My first pcb will include two SMD voltage regulators. Everything else is THT.
I made a screenshot of the power supply. On the pads to the right of U1 (5V) and U3 (3.3V) do I need solder mask? The copper pour on top will be 3.3V.
On the ground pads I added vias. Is that the correct thing to do? I’m in unfamiliar territory so I want to be sure I get this right.
Not sure what you mean as solder mask is negative in PCB design software.
You need pad like rectangles at solder mask layer to have no solder at these pads as these elements are solded there. I think in many voltage stabilizers these pads are connected to GND.
When I use 78M05 in DPAK I put 8 vias around its biggest pad jut to have a good thermal connection to GND for heat dissipation.
As @Piotr mention, the SolderMask layer is a negative. This means that there should be a SolderMask graphic over these tabs.
And, again, run into the topic of whether or not these tabs should be called tabs, or given separate pin numbers. Most common voltage regulators of this package have the center pin and the tab physically connected together.
ON EDIT: I didn’t look close enough, there is no copper pour at the moment.
Thanks for your feedback. They are SOT_223 and they are the only devices I don’t have any experience with. I will order a supply to test physically before I finalize the board.
Is a via on the ground pad correct?
That’s what I wanted to know. It would make sense that they would be connected to ground.
I deleted both copper pours to make corrections in the traces.
I’ll edit the board for another week or two before I have it made. I will have a second version but for this board I want to make sure there are no fatal mistakes.
I understand that you plan to test if that 4-th pin is connected to 2-nd pin. It is not written in their pdfs? What type of stabilizers do you use?
If you have GND at bottom (no info about it in what you have written) then it is OK to connect GND pin to GND by via. In my designs all bottom is GND (no other connections). I put via to GND as close to GND pads as reasonable (via copper circle touches the pad).
Thank you. I will follow that procedure from now on.
I built the power supply based on the schematic in the manual. I sometimes replace 1K uF caps with 10 uF tantalums based on the recommendation of an engineer who has helped to guide me for a number of years. Both electrolytic and tantalum filter capacitors work well.
The wide pin (tab) on the right side of your voltage regulator is big for a reason.
It is meant to be soldered to a copper pour which acts as a heat sink.
SMD packagages overheat very easy, and with a bit of copper around it you can increase it’s current capability without overheating 10x or more.
Often this is also combined with “thermal via’s”, an array of closely spaced via’s which connect coppoer pours on top and bottom side, which again almost halves the thermal resistance.
I see some bulk elco’s, but I do not see any decoupling capacitors, and these are almost always mandatory.
Your via’s look very small. Probably the 0.4mm default I usually make them a bit bigger.
These are about the coarsest SMD packages you can find, and therefore a good start.
For soldering I recommend first heating a pad, and then putting a bit of solder on it and remove the iron as fast as possible to prevent the flux from evaporating.
Then take some good quality tweezers to hold your SMD and align it on the Footprint.
Then re-heat the pad with solder to make one connection. After that you can let go of the tweezers and solder all the other pads.
You can also re-visit the first pad with your soldering iron then. Remember that the first connection is only for keeping the SMD at it’s place. I use the same technique with QFP’s and other high pin count components, and with the first solder connection I often short 2 or 3 pins together. I don’t care, the goal is perfect alignment of the chip on the footprint.
After the other pins are soldered the chip won’t move again and the short can be easily fixed.
With higher pin count components, using extra flux is mandatory, but with these big pads you can get away without it. If you have trouble soldering a pin, then remove the old solder with solder wick and apply new solder (with built-in flux).
In 2002 (I think) when earthquake made tantalums difficult to get I switched from tantalums to ceramics. At voltages like 3V3, 5V I use ceramics X7R or X5R (like 22uF X5R 0805).
Ceramics have wery low ESR. It is generally good except you have to check (read in datasheet) if your stabilizer is stable loaded with Low ESR capacitor. Standard stabilizers are stable. Low-drop - olders not, newest yes.
I suggest to use SMD elements. Assume 0603 (or 0805) as your basic standard. It is not a big problem to assemble them manually.
It’s probably difficult to see my respect for your knowledge because I’m still new to the forum but I am paying attention.
I remember when I was first learning to program the Parallax Propeller microcontroller. I was in way over my head and ready to quit but a professional engineer from that forum guided me through the entire project.
