[Arm-netbook] EOMA68-A20 2.7.4 pre-production prototypes received and working

[Luke Kenneth Casson Leighton]

On Wed, Jun 21, 2017 at 6:26 PM, Wolfram Kahl <kahl at cas.mcmaster.ca> wrote:
> On Wed, Jun 21, 2017 at 12:00:07PM +0100, Luke Kenneth Casson Leighton wrote:
>> .On Wed, Jun 21, 2017 at 9:35 AM, mike.valk at gmail.com
>> >>  in a stack you tell the factory what thicknesses you want, as well as
>> >> what material in between, and what thickness of that, too.  so you get
>>
>> > So it doesn't have to be a problem. As long as you control the layers
>> > that have traces have the same thicknesses.
>>
>>  technically correct but far too much risk and hassle.  you end up
>> tying the PCB layout to a specific PCB manufacturing factory.
>
> Are there high-frequency risks/problems with switching back and forth
> between the layer pair

 there are.  the more VIAs you have the more EMI there is.  R.F. (and
HDMI is R.F.) does not travel in a straight line: if you have an
abrupt change of direction (a corner or a VIA) the signal actually
tries to just keep going in a straight line!

 the best track layouts use curves not 45 degree transitions.  the
best layouts have no track changes at all, are as symmetrical as
possible, are completely surrounded symmetrically by the exact same
amount of space on either side, and the exact same number of vias on
both sides of the track.  and also are impedance matched in terms of
distance between the pairs, width of the tracks, *and* the distance
between layers *and* the dielectric constant of the insulation between
the layers.

it's a pretty heavy-duty amount of requirements.

l.