| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Hi Neven, Hi all,
Thanks for your comments, See my remarks/additional comments inline. I update the slides to integrate your feedback. (especially new slide 7 and updates on slide 9, as well as some corrections, like SMAà3.5)
Regards, Stefan
Von: Neven Pischl [mailto:npischl@xxxxxxxxxxxx]
Hello Stefan, thanks for the material. I have a couple of comment that I hope can help out.
Slide 4/9: If you want to de-embed I strongly suggest single-ended traces, which means the spacing between traces must be many times the dielectric thickness in the entire routing. I still do not recommend de-embedding in general, and instead suggest using the results as-is with the jig. It makes it much simpler for the users and with good jigs it should not matter for the measurements that we want to do. è Sure, the traces from 3.5 connector to the pin are single ended, 50 ohm, and if you have a multi-Layer PCB you can have them in between two GND Layers, so coupling is reduced. However you carefully have to design the 50 ohms. For “simple” measurements (e.g. IL of cables) you can ignore the influence of the traces, however for measurements of an connector (without cabeling) the influence of the traces will become recognizable. Therefore such a jig allows both options – and if you do it correct, the de-embedding does not produce an additional failure, especially if all traces are identical (same length, impedance, etc) you can use ONE de-embedding model for all traces and therefor if you make an mistake
“Low-impedance” connection to the GND plane or another PCB should be defined. It could be defined physically, such as using wide metal brackets and continuous connections between the PCBs and them as well as to the metal GND plane. è Once a PCB is defined, we could also define a corresponding metal block (size, material, CAD-drawing,…) where the PCB could be fixed (screws) with proper GND connection to the GND plane (if the block is heavy enough)
SMA – I suggest using 3.5mm connectors instead of SMA. I use Huber&Suhner air-dielectric 3.5mm connectors which I measured as much better than e.g. Johnson SMA which are routinely used for “miscellaneous” lab measurements. I am not sure if these H&S connectors come in all shapes we might want to use, I use straight edge-mount type. è You are right, when I wrote SMA I instead had in mind “SMA compatible high quality RF connector”. The edge-mount types of such a connectors however have the disadvantage, that it is hard or even not possible to keep all traces the same length (which reduces the error probability of de-embedding as mentioned above). Another proposal which was sent to me was a right angled high quality 3.5 test connector. On http://www.molex.com/molex/products/datasheet.jsp?part=active/0733870020_RF_COAX_CONNECTORS.xml you can find the suggested product. Such a type of connector would give us the best of both worlds: freedom to place on the PCB anywhere and also good RF properties.
Slide 5/9: Again, I envision a metal bracket (typically L-shaped) to provide connection to the metal GND plane. è As I mentioned above, if the PCB is defined, a CAD-drawing of a metal block with 50mm height could be defined as GND connection and then there have to be defined basic layout rules for cable harnesses on the GND Plane (e.g. 50mm above GND, 150mm distance between wires, …)
Slide 6/9: Again, I generally prefer taking data with the jig included, providing I can design and demonstrate that the effect of the jig does not alter the data in any significant way and we can live with it. De-embedding should be done only on special occasions when absolutely needed.
FR4 vs. Rogers – it depends on what you want to measure and how the jig is designed. FR4 might be used for some (but not balance) measurements. In general, to avoid confusion and issues that people who are not so familiar with little but very significant details, I prefer specifying Rogers and keep it simple and on the safe side. è Yes, I also would propose to define 1 material at the end (Rogers RO4003), however, I didn’t want to define one by myself, so I wait if there are any other opinions. Is “Rogers RO4003” complete defined, or are there different variants?
One general comment on the test jigs shown in the photos is that they are designed to measure the connector only. While this is important to know the connector-only performance and it needs to be done, the performance of the connector mounted on the cable is what really matters in the end. E.g. the un-twisted section of the wires, and their relation to the other wires in the vicinity are so critical that they can completely destroy an otherwise “good-looking” connector. Therefore, in my very strong opinion, these jigs must include a section of the cable, perhaps something like 5-10 cm or so – as appropriate for each connector type under test – and the influences of various possible wire arrangements in that area should be included in the measurement. I am sure this would show that even with the best connectors and cables in the world, this area can influence the overall performance drastically and prove to be the weakest segment of the link. Such a test setup would also allow understanding any wire-management that might help improve it. è With the shown test jigs you can also measure connector + cable (for inline, just increase the distance, for PCB-to-harness the metal plates which are shown allow to shift the PCB along the longer edge) I added a picture to the presentation and updated acc. to your comments
I hope this helps.
Neven
From: Stefan Buntz [mailto:stefan.buntz@xxxxxxxxxxx]
Dear colleagues,
attached you find a some additional notes on measurement adapter/test setup and a clarification of what – at last – I understood as “extended link segment”. Please don’t understand this slide as a finished proposal for a unified test setup – I would like to try to add more additional information to this slideset and summarize all the inputs we have (please give feedback if you have any additional input, most probably I didn’t get all comments during our F2F meeting) before we go on an try to define the unified test setup/adapter.
Regards, Stefan
Von: Mehmet Tazebay [mailto:mtazebay@xxxxxxxxxxxx]
Dear Colleagues,
Please find the attached meeting notes from our F2F meeting in Geneva. We look forward to working with you.
Best Regards, Mehmet Tazebay
|
Attachment:
130801_Buntz_IEEE802.3bp_TestAdapterConsiderations_V0.2.pdf
Description: 130801_Buntz_IEEE802.3bp_TestAdapterConsiderations_V0.2.pdf