What bothers me about some of those boards including the guitar switcher you did is you put the U numbers under the IC so once the part is soldered on the U numbers are hidden.
I don't like to put the component ID under the component. It's useless except for hand assembly. You can't check the assembled boards or troubleshoot them easily. I see you leave all the extra copper in place. Excellent. It makes no sense to remove it, removing it uses up etchant, and once grounded, all that extra copper makes the boards quieter and helps keep the ground at ground.
I agree about the component ID. I prefer to put component values under the component and the ID's next to the component. When assembling the board the component values are useful and the fact that they are covered up after the part is installed is of no consequence. Most of the default KiCAD footprints put the component ID under the component and the value next to it which is opposite of what is useful. As having as much flood fill as possible, this is ok for low speed analog designs, but for high speed digital circuits, I would recommend avoiding this as it is easy to unwittingly create Ghz order resonators and antennas on a board with these islands of copper between traces. Even a 100mhz digital circuit will have fast enough edges to excite Ghz EMI.
Good explanation of PCB design / production techniques. Re PCB sizing (and mounting), I always tend to start with the intended enclosure before even starting PCB design. Unless your runs are high in quantity and you have the money for injection moulding, then I find it's best to stick with off-the-shelf enclosures.
Love on that last board that agilent decided they needed to slap not one, but two LM399s into the design :D My suggestion is, unless other requirements preclude it, may as well always give boards slight rounded edges and some screw holes for future mounting possibilities.
All those signal labels are a solved problem in KiCad. Create a custom copy of a footprint and use ${} variable references to get signal names as values of variables from the schematic component. Then for the schematic component put the signal names on the pins and on the variables names like, say, Pin01, …, Pin28. Then in the footprint the text field that had ${Pin01} value will show the name of signal on pin 1 etc. I do this all the time with edge connectors on PDP logic modules and similar modular logic. The footprint for the edge conn has all the names laid out nicely and just picks up variables from the schematic symbol. The schematic symbol can be generic just have those signal legend fields next to each pin. On the schematic you name them according to signal names and done. The rest syncs up. Same goes for things like ${REVISION} on the legend that shows the revision number etc. Use the variables :) I do the same with positioning of references: got a bunch of copied footprints in a custom library that has the references positioned the way that will make most sense in a given layout and just use those. Can also be used to pick a uniform custom font for references. I also like silkscreened component values on low density boards so that goes into the footprints and gets handled automatically. There’s also lots that can be done using Python scripts on the PCB.
When I first learned PCB design, my first year in college in 1990, we just made rudimentary designs. We would print our design on a clear plastic sheet with a laser printer. We would then use a flat iron to transfer traces to copper-clad fiberglass boards. We would often have to manually touch-up a few spots on the board with a pen. We’d then etch the board with that nasty chemical etchent liquid. We’d then rinse and use fine steel wool to rub off the black stuff on the copper (the mask). We then used a tinning solution that would chemically plate copper to prevent oxidation. We drilled our holes manually. We knew there were better processes then but this was how we’d prototype.
i did love to see SMD components flying off a PCB when doing a resonance search on equipment. Amazing to see how much a PCB can bend when it come in to resonance and how unforgiving smd chips are to PCB moving, stayed with TPH for quite a while as they handle a flexing PCB way better. But the aim was to make sure any resonances in the equipment was away from common vibration frequencies known to exist on what the equipment was mounted on.
Half the time it seems no one is spending time to arrange references on the through hole designs. On dense designs it’s like with SMT layouts: sometimes the ref just doesn’t fit and needs to go elsewhere with a leader line.
maybe I'm not explaining it well. if you cut a trace with an exacto it does not act the same as the 1980 boards. those boards had the copper directly applied, modern boards use an extra layer (same process for many layer boards) and that adds some softness
@@IMSAIGuy I wonder if what you are experiencing is on modern boards there is normally a solder mask that you have to cut through as well as the trace?
no it is some layer under the copper. think of a 4 layer board. most boards just have two outer layers. so there is some insulator between those two and I think adding the 4th layer is the same as adding the 3rd to the FR4 if that makes sense.
