Critique on PCB layout wanted

[permal]

Here's you chance to smack my fingers and tell me all the wrongs I've done in this layout. Please keep it friendly

For various reasons I need to connect both the SPI and MMC interface to my MCU. To do that I opted to use a MAX4948 which will be controlled by a switch (previously discussed here), with the MMC interface on the NC-path. The wires are all matched to the same length between MCU/MAX4948 (40mm) and MAX4948/SDCard (31mm) holder, except for the CLK signal that is 1 mm longer (I read that was important somewhere). The image only contains the signal wires, no ground planes (obviously) or power.

Personally I think there are too many wires crossing, but I can't move the pins on the MCU and no matter how I turn the MAX4948, there are always some wires that gets crossed so this is as few as I can get.


Thank you in advance.

[ataradov]

What is the MCU? Are you sure it can go anywhere close to the frequencies at which length-matching does anything?

I would say that you 100% don't need it.

Where are the board outlines?

With all the meandering, it is hard to tell, but it looks like the number of vias can be reduced.

[permal]

The MCU is an ESP32.

If I understand it correctly, it can go up to 40MHz with the MMC ifc.

[ataradov]

This is not fast enough on such a small board, don't worry about it.

[In Vacuo Veritas]

What is it with n00blers putting length matching everywhere? 

Do you also put thrust-reversers on bicycles?

[chrisl]

Get rid of the length matching.. don't need one at this speed.
Also please post the schematic and the board stack-up as well if you want meaningful feedback.

[mc172]

Also please post the board stack-up as well if you want meaningful feedback.

It's a double sided PCB. What meaningful feedback could you possibly give?

[permal]

What is it with n00blers putting length matching everywhere? 

Do you also put thrust-reversers on bicycles?

We're noobs, still learning, oh great master 

And what's wrong with thrust-reversers anyway?

---

The consensus seems to be length matching isn't needed, thanks for that feedback. 

[MiDi]

I have a bodged wired SPI (~20cm) @80MHz on an ESP8266 and it runs fine since months...
The crossings with large overlay could be a bigger problem than the length -> crosstalk.
Signals on a board run at ~2/3 of speed of light, so ~200E6 m/s.
A 1cm difference turns into 50 pico seconds 

[stmdude]

Seems to lack a bit of VDD, decoupling, ground-plane (might be outside the screenshot though)

The amount of vias seems fairly okay. The pinout of the MAX chip isn't doing you any favors.

As for the meandering.. As others have pointed out, it's kind of not really needed. However, you've learned something in the process (clock-line should be longer than the other lines), and if you think it looks cooler, keep it.

[mc172]

Some of the tracks are very close to the pads. Make it easy for yourself - if you had to hand solder that, you'd probably short out pins 14 to 19 on the larger IC because of the tracks needlessly close to them.

[The Doktor]

Some of the tracks are very close to the pads. Make it easy for yourself - if you had to hand solder that, you'd probably short out pins 14 to 19 on the larger IC because of the tracks needlessly close to them.

Wouldn't the solder resist prevent this?

[coppercone2]

What is wrong with length matching? Its good practice so is CMRR.

why encourage a lazy attitude?

If your gonna say don't do it at least give a analysis number on the protocol degradation %. Instead of doing basic communications engineering calculations to educate you just call some guy a noob. wtf

the fact of the matter is, it may work, but at what capacity and under what conditions, how does interference susceptibility change based on mismatch, all sorts of issues can be analyzed here instead of answers like that

[Simon]

Right, thread sort of tidied up, now play nice please!

[mc172]

Some of the tracks are very close to the pads. Make it easy for yourself - if you had to hand solder that, you'd probably short out pins 14 to 19 on the larger IC because of the tracks needlessly close to them.

Wouldn't the solder resist prevent this?

Yes, if you get a board made rather than roll your own. There is nothing forcing the tracks to be so close to the pads so why make it more difficult than it needs to be and rely on the solder resist?

[FotatoPotato]

My best advice is to try to make it a bit more compact, there's lots of wasted space where the traces are, and also while in many cases length matching is a good thing and when you are dealing with high speed data that needs to get to the same place at the same time, it doesn't seem like its needed in this case, lastly I would add a flooded ground plane on both layers to keep the noise low and to add heat sinking for the ground pad on your main IC (Sorry I cant see the name underneath the traces).

As an after thought, I always like to make 3 - 5 versions of the same board, I always seem to make the next one better than the last. I would urge you to do the same as it gives you extra practice and the end result its usually better than the first try . Also, try to use only one layer. It might be impossible and that's ok, but its always a good idea to challenge yourself to route smarter and reduce layer count.

Good luck!  :