Author Topic: 6 layer PCB stack up  (Read 2013 times)

0 Members and 1 Guest are viewing this topic.

Offline ali_asadzadeh

  • Super Contributor
  • ***
  • Posts: 1755
  • Country: ca
6 layer PCB stack up
« on: April 29, 2020, 08:16:00 am »
Hi,
I'm designing a 6 layer board, I need to have these impedances, 40 Ohm for DDR3, 80 Ohm for differential signals of DDR3 and 50 Ohm single ended and 100 ohm differential for other interfaces, I'm stucked with selecting reasonably cost effective Stack up to reach my design goals,
I have calculated this stack up, with these spaces.
the problem is that,I can not make 40 ohm single ended, because the trace width would become too large(around 10mils)! so I can not route the board efectively, I need the trace to be around 4-6 mils, any Ideas what should I do?
ASiDesigner, Stands for Application specific intelligent devices
I'm a Digital Expert from 8-bits to 64-bits
 

Offline Miyuki

  • Frequent Contributor
  • **
  • Posts: 837
  • Country: cz
    • Me on youtube
Re: 6 layer PCB stack up
« Reply #1 on: April 29, 2020, 09:00:31 am »
And increasing impedance ? Or you have to go to really high clock rates ?
Check signal calculation for 50 Ohm
Or even higher, all depends on distance and clock needed to keep eye diagram enough
« Last Edit: April 29, 2020, 09:02:10 am by Miyuki »
 

Offline TheUnnamedNewbie

  • Super Contributor
  • ***
  • Posts: 1184
  • Country: 00
  • mmwave RFIC/antenna designer
Re: 6 layer PCB stack up
« Reply #2 on: April 29, 2020, 09:48:24 am »
A single ended microstrip/stripline is always gonna give you much wider traces than a differential with two time that impedance, because you don't have all that capacitance of the nearby trace to make the thing smaller.

If you really need single-ended transmission lines, consider using something like CPW.
The best part about magic is when it stops being magic and becomes science instead

"There was no road, but the people walked on it, and the road came to be, and the people followed it, for the road took the path of least resistance"
 

Offline OwO

  • Super Contributor
  • ***
  • Posts: 1250
  • Country: cn
  • RF Engineer.
Re: 6 layer PCB stack up
« Reply #3 on: April 29, 2020, 10:42:14 am »
DDR3 system impedance is scalable. I haven't tried this yet but you may be able to use 60ohm traces if you change the on-chip termination calibration resistors (Zq on the dram chip side and VRN/VRP on the FPGA side).
Email: OwOwOwOwO123@outlook.com
 

Offline ali_asadzadeh

  • Super Contributor
  • ***
  • Posts: 1755
  • Country: ca
Re: 6 layer PCB stack up
« Reply #4 on: April 30, 2020, 06:09:36 am »
Quote
If you really need single-ended transmission lines, consider using something like CPW.
whats CPW?

Quote
DDR3 system impedance is scalable. I haven't tried this yet but you may be able to use 60ohm traces if you change the on-chip termination calibration resistors (Zq on the dram chip side and VRN/VRP on the FPGA side).
Oh, that's good advice, and I have not tried it yet!

my DDR3 speed is 1866, so xilinx recommended 40 Ohm traces |O
ASiDesigner, Stands for Application specific intelligent devices
I'm a Digital Expert from 8-bits to 64-bits
 

Offline cgroen

  • Supporter
  • ****
  • Posts: 629
  • Country: dk
    • Carstens personal web
Re: 6 layer PCB stack up
« Reply #5 on: April 30, 2020, 06:50:55 am »
Quote
If you really need single-ended transmission lines, consider using something like CPW.
whats CPW?

Quote
DDR3 system impedance is scalable. I haven't tried this yet but you may be able to use 60ohm traces if you change the on-chip termination calibration resistors (Zq on the dram chip side and VRN/VRP on the FPGA side).
Oh, that's good advice, and I have not tried it yet!

my DDR3 speed is 1866, so xilinx recommended 40 Ohm traces |O

https://en.wikipedia.org/wiki/Coplanar_waveguide
 

Offline ConKbot

  • Super Contributor
  • ***
  • Posts: 1331
Re: 6 layer PCB stack up
« Reply #6 on: April 30, 2020, 01:50:53 pm »
A single ended microstrip/stripline is always gonna give you much wider traces than a differential with two time that impedance, because you don't have all that capacitance of the nearby trace to make the thing smaller.

