Signal Integrity Extension CVS File Data

[garyKGTM3]

Hello,

Is there a document for the Signal Integrity Extension that explains the data in the exported CVS file? What are the paths and segments and how are they defined? Also, why is the CVS data in a non-number format having a text unit after each number? This text must be stripped out before analyzing the numbers in a spreadsheet. The numbers themselves are repeated in each column in groups of segment and path totals, making it impossible to simply sum columns of numbers. And speaking of sums, the trace length sums (in mils) are much longer than the trace lengths (in mm) in the CVS file exported by the Eagle ULP data export script.

Thanks,

Gary

[jorge_garcia]

Hello @garyKGTM3,

I hope you're doing well. There's not documentation as far as I'm aware of for the columns in the CSV. As far as I know the columns in the extension itself are intended to describe it.

Path is the total of all the segments. In my case all the values are exported in mm and the sum checks out. Where is the conversion to mils coming from? Is that something you are doing on your own?

I don't think you need to individually to the sums if the path gives it to you. But you want to do it anyway you can delete the rows that have paths in them. Additionally you can use the text to columns functionality in Excel (I'm sure Calc has something similar) to easily strip the units if you don't want to see them.

Let me know if there's anything else I can do for you.

Best Regards,

[garyKGTM3]

Hi Jorge,

Thanks for your reply.

Let's compare the length for one signal trace on the PCB, M_A0. Here is data exported by the length-freq-ri ULP, showing data for the M_A0 trace:

EAGLE Version 9.7.0 Copyright (c) 1988-2023 Autodesk, Inc.
List of signals with length and its max. frequency / current
exported from C:/Users/gembl/AppData/Local/Temp/Neutron/ElectronFileOutput/26376/brd-45b6210a-145e-49af-9266-ddeb05093221/EpiBack.brd
at 11/20/2023 10:21 PM
Signal	f max. [MHz]	l [mm]	A [mm2]	R [mOhm]	w min [mm]	w max [mm]	Imax [A]
GND	241.48	1241.533	0.002	12149.98	0.051	1.016	0.19
...	...	...	...	...	...	...	...
M_A0	3810.95	78.668	0.004	384.93	0.102	0.127	0.35
...	...	...	...	...	...	...	...

Here is data from the SI Extension:

Signal Path	Length	Impedance	Time Delay	Resistance	Capacitance	Inductance
M_A0    11/14/2023 16:14 - 1.00GHz, Target Z0 50 Ohm +/- 10 %
Path-U2(P3)-U1(AK25)	1417.44 mil	40.58 - 128.42 Ohm	255.78 ps	2283.82 mOhm	0.00 - 3.32 pF	11.81 nH
Seg-1	13.27 mil	52.07 Ohm	1.98 ps	14.52 mOhm	0.038 pF	0.10 nH
Seg-2	5.27 mil	116.37 Ohm	0.63 ps	4.58 mOhm	0.005 pF	0.07 nH
Seg-3	0.41 mil	76.21 Ohm	0.06 ps	0.39 mOhm	0.001 pF	0.00 nH
Seg-4	3.31 mil	76.21 Ohm	0.45 ps	3.11 mOhm	0.006 pF	0.03 nH
Seg-5	3.31 mil	62.58 Ohm	0.60 ps	4.07 mOhm	0.010 pF	0.04 nH
Seg-6	0.41 mil	62.58 Ohm	0.08 ps	0.51 mOhm	0.001 pF	0.00 nH
Seg-7	5.27 mil	128.42 Ohm	0.82 ps	5.77 mOhm	0.006 pF	0.11 nH
Seg-8	204.88 mil	43.30 Ohm	37.27 ps	264.19 mOhm	0.861 pF	1.61 nH
Seg-9	5.27 mil	128.42 Ohm	0.82 ps	5.77 mOhm	0.006 pF	0.11 nH
Seg-10	3.73 mil	62.58 Ohm	0.68 ps	4.58 mOhm	0.011 pF	0.04 nH
Seg-11	2.70 mil	69.87 Ohm	0.48 ps	4.29 mOhm	0.007 pF	0.03 nH
Seg-12	1.02 mil	69.87 Ohm	0.18 ps	1.62 mOhm	0.003 pF	0.01 nH
Seg-13	5.27 mil	127.18 Ohm	0.78 ps	7.81 mOhm	0.006 pF	0.10 nH
Seg-14	111.94 mil	45.32 Ohm	20.42 ps	195.68 mOhm	0.451 pF	0.93 nH
Seg-15	188.72 mil	40.58 Ohm	34.40 ps	280.88 mOhm	0.848 pF	1.40 nH
Seg-16	825.90 mil	45.32 Ohm	150.66 ps	1443.74 mOhm	3.324 pF	6.83 nH
Seg-17	5.27 mil	93.76 Ohm	0.89 ps	7.83 mOhm	0.010 pF	0.08 nH
Seg-18	3.73 mil	69.85 Ohm	0.66 ps	5.86 mOhm	0.009 pF	0.05 nH
Seg-19	3.64 mil	76.21 Ohm	0.49 ps	3.42 mOhm	0.006 pF	0.04 nH
Seg-20	5.27 mil	116.37 Ohm	0.63 ps	4.58 mOhm	0.005 pF	0.07 nH
Seg-21	18.84 mil	52.07 Ohm	2.80 ps	20.61 mOhm	0.054 pF	0.15 nH
Path-U2(P3)-U3(P3)	469.74 mil	43.30 - 128.42 Ohm	83.30 ps	590.99 mOhm	0.00 - 1.71 pF	3.93 nH
Seg-1	13.27 mil	52.07 Ohm	1.98 ps	14.52 mOhm	0.038 pF	0.10 nH
Seg-2	5.27 mil	116.37 Ohm	0.63 ps	4.58 mOhm	0.005 pF	0.07 nH
...	...	...	...	...	...	...