Re: auto route, the designer spends most of his/her time setting up the parameters and nets. Then a lot of time getting those last few nets routed which the auto was unable to route. For me the death-nell mistake are those getting all the footprints correct. It really pisses me off. Some many details. Sometimes the best thing is to have the damn thing made and count on having to re-run after making corrections.
I agree about the footprints. I keep getting boards back where I made some mistake because of a detail about the footprints. One of the challenges that I have is that on the screen you usually have little idea of how big things actually are. I think I may start making scale printouts of the copper traces so I can judge how part placement will go before sending things off to the fab.
@@argcargv Well if you have the stomach for it I suggest: Print the completed design 1:1 then place parts on the print to check the footprint. This can help but is only half the story. The other half is: "are the pin numbers correct?" This assumes you have ordered the parts and have them in hand.
@@mr1enrollment Yes, I have run against both problems, getting the PIN numbers incorrect and getting a similar but smaller dimension part. I thought I was doing good to just double check the PIN numbers but then I found that the mistakes can also be dimensional. There are many times that I have had to have a second revision of a board because of footprint problems or scale problems where clearance or something like that doesn't make sense. The design rules all check out, and the 3D vis looks good, but when you actually get to assembling things you find out all of the things that weren't checked out.
Agree on the issue with footprints, if there is anything that makes PCB design most problematic / time-consuming then footprints are it. The few online libraries include multiple errors, especially with dimensions. In my opinion, all the most significant component distributors should include links to accurate footprint files (not Library Loader, which is some sort of joke).
What bothers me about some of those boards including the guitar switcher you did is you put the U numbers under the IC so once the part is soldered on the U numbers are hidden.
I don't like to put the component ID under the component. It's useless except for hand assembly. You can't check the assembled boards or troubleshoot them easily. I see you leave all the extra copper in place. Excellent. It makes no sense to remove it, removing it uses up etchant, and once grounded, all that extra copper makes the boards quieter and helps keep the ground at ground.
I agree about the component ID. I prefer to put component values under the component and the ID's next to the component. When assembling the board the component values are useful and the fact that they are covered up after the part is installed is of no consequence. Most of the default KiCAD footprints put the component ID under the component and the value next to it which is opposite of what is useful. As having as much flood fill as possible, this is ok for low speed analog designs, but for high speed digital circuits, I would recommend avoiding this as it is easy to unwittingly create Ghz order resonators and antennas on a board with these islands of copper between traces. Even a 100mhz digital circuit will have fast enough edges to excite Ghz EMI.
Good explanation of PCB design / production techniques. Re PCB sizing (and mounting), I always tend to start with the intended enclosure before even starting PCB design. Unless your runs are high in quantity and you have the money for injection moulding, then I find it's best to stick with off-the-shelf enclosures.
Love on that last board that agilent decided they needed to slap not one, but two LM399s into the design :D
My suggestion is, unless other requirements preclude it, may as well always give boards slight rounded edges and some screw holes for future mounting possibilities.
All those signal labels are a solved problem in KiCad. Create a custom copy of a footprint and use ${} variable references to get signal names as values of variables from the schematic component. Then for the schematic component put the signal names on the pins and on the variables names like, say, Pin01, …, Pin28. Then in the footprint the text field that had ${Pin01} value will show the name of signal on pin 1 etc. I do this all the time with edge connectors on PDP logic modules and similar modular logic. The footprint for the edge conn has all the names laid out nicely and just picks up variables from the schematic symbol. The schematic symbol can be generic just have those signal legend fields next to each pin. On the schematic you name them according to signal names and done. The rest syncs up. Same goes for things like ${REVISION} on the legend that shows the revision number etc. Use the variables :)
I do the same with positioning of references: got a bunch of copied footprints in a custom library that has the references positioned the way that will make most sense in a given layout and just use those. Can also be used to pick a uniform custom font for references. I also like silkscreened component values on low density boards so that goes into the footprints and gets handled automatically. There’s also lots that can be done using Python scripts on the PCB.
