@budude: When I pour copper, DesignSpark automatically substitutes thermal relief pads (spokes).
You are correct, and I'm now considering the actual connectivity (after copper pour) vs what the tool sees until then. I think there are some real considerations, though. For instance, as I mentioned above, if I use top copper for a ground strap, will I be introducing ground loops? If I use a bottom copper strap, then that becomes an "artificial" blockage for one signal trace. See that in the attachment. But is it really? I now see that if I rely on the copper pour as a "hidden ground strap", then that same signal trace (on bottom copper) becomes a "hidden blockage" for the ground plane, possibly resulting in some coupling issues/bounce?? This is where the value of real experience becomes so obvious to me: I can understand the theory and argue for all sides at the same time. Without real practical experience I may be worried about things that just don't matter and miss the things that do.
BTW, I'm glad that it looks like I've won the "human vs auto-router" competition. However, I took me about 20 hours and the router completed in about one minute...
this:
vs this:
vs this:
Is the question I'm trying to ask clear?
Thanks.
I'm not sure I'm understanding what you are saying, but...the first picture (of three in post 21) is the best one of the three.
What I would do (if your tools allow it) would be to remove that ground trace on the blue layer and use the copper pour to replace it. And move every possible signal trace from the blue layer to the other layer, going to the extent of moving components around in the process. For example, I would try to move HDR3 and HDR4 so their traces can be routed on the signal layer, rather than going through to the blue layer. I realize that this can't happen for all of the signals and that this requires a lot of design time, but every bit will help.
One last point - you will usually capacitors connecting the two nodes of a crystal oscillator to ground. I don't know if this is necessary for the micro that you are using, but the absence of two capacitors from the crystal leads to ground does lok unusual.
That's looking much better. The back-side traces (blue) are still cutting many slots in the ground plane. By adding a few vias, you could shorten the part of the trace that's on the back side. For example. come out of the '595 pin 15 on the back and via to the front just after the Vcc trace.
If you are planning to go to another board, would it be better to use 2x8 headers and a 16-pin ribbon cable instead of RJ45s?
Double check the hole sizes on CONN1; they look a bit small to me.
I'd try to move HDR3 and SW1 down to where the I2C EEPROM is. Routing SDA/SCL topside is easier than dealing with 5 signals.
/mike
I am by no means an expert, but I *think* a ground is missing on IC2 (bottom left pin in the image from post 17).
tlorek wrote:I am by no means an expert, but I *think* a ground is missing on IC2 (bottom left pin in the image from post 17).
I believe that's 5V actually but good catch on that connection missing.
Yes, you did catch a misising connection. I'm amazed at the level of scrutiny folks apply here. Thank you! (no sarcasm intented - really, thank you!).
So I'm still working away at this project. I guess this is what it must feel like to be an artist - no creation is ever "the best I can do" but always offers additional ways to make it better. I'll have to just let go and send it to fab sometime soon.
I suppose I have a fundamental question: to me, the ground plane seems like an enormous conductor. The few traces I have on bottom copper are cutting that plane, but how much does it matter? If I look at my VCC power trace, it's just 50 mils everywhere. There is no place on the entire boards where there would be a virtual trace less than that size, even with the bottom copper signal traces I have. On the other hand, I'm wondering how much degradation of signal there will be when using a via on a signal line. I have several traces that will run at 20MHz clock. Isn't it worse to insert a via in those signal traces than allow for a longer cut in the ground plane?
This is where the art of layout eludes me - and why I have a respect for those that do this daily, elegantly, and efficiently.
Doing a schematic, checking it carefully and performing a DRC will catch those missing connections. Of course if you don't connect a pin in the schematic in the first place, no DRC will help find that! 😉
PCB layouts are indeed a learned art. I look back at my early efforts and shudder. Over 700 designs later, arranging the parts and routing has come much easier.
David_AVD wrote:Doing a schematic, checking it carefully and performing a DRC will catch those missing connections. Of course if you don't connect a pin in the schematic in the first place, no DRC will help find that! 😉PCB layouts are indeed a learned art. I look back at my early efforts and shudder. Over 700 designs later, arranging the parts and routing has come much easier.
David, how do you apply a DRC on a schematic? With the experience of 700 boards behind you i'm sure you'll be able to explain this very simply.
David_AVD wrote:Doing a schematic, checking it carefully and performing a DRC will catch those missing connections. Of course if you don't connect a pin in the schematic in the first place, no DRC will help find that! 😉PCB layouts are indeed a learned art. I look back at my early efforts and shudder. Over 700 designs later, arranging the parts and routing has come much easier.
Yes, I understand and agree. I did start with a schematic and have rigorously checked for correlation between it and the PCB layout. What I posted somehow was an "in between" stage.
I am now (again, again) "done" with this design. I have to admit that I was frustrated by those that challenged me to re-do the upper-left corner of the design, where I had a somewhat messy bit of bottom copper signal traces winding out from under the Prop chip. They were just reset and "expansion" signal pin lines (no need for them now, but maybe later) so I just jammed them in. I didn't see how I could avoid the mess, and like I said, they aren't crucial signal lines.
The DRC check ended up complaining about some pads. I looked again and decided that I should try to heed the advice given. How will I ever do that? Well, 5 hours later, I am done (again, again) and I was somehow able to remove all bottom copper signal traces from that region. I amazed even myself. It turns out that resistors are really good "free" jumpers across other traces, without any vias or extra pads. I surprised myself with that revelation.
Thanks for challenging me.
mrpackethead wrote:David, how do you apply a DRC on a schematic? With the experience of 700 boards behind you i'm sure you'll be able to explain this very simply.
You don't Andrew. A DRC is applied to the PCB. Such a n00b ! LOL
