Dear colleagues,

I have to design a flex PCB with a 90 Ohm differential pair. (Edge coupled microstrip)
There are different calculators on the web to ease my life. My problem is: I don't know, which one gives the most reliable result?
When I enter the same properties, I got completely different values from different calculators.
Just as an example:
Board thickness (h) = 50 micron
Copper thickness (t) = 18 micron
Track width (w) = 137 micron
Distance between tracks (s) = 254 micron
Dielectric constant = 3.4
(Sorry guys, I live and I have grown up in a metric world )
With these values, Saturn PCB's off-line calculator gives 57 Ohm differential impedance.
Altium's on-line results 88.6 Ohm ( https://resources.altium.com/p/differential-microstrip-impedance-calculator )
PCBway's on-line gives 90.8 Ohm ( https://www.pcbway.com/pcb_prototype/impedance_calculator.html )
By Digi-Key I got an error message and it calculates 67.3 Ohm ( https://www.digikey.com/en/resource...ors/conversion-calculator-pcb-trace-impedance )
Multi-Circuit also has a calculator that gives 67 Ohm ( https://www.multi-circuit-boards.eu/en/pcb-design-aid/impedance-calculation.html )
Eeweb's calculator results 91.8 Ohms. ( https://www.eeweb.com/tools/edge-coupled-microstrip-impedance/ )
So, the election ended with 3:3 But I'm afraid, there's only one proper calculation.

None of the above calculators are able to give data on the standard 25 micron thick material for 90 Ohms.

Which one shall I trust? Do You have experience with these calculators? Unfortunately, I have no field solver and I don't feel motivation and knowledge to manually solve Maxwell's equations

 

Jean82, I use Saturn and my CAD tool's built in Stackup editor, (Mentor Xpedition) to do my stackups for both rigid and flex boards. After creating what I believe are the right/best combinations to get what I'm looking for, I will submit the information to that FAB shop for review. The people that are going to build your board are going to know what they can do and will have empirical data to support their processes, etc. Talk to your FAB shop for better understanding and materials, etc. for your application, they will have good information for you for sure and will want to make sure you do the right thing for the best outcome. I will also say this, if your flex is a short length, consider that it may not need to be controlled impedance which should make things a lot easier. If you do need it, be mindful of thicker dielectrics and the ability for flexing depending on your application of either dynamic or flex to install requirements. Also with thicker dielectrics and a dynamic flex or even one with a vib requirements can be tricky if your bend radius is small since stretching and compression can break smaller traces. (I know from experience ) As for the calculators question, I would pick Saturn and if your CAD system has a stackup editor solver in it, use it but always run the numbers by your FAB shop... If they use a TI calculator or Abacus to get what they can produce and what you need, so be it. ;-) Good Luck!

 

Hi PadMasterson,
Thanks for Your answer.
According to the FAB, they can only produce the flex circuits on 25 micron material. Its dielectric constant is 3.3.
With this thickness, I cannot set up a manufacturable construction that results 90 Ohm differential impedance. 32 Ohm is the maximum I can reach - doesn't seem to be ideal for USB2.
Since I don't have a built-in calculator in my PCB SW, (I know Mentor Xpedition but here we don't have it, we use a much simpler program) I've used Saturn's calculator. The calculator on the FAB's site doesn't help - it only knows FR4, min 0.6mm thick - way too far from the flex circuit.
The circuit contains only the wires from one board to the other. It will be assembled once and then doesn't move.

 

Jean82, So 25 micron material seems pretty dang thin for a material that has the copper on it? by my calculations, 25 microns is .00098". I can see that as a coverlay without adhesive but I'm wondering why they can't do a more conventional material for Flex? BTW, how long of a flex are you looking at, 2 - 3" worth? Might not even be worth trying to get a controlled impedance for something that short, but I'm not up on USB2 edge or data rates, so maybe it's needed? Do you have a width restriction and how are you terminating the ends of the flex, connectors on a rigid section? The Saturn stand alone calculator can do a custom material you can set your 3.3 Dk value in, but the limits are stated for material thicknesses and the like, so it may not get you a good answer. Why can't the FAB shop give you a detailed configuration for what they need to get you those values? If they can't or don't know how, I would look for another FAB shop... If nothing else, check with another shop for the info and use their numbers if you can. This is not something out of the ordinary as far as creating a diff pair on flex, true you have more limitations in many ways, but I don't think you're asking for something that's never been done before either. Good luck, let us know what you find out too.

 

Dear PadMasterson,

thanks again for Your help. The FAB is JLCPCB in China. They are not my own choice, but our management's. I can't change that even if I would. They have said that the base material is 25 micron and that's the only choice they have. Big land, big company, no flexibility.
I have only these flex PCBs in my hand. The other parts of the whole is designed by colleagues. Our customer wants USB2 with 480Mbit/sec, and I'm pretty sure they will test it too. To my knowledge it requires some 2GHz of bandwidth - hard to believe it is realistic, since we're in the microwaves already, but the 5th harmonic of 480MHz is there. And this is just the bandwidth, reflections are not considered yet. The length varies from 5cm to 35cm (2" to 13,5" if I'm correct). The connectors are said to be specified for USB2 - also out of my competence.
During the time I had to finish this job, anyway. I did it as I can: 0.1mm wide tracks, 0.1mm spacing (4 mil), 25 micron (0.1 mil) base material and a GND area on the other side. Copper is 18 micron (don't know how many oz it is) on both sides. We will see what happens, I have constraints if the 480MBit/s will work.