I will never visit Australia but if I did I would have to shake his hand. He was a tough and demanding teacher but he stuck with me.
My second PCB will have SMD elements that I can solder with a fine tip. I need the confidence of successfully designing my first PCB.
Everything you wrote in this post is essential info. Thank you.
edit: On this board I will add 1uF decoupling caps near VDD and VSS on both sides of the chip.
Don’t worry. I need not to see it 
And here you are better. I have never written any program for microcontroller (it is my brothers task, I write only for PC).
1uF seems a bit large. A more typical value is 100nF but with faster MCUs it might even be better to use 10nF
Look at the frequency vs impedance curve of your chosen cap and make sure that its resonant frequency (the place where it stops being a cap and starts being an inductor) is larger than the frequencies expected to be produced by your system. (Remember trapezoid signals like found in digital circuits have quite high frequencies in them mostly depending on the speed of the voltage [du/dt] change not really the frequency of the signal itself.)
And also remember that the size of the package does influence this heavily. The smaller the package the higher the resonant frequency (typically).
This is also why you sometimes see suggestions of having a 1u cap in parallel to a smaller cap (The large cap should then have a higher ESR to avoid introducing another set of high impedance points. This is why you might use tantal caps for the larger of the two.)
The 1u cap then takes care of low frequencies while the 100n (or 10n) takes care of the higher frequencies. Make sure the smaller cap is nearer to the supply pin.
1 Like
Busted! Yes, the decoupling caps are 100nF. I had to take a closer look at one of my existing boards. Experienced people don’t miss those details.
Make no mistake, experienced people do. They just have more robust workflows to then catch those details before passing the work to the next person. But even then, occasional errors slip through giving the next person something to do. 
I could never have learned to program an 8748 in 1987 with the resources available to you at that time! Absolutely amazing.
Just a month ago it took me and two other people a full working week to find out that the main tantal cap was soldered in wrong (Remember the mark on a tantal shows the positive terminal!)
You provoked me to write something about those times.
With 8748 we designed our first product - the EPROM programmer called “Piccolo” (first piece sold in 1988).
- Our programmer was very small compared to others. They used controller with sofware in external EPROM so they needed external latch to drive bus and they had too few microcontroller pins left so they needed to add external buffers to drive/read all textool socket pins. So the whole system instead of one microcontroller but even so our solution was more expensive (but we named our firm ‘MicroMade’ so we were obliged to do small products )
- To upgrade the programmer (for new EPROM on market) their users had to send them the programmer and they erased and then programmed the EPROM. We parametrised all programming processes so we just published the new software versions. Software from 2008 could work with 20 years old device - no problem.
With next our product - ‘Picco-GAL’ - the GAL & serial EEPROM programmer we have done something I am proud of. ‘Piccolo’ knew how to program EPROMS only programming was parametrised (how many and how long pulses, at what pin, what voltage, and so on). Picco-GAL know nothing about programing GALs or serial EEPROMS. Picco-GAL had an FORTH like language interpreter and the programming algorithm was send to it and then interpreted. But we had oly 128 bytes of 8751 internal RAM.
We divided it for:
- microcontroller stack (8 bytes + 2 bytes for each loop level)
- FORTH stack (data manipulated by FORTH commands - loop sizes, pin numbers,…) (depend on program - about 6…12 bytes)
- program (not sure but max probably about 114 bytes)
- programming data stack (bytes to be programmed in GAL, EEPROM) (typical page - 16 or 32 bytes)
1 and 2 rised agains each other, and 3 and 4 rised against each other. The division betwean these two sets was dynamically controlled by FORTH program. As program typically has some introduction to be done before it starts its main loop it was standard that when it comed to load first page of data to be programmed that data overwrited program introduction.
For some more complicated GALs the program had to have two parts. The first ended with instruction to load and run the next one. But the main loop always had to fit in that second part.
Imagine now the real task makeing program being 100 bytes long
That’s something only an engineer could do.
I took a look at the Picco-GAL webpage translated into English… That’s genius level stuff.
I didn’t know there is such webpage
When we (Poland) joined EU and from 1.05.2004 we were not allowed to sell anything without CE we just canceled over 30 devices we had those time in our offer. Picco-GAL was designed about 1991 when I know nothing about EMC. I know it will not pass ESD tests. We decided to not redesign it so since then we don’t sell Picco-GAL.