There is a total of 269 rows of data for this one trace. There are 15 rows labeled "Path..." with the rest labeled "Seg...".

I already determined the "Seg-1", "Seg-2", etc. refer to portions of the "Path-" part of the trace listed above these. The path name, such as the "Path-U2(P3)-U1(AK25)" and "Path-U2(P3)-U3(P3)" in this example, appear to be the start and end pins of a portion of the trace.

Summing the length column and dividing by 2 to account for the duplicate lengths (paths + segments), then converting from mils to mm by multiplying by the conversion factor 0.0254 mm / 1 mil, I obtained 492.99 mm, which is much longer than 78.668 mm from the ULP. Deleting the path values and summing just the segments (without dividing by 2) gives the same length. Could you help with this discrepancy?

The data format conversion of stripping out the "mil" text and converting to a number is a nuisance. However, keep in mind that if analyzing a bus, there may be over a hundred signals that need data format conversion, not to mention summing and collating.

Thanks,

Gary

[jorge_garcia]

Hello @garyKGTM3,

I hope you're doing well. I would need the full CSV to confirm the math. Just looking at what you have posted it's within the realm of possiblity. The ULP is reporting 78.668mm which is roughly 3100 mils. In the sections you have included  you have about 1900mils of signal length.

If you want to simplify things you can just use the Path results and exclude the individual segments that would simplify things considerably.

Let me know if there's anything else I can do for you.

Best Regards,

[garyKGTM3]

Hi Jorge,

Here are the text and CSV files. I rechecked the total length and conversion factor and couldn't find the problem. Could the SI Extension be counting trace length paths or segments multiple times?

Thanks,

Gary

[jorge_garcia]

Hey @garyKGTM3,

I hope you're doing well. I found something interesting, the length-freq-ri ulp doesn't take into account parallel traces and according to it's own documentation is a simplified calculation. I would be more inclined to trust the results of extension over the ULP which was made many years ago (20 at this point).

I haven't seen the layout but you can confirm if the above limitation applies to your design.

Let me know if there's anything else I can do for you.

Best Regards,

[garyKGTM3]

Hi Jorge,

I manually added the M_A0 trace segment lengths (from the segment properties), plus the lengths of the vias traversed through the board and obtained the same total length as the ULP. It looks like the extension is in error.

Thanks,

Gary

[jorge_garcia]

Hi @garyKGTM3,

I hope you're doing well. At this point I would like to request access to the the files or a simplified version that we can use for testing. I know our developers are going to request it when I take this to them.

We definitely want to get to the bottom of this.

Let me know if there's anything else I can do for you.

Best Regards,

[garyKGTM3]

Hi Jorge,

I can send the brd file directly to you. Let me know how to do this.

Thanks,

Gary