A few notes:
- Post a schematic; it will make reviewing this easier.
- It looks like you are using a constant current shift register; if so, you can't just parallel the channel indicator LEDs. If you are using a voltage-output part, you need current-limiting resistors
- As you are driving SSRs with this, put the channel indicator LEDs in series with the output to reduce driver current and heating.
- the ground pour is very swiss-cheesed by the signal traces going to the RJ45s
- traces should bend at 45 degrees rather than 90 degrees
- traces should connect to the centers of pads rather than generating acute angles which can form acid traps
- you don't have any mounting holes
- I'd move the ground vias for the driver chips right next to the pin. All of the LED current is flowing through it. As above, you also need to improve the ground pour from the power connectors to the driver chips. Moving the power connector to the top edge (above all the RJ45s) will help here
- make sure the driver chips really are in wide-soic packages; some may only come in narrow.
- beef up the traces for the regulator
/mike
Mike,
Thank you for the advice. This is by far the most complicated PCB I've designed so your advice is much appreciated.

- I'll likely just remove the channel LED indicators. I wanted to option the board to use a through hole or surface mount IC and I'll need more room to make that happen.
- I plan to drive Darlington transistors at the relay boards.
- My concern with a homemade board is that I will not be able to make a good solder connection if I run the traces to the topside of the RJ45 (and other) connectors. That's why I have all input and output traces run to the bottom side of the board. Please advise.
I'll get to work on the rest of the suggestions. I'm glad you posted because I was about to print off the PDF's and get started.
Here is the code I wrote to test the capability of a software generated PWM signal. This is at 250hz
#include <iostream>
#include <sys/io.h>
#define base 0x378
using namespace std;
int timeout;
int i = 0;
int runPWM(int powerLevel, int timeout)
{
ioperm(base,1,1);
//Loop to generate PWM at a given percentage
while(i != timeout)
{
outb(0,base);
usleep(4000-powerLevel*40);
//cout << i << "msecs
";
outb(255,base);
usleep(powerLevel*40);
i++;
}
return 0;
}
int main()
{
int powerLevel = 0;
cout << "Power level %: ";
cin >> powerLevel;
cout << "Time to run (in seconds): ";
cin >> timeout;
timeout = timeout * 250;
runPWM(powerLevel, timeout);
return 0;
}
I added the ability to change the frequency:
#include <iostream>
#include <sys/io.h>
#define base 0x378
using namespace std;
int timeout;
int i = 0;
int freq, freq2, freq3;
int runPWM(int powerLevel, int timeout, int freq2, int freq3)
{
ioperm(base,1,1);
//Loop to generate PWM at a given percentage
while(i != timeout)
{
outb(0,base);
usleep(freq2 - powerLevel * freq3);
//cout << i << "msecs
";
outb(255,base);
usleep(powerLevel*freq3);
i++;
}
return 0;
}
int main()
{
int powerLevel = 0;
cout << "Power level %: ";
cin >> powerLevel;
cout << "Time to run (in seconds): ";
cin >> timeout;
cout << "Frequency in herz?";
cin >> freq;
timeout = timeout * freq;
freq = freq ^ -1;
freq2 = freq * 1000000;
freq3 = freq * 10000;
runPWM(powerLevel, timeout, freq2, freq3);
return 0;
}
I hope to test this tonight in class. If not, I'll get it done tomorrow at work.
For a home-etch board, you do need to take into account the lack of plated thru holes. If you are OK with SMT, I'd go that route for all the parts since you don't need to drill holes for the leads. I'd put the parts on the same side of the boards as the pins of the connectors (back side if the connectors mount on top) as the solder pads for the connectors and chips end up on the same side of the board.
/mike