When I first learned PCB design, my first year in college in 1990, we just made rudimentary designs. We would print our design on a clear plastic sheet with a laser printer. We would then use a flat iron to transfer traces to copper-clad fiberglass boards. We would often have to manually touch-up a few spots on the board with a pen. We’d then etch the board with that nasty chemical etchent liquid. We’d then rinse and use fine steel wool to rub off the black stuff on the copper (the mask). We then used a tinning solution that would chemically plate copper to prevent oxidation. We drilled our holes manually. We knew there were better processes then but this was how we’d prototype.
i did love to see SMD components flying off a PCB when doing a resonance search on equipment. Amazing to see how much a PCB can bend when it come in to resonance and how unforgiving smd chips are to PCB moving, stayed with TPH for quite a while as they handle a flexing PCB way better. But the aim was to make sure any resonances in the equipment was away from common vibration frequencies known to exist on what the equipment was mounted on.
Fantastic overview.
If the FR4 is gummy, insufficient hardner is used in the mix
Yester days stencils are todays SM parts 😮 - and yes ibknoev
Half the time it seems no one is spending time to arrange references on the through hole designs. On dense designs it’s like with SMT layouts: sometimes the ref just doesn’t fit and needs to go elsewhere with a leader line.
Might that first one at 1:17 that's nice and hard be FR2 bakelite?
how about PCWAY
Never tried them. I first tried JLCPCB and was happy with the quality and stayed there.
I've used both JLCPCB and PCBWAY. They seem nearly identical.
I never run across gummy fr4. Both from US manufacturers or from China. Who is sending you gummy boards?
maybe I'm not explaining it well. if you cut a trace with an exacto it does not act the same as the 1980 boards. those boards had the copper directly applied, modern boards use an extra layer (same process for many layer boards) and that adds some softness
@@IMSAIGuy I wonder if what you are experiencing is on modern boards there is normally a solder mask that you have to cut through as well as the trace?
no it is some layer under the copper. think of a 4 layer board. most boards just have two outer layers. so there is some insulator between those two and I think adding the 4th layer is the same as adding the 3rd to the FR4 if that makes sense.
@@IMSAIGuy it's called prepeg it goes between a copper layer and core. it's what sticks everything together... i.e. copper - prepeg-core-prepeg-copper
Re: auto route, the designer spends most of his/her time setting up the parameters and nets.
Then a lot of time getting those last few nets routed which the auto was unable to route.
For me the death-nell mistake are those getting all the footprints correct. It really pisses me off. Some many details. Sometimes the best thing is to have the damn thing made and count on having to re-run after making corrections.
I agree about the footprints. I keep getting boards back where I made some mistake because of a detail about the footprints. One of the challenges that I have is that on the screen you usually have little idea of how big things actually are. I think I may start making scale printouts of the copper traces so I can judge how part placement will go before sending things off to the fab.
@@argcargv Well if you have the stomach for it I suggest: Print the completed design 1:1 then place parts on the print to check the footprint. This can help but is only half the story. The other half is: "are the pin numbers correct?" This assumes you have ordered the parts and have them in hand.
@@mr1enrollment Yes, I have run against both problems, getting the PIN numbers incorrect and getting a similar but smaller dimension part. I thought I was doing good to just double check the PIN numbers but then I found that the mistakes can also be dimensional. There are many times that I have had to have a second revision of a board because of footprint problems or scale problems where clearance or something like that doesn't make sense. The design rules all check out, and the 3D vis looks good, but when you actually get to assembling things you find out all of the things that weren't checked out.
@@argcargv welcome to the PCB world. ;-}
Agree on the issue with footprints, if there is anything that makes PCB design most problematic / time-consuming then footprints are it. The few online libraries include multiple errors, especially with dimensions. In my opinion, all the most significant component distributors should include links to accurate footprint files (not Library Loader, which is some sort of joke).