Dancing is not easy with tied legs

 

Jean82, Best of luck on this endeavor. I fully understand the issues of not having control over vendors and related things, it just makes the job harder sometimes. To be responsible for a product/part/PCB and not have control over the vendor or able to get information from them is like painting in the dark, you know the paint is going on the wall but you're not sure if you're covering it right or missing places... On that shorter length flex, I suspect it will be OK, but the longer ones may be more of an issue if not close to right. As for the connectors, I haven't seen a USB 2.0 requirement yet, but unless the connector and interface is just right, it's going to be an impedance hit at both ends regardless. It may not be much of one but a hit is a hit and all you can do is hope it's not a show stopper, but then you don't have the control over that so you shouldn't worry about that part anyway.
I ran your numbers in Saturn and they looked pretty good. The 18 micron copper is 1/2 oz. and your trace and space if I understood everything gave me like 95.5 ohms assuming you have another 1/2 oz. plated on that first 1/2. If no additional plating is done, I think you could be on the high side at 105 Ohms, but I don't know if you have/need plating on the flex... I think you will be in good shape in the long run. Drop a note if you can when you find out how it worked out, I'd like to know myself. Have a great day!

 

Hi PadMasterson,

I'm sorry, I'm not sure that this board comes back to my hand. Also, we don't have instruments to measure their differential impedance - a TDR or a VNA would be ideal, but they're not here. By the way, I think this specification is a bit of overshoot. I mean, that for the concrete product, even 1Mbit/sec is more than enough. I don't know why our customer wants 480Mbit/sec.

On the other hand, I made again the calculations, and I cannot get the same results like You. Where is the difference, am I doing something wrong? Yes, the bottom side it plated, 18 micron, and it is GND.
I attach two pictures, one in metric, the other is imperial calculation of the same thing. ( I was wrong in the previous post, 25 micron equals 1 mil, not 0.1 - I am sorry.) Both gives me only 20 Ohm differential impedance. In case I reduce the width and the distance of tracks to 0.08mm - this is the border of manufacturability, I don't think it is practical to go there but for the calculation it is just fun - and I get some 31Ohms. If I calculate it for embedded edge coupled differential pair, it is still only 39 Ohms. I didn't set the material's name but I set the dielectric constant to 3.3. The results are far away form 90 Ohms in all cases.

 

Jean82, So I noticed your version of the Saturn calculator and realized my version was out of date so I loaded the latest, V8.34 in case you are interested. I put in your numbers and indeed we get the same exact numbers. I am not sure what I did the other day to get the numbers I got, I didn't save the input data, so I may have entered something incorrectly, I can't say for sure now, sorry about that, I was even wondering what the heck I did for that too... As for the newer calculator, I did see some slight differences between your version and the newer one but not anything I think is a major issue, but here are some things that may be different. In your version above, you have a .5 oz. copper layer and an additional .5 oz. of plating which should total to about 1.4 mils as the calculator shows. I made the assumption that you are not going to have exposed copper running on this flex so a cover layer would be placed over the copper which would embed the traces. Since you had a plating process for your numbers, I started with a 1 oz. copper and added a .98 mil cover layer and used the embedded calculator and added the total thickness to the H2 value to get about 2.66 mils.I also noticed that you had FR-4 STD selected but have the Er at 3.3. In the new version, if you change the Er value the Materials change to Custom. I don't think this is a big deal, but it was something I noticed. So by adding the coverlayer and overall thicker copper, using the same other values, I get the diff impedance of 36.4 Ohms which is not anywhere near the 90 but is more than the 20 from your numbers. I don't know if this helps or not and I understand you don't have the tools available to measure the end result, but as I mentioned before, I think having a better dialog with the FAB shop would have helped a lot and it's sad you don't have that ability. Here is that last calculation setup I did to get the 36 Ohm values for reference.

One additional thing I noticed and may or may not be a factor is in your values, there is a warning regarding the W/H and S/H which may indicate the values are very close to the calculators ability to accurately calculate the correct numbers. When I entered my numbers above, that warning was gone so it may be a slight bit better for accuracy in this case, but in either case, the result is not the 90 Ohm differential impedance value you were trying for. One last thing I checked also, was using Kapton as the material since it's Flex, but the Er was only a .1 difference at 3.4. Lastly, the material thicknesses for the dielectric seem odd that adhesive thickness was not included or mentioned? If I add 1 mil of thickness to the overall thickness in my setup above, I get a total of about 3.66 mils for H2 which did drop the impedance down a bit to 33.49 Ohms. Sorry I couldn't get the numbers I had yesterday, I wished I had them again to see what I did, but at least I get the same as you now and we agree it's not great, but it is what it is and physics doesn't change just because we want it or would like it to. Good luck on the build and have a great day!

 

Dear colleagues,
I have to excuse me. Although the flex prints are manufactured, they didn't came back to my hands, but to a local colleague's instead. I don't know how and what he tests, therefore I cannot report the results. I was just an assistant in this task.