatenously.
The testing we have done in our labs (for 2mm HM product, as well as
any
other separable conductive interface) has shown that we ALMOST NEVER
connect
the multiple beams of the contact at the same point in time. For
typical
card insertion speeds (this also includes all speeds within certain
specifications), we have measured the time between these connections
(of
different contact beams ON THE SAME PIN) on the order of
micro-seconds.
For different pins within the same connector, the time between these
connections can be much longer. This separation time depends highly
on the
tolerances and design of the mechanical cage supporting the inserted
cards,
the tolerancing in the alignment and guidance of the connector
housings, and
the human factor (operators or maintenance personnel will insert
cards at
different speeds, sometimes not insert them fully, insert them
partially
before finishing the insertion, etc., etc.).
Sooo...we have definitely seen (during live insertion scenarios)
occurrences
of inserted cards impacting the individual legs of a differential
pair
separately. This impact could be separated by micro-seconds or
longer.
In addition to these relatively long (with respect to a high-speed
waveform)
micro-second separations, we have also recently discovered a much
faster
discontinuity in our labs. This new phenomenon also occurs during
hot-swap
events and occurs right at the initial time when the conductive
interfaces
come in contact with one another. The connections/disconnections
between
backplane contact and card contact happen as quickly as nanoseconds
and as
slow as micro-seconds. We have been referring to the phenomenons as
nanosecond discontinuities. We have also seen this occur at any
highly
conductive separable interface.
The nanosecond discontinuities are extremely erratic and never seem
occur
at the same positions or in the same timeframes consistently (even
for the
same contact, let alone discrete contacts within the same
connector).
The combination of mechanical tolerancing and nanosecond
discontinuities
provides us with an extremely wide range of time with which we can
separate
the impacts of a hot-swapped card on the individual legs of a
differential
waveform. Depending on the logic family you are using, opposing
potential
levels (i.e. 0.0-5.0V, 0.0-3.3V, 0.5-1.5V, etc) during an insertion
can
resulting in significant current flow to/from the backplane from the
inserted card. The capacitive loading, termination, topology,
driver
capabilities, and sampling of the system will all play key roles in
the
worst-case analysis of such 'glitches'. We have measured (on 5V
systems)
voltage disruptions as high as ~2 V (on a logic level 'low'
waveform). On
lower voltage systems (like LVDS or ECL or GTL), we can measure
similar
percentages of voltage alteration to an operating waveform.
For those who are interested in a presentation we gave on this
phenomenon,
go to www.amp.com/simulation <http://www.amp.com/simulation> . Go
to the
"documents" section, and choose "Conference Presentations." Select
"Nanosecond Discontinuity Impact on Hot Swap."
Have a nice day!
John D'Ambrosia
AMP Circuits & Design
Program Manager, Strategic Accounts
Tel (717) 986-5692
Fax (717) 986-5095
Email: [email protected]
<mailto:[email protected]>
URL: http://www.amp.com/simulation <http://www.amp.com/simulation>
**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE si-list
or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****
**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE si-list
or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****
**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****
This archive was generated by hypermail 2b29 : Wed Nov 22 2000 - 10:50:05 PST