For 99% of all cases this design is the right way to start with a '328 design...
And, if using a breadboard, please start neat and tidy and keep your wires short. This is the above circuit, without the reset switch and IO connections to the ISP pins...
And as a PDF.
The ISP connectors viewed from the top of the connector looking down towards the PCB...
I would add pins and pinout to ISP.
David
I would add pins and pinout to ISP.
Good idea. Done.
Brian, Your ISP connector (first schematic) looks a bit odd, it seems to imply all the pins are shorted together.
Other than that, a very nice posting!
Jim
gchapman 
Microchip's reference for AVR :
up one level to
Unused XTAL Pins
Also, show an optional Xtal for Uart operation
Brian, Your ISP connector (first schematic) looks a bit odd, it seems to imply all the pins are shorted together.
It's just a style thing, all my schematics, even the CAD ones, have that style of connector.
Also, show an optional Xtal for Uart operation
Done.
Thanks for the suggestions. I'm probably going to stop there as any further tweaks will take it away from my 99% goal.
(optional) filter inductor for AVCC ?
The reset circuitry looks more complex than I've seen in real life. "Decrease life of switch"? Because of those harsh 0.25uJ discharges?
Hi Brian,
Why do you have 4.7k ohm resistors in series with the SPI signals? For impedance matching to to typical PCB traces, it is more common to use values around 33-39 ohms (allowing some 10-15 ohms for the on-chip driver impedance). At 4.7k ohms, parasitic capacitances may severely limit the speed of the SPI bus.
Also, I'm not sure about this device - I exclusively have used XMEGAs - but the RESET# signal is used by the PDI, and putting capacitance on RESET# will strangle PDI operation. Does the ATmega328 do something different?
Yes - impedance matching! Not for transmission line effects but so that the ISP signals can override other devices connected to SPI when programming the flash.
Yes - impedance matching! Not for transmission line effects but so that the ISP signals can override other devices connected to SPI when programming the flash.
Ugh. It appears that my decision to stick with the XMEGAs was a good one. I can get maximum speed out of the SPIs. Thanