First time poster here. Going through, it looks like people here really take the time to give quality advice on board design. Hopefully I can get the same awesome advice for my board!
My design is a modification of an existing, open-source design for a wideband oxygen sensor display. The original design calls for two boards - a logic board and a display board. My goal was to put everything on a single board for purposes of fitting it into an off-the-shelf enclosure and supplying people with a viable product, DIY or pre-made.
The design works already. What I’m looking for is advice/constructive criticism/wisdom on the board layout. I tried my best to group power components in one section, display components in one section, filtering components in one section, etc etc. I’m very much a newbie, though, so I’ll take any advice/constructive criticism/wisdom I can get
I’ve included a PDF of the schematic, a picture of the board, and the EAGLE schematic/board files. Hopefully this is enough to satisify all the people here, EAGLE users or not.
Is the LED display mounted on the bottom? You might have trouble if you need to desolder components that have pins under the LED.
I assume you are only using 3 digits.
The VReg is on the bottom and it looks like you have a power trace in the keepout area. I think the 317 has the heat sink connected to output so it’s probably ok though DRC should have flagged it.
500 ohm as a current limiter might be a bit high - how bright do you want your LEDs to be?
Your 8.2K base resistors on the transistors seems kind of high. I’d probably use 2.7K.
This is a design that is crying for surface mount.
Philba:
Is the LED display mounted on the bottom? You might have trouble if you need to desolder components that have pins under the LED.
I assume you are only using 3 digits.
The VReg is on the bottom and it looks like you have a power trace in the keepout area. I think the 317 has the heat sink connected to output so it’s probably ok though DRC should have flagged it.
500 ohm as a current limiter might be a bit high - how bright do you want your LEDs to be?
Your 8.2K base resistors on the transistors seems kind of high. I’d probably use 2.7K.
This is a design that is crying for surface mount.
The first revision will be through-hole. This is because I don’t have a market to justify SMD - the only way I’d go SMD is if I got the boards assembled for me, and that will be expensive unless I’m doing large runs. The through-hole also works as a lowest common denominator for DIY as not everyone is able to or willing to solder 0805/0603 pitch parts.
Yes, the LED display is on the bottom. Yes, it will be a slight pain to take out components on the opposite side of the display. I view this as something people will have to accept if they want a low-priced wideband display. This is also why I’m trying to get the board design very nice to avoid problems that would require component replacement. also, I’m using all 4 digits.
The heatsink on the regulator is indeed the output so the power traces need to connect to it.
As far as the resistor values, I’ve yet to receive/use my reference display kit to see what the brightness is like. If need be, I will change the values.
Besides that, anything else you think could be done differently on the board?
One thing I have found is that its much more useful to omit part reference numbers (R1, D1, etc) and just include the actual values (580, 8.2K) or real part number (1N4002). When its time to assemble, its a royal pain to have to keep looking at the schematic or parts list to remember what the actual parts are supposed to be.
Typically when I am including a display (which is often the most expensive component), I will socket it. For a production board it probably won’t make sense, but for a prototype it can be useful. In your case, it would also allow you to put the display on the front since it the pin header would effectively lift the LED above the other components.
Philba:
it’s funny, I find surface mount easier to use than through-hole.
You don’t HAVE to connect a trace to the heatsink.
I’d rather do all SMD myself, but to solder them all by hand would be a pain, and releasing an SMD DIY kit could very well spell total financial disaster for me.
Vraz:
One thing I have found is that its much more useful to omit part reference numbers (R1, D1, etc) and just include the actual values (580, 8.2K) or real part number (1N4002). When its time to assemble, its a royal pain to have to keep looking at the schematic or parts list to remember what the actual parts are supposed to be.
Typically when I am including a display (which is often the most expensive component), I will socket it. For a production board it probably won’t make sense, but for a prototype it can be useful. In your case, it would also allow you to put the display on the front since it the pin header would effectively lift the LED above the other components.
I actually thought up until yesterday that I was stuck with the text placement of the existing symbol in EAGLE. I didn’t know you could smash them out and put them where you want. I actually plan on rearranging that stuff and just putting values. I wanted to keep the part numbers so people could more easily refer back to the schematic if they were critiquing the board.
I like the thought of socketing the display itself for prototyping.
TheDirty:
LM317 is not a automotive spec regulator. You might want to try something with internal protection that is speced for automotive use, like a LM2937.
You can then remove the trim resistors, flyback diode and reverse protection diode.
I could really remove all that? That’s pretty damn sweet. At ~$1, that looks like a good move for an automotive product. Thanks for the tip; I think I will swap that in right now.
Removing the reference designators would make troubleshooting that much harder. Maybe place the refdes next to the part and the value inside the silkscreen body?
I make my boards with part number but not values. If you are trouble shooting, having the schematic is invaluable. When building, I don’t find it a problem to have the schematic available. I know nathan does it differently, each to his own way.
Same here. When I’m assembling boards I include the component values on a printout of the silk screen layer, which minimises the chance of making a mistake.
Philba:
it’s funny, I find surface mount easier to use than through-hole.
You don’t HAVE to connect a trace to the heatsink.
I’d rather do all SMD myself, but to solder them all by hand would be a pain, and releasing an SMD DIY kit could very well spell total financial disaster for me.
Vraz:
One thing I have found is that its much more useful to omit part reference numbers (R1, D1, etc) and just include the actual values (580, 8.2K) or real part number (1N4002). When its time to assemble, its a royal pain to have to keep looking at the schematic or parts list to remember what the actual parts are supposed to be.
Typically when I am including a display (which is often the most expensive component), I will socket it. For a production board it probably won’t make sense, but for a prototype it can be useful. In your case, it would also allow you to put the display on the front since it the pin header would effectively lift the LED above the other components.
I actually thought up until yesterday that I was stuck with the text placement of the existing symbol in EAGLE. I didn’t know you could smash them out and put them where you want. I actually plan on rearranging that stuff and just putting values. I wanted to keep the part numbers so people could more easily refer back to the schematic if they were critiquing the board.
I like the thought of socketing the display itself for prototyping.
TheDirty:
LM317 is not a automotive spec regulator. You might want to try something with internal protection that is speced for automotive use, like a LM2937.
You can then remove the trim resistors, flyback diode and reverse protection diode.
I could really remove all that? That’s pretty damn sweet. At ~$1, that looks like a good move for an automotive product. Thanks for the tip; I think I will swap that in right now.
** Because the LM2937 is a LDO (Low-DropOut) regulator, it is “very fussy” about its output capacitor. The output cap’s ESR must be in a specific range. See page 7 of the datasheet at - http://www.national.com/ds/LM/LM2937.pdf
As far as the link to the EAGLE files, if you copy the whole line and put it into you address bar, it works. phpBB apparently sucks as handling links with ( or ) in them.
I did see the datasheet and the section regarding ESR on the output capacitor. I believe I found a suitable tantalum capacitor for the time being. Once I can wrap my head around all the math, I’ll most likely switch to an alum. electro/tantalum cap pair rather than a single, big tantalum cap. (like the datasheet suggests)
http://www.vishay.com/docs/40005/593d.pdf is the tantalum cap I’ve selected. It looks like it should provide sufficiently low ESR in colder temperatures for the LM2937.