Only for strongly coupled diff pairs.  Once you start pushing up layer counts, or using thin dielectrics for other reasons, you start running into manufacturing limits before you can do a strongly coupled pair.  I.e. A 4mil dielectric can end up using ~4 mil trace and ~6 mil trace (don't have exact numbers in my head) which ends up with them weakly coupled. I.e. 90 ohm differential impedance, 25-30 ohm common mode impedance.
But at that point you're past, or at least toeing the line for what impedance approximation equations can do accurately and a 2D field solver becomes a lot more useful.
 

Offline OwO

  • Super Contributor
  • ***
  • Posts: 1250
  • Country: cn
  • RF Engineer.
Re: 6 layer PCB stack up
« Reply #7 on: May 08, 2020, 01:01:57 pm »
UPDATE: I just tested that DDR3 can be operated with 67ohm traces (~0.15mm width on 0.2mm prepreg), if Zq is set to 360ohms and Zynq on-chip termination resistors VRN/VRP are 120 ohms. This sets a 60ohm system impedance. Booting linux and running memtest for 2 hours so far hasn't shown errors, but I'll let it run for 2 days and post here again. This is running at 1066Mb/s.
Email: OwOwOwOwO123@outlook.com
 
The following users thanked this post: ANTALIFE, SiliconWizard

Offline ali_asadzadeh

  • Super Contributor
  • ***
  • Posts: 1755
  • Country: ca
Re: 6 layer PCB stack up
« Reply #8 on: May 09, 2020, 10:20:23 am »
Thanks Owo,But I'm targeting for 1866Mb/s :'(
ASiDesigner, Stands for Application specific intelligent devices
I'm a Digital Expert from 8-bits to 64-bits
 

Offline OwO

  • Super Contributor
  • ***
  • Posts: 1250
  • Country: cn
  • RF Engineer.
Re: 6 layer PCB stack up
« Reply #9 on: May 13, 2020, 07:19:11 am »
It passes memtest running for over 2 days, but so far it's only one sample. I will be building a few more samples later to test.
Email: OwOwOwOwO123@outlook.com
 

Offline ali_asadzadeh

  • Super Contributor
  • ***
  • Posts: 1755
  • Country: ca
Re: 6 layer PCB stack up
« Reply #10 on: May 14, 2020, 06:45:34 am »
Quote
It passes memtest running for over 2 days, but so far it's only one sample. I will be building a few more samples later to test.
Thanks for sharing, what FPGA or CPU are you using? any DDR part number to share?
ASiDesigner, Stands for Application specific intelligent devices
I'm a Digital Expert from 8-bits to 64-bits
 

Offline OwO

  • Super Contributor
  • ***
  • Posts: 1250
  • Country: cn
  • RF Engineer.
Re: 6 layer PCB stack up
« Reply #11 on: May 14, 2020, 07:10:19 am »
Zynq 7020. H5TQ4G63AFR. Get the -TEC suffix (2133 speed grade) if you want to go 1866Mb/s. Always get DDR3 components faster than the speed you are targeting, this way you will have more margin and a wider eye.
Email: OwOwOwOwO123@outlook.com
 
The following users thanked this post: ali_asadzadeh

Offline ali_asadzadeh

  • Super Contributor
  • ***
  • Posts: 1755
  • Country: ca
Re: 6 layer PCB stack up
« Reply #12 on: May 15, 2020, 10:24:04 am »
Thanks OwO, is it a 4 layer job again? ;D ;) also open source or closed?
ASiDesigner, Stands for Application specific intelligent devices
I'm a Digital Expert from 8-bits to 64-bits
 

Offline OwO

  • Super Contributor
  • ***
  • Posts: 1250
  • Country: cn
  • RF Engineer.
Re: 6 layer PCB stack up
« Reply #13 on: May 16, 2020, 05:37:50 pm »
https://github.com/gabriel-tenma-white/sdr6

New board, same DDR3 layout as sdr5. This time JLC doesn't offer PCBA on 0.1mm prepreg boards anymore, so I went with 0.2mm and adjusted resistors.
Email: OwOwOwOwO123@outlook.com
 

Offline ali_asadzadeh

  • Super Contributor
  • ***
  • Posts: 1755
  • Country: ca
Re: 6 layer PCB stack up
« Reply #14 on: May 17, 2020, 05:37:39 am »
Thanks for sharing,I have made suggestions in github  ^-^
« Last Edit: May 17, 2020, 05:40:37 am by ali_asadzadeh »
ASiDesigner, Stands for Application specific intelligent devices
I'm a Digital Expert from 8-bits to 64-bits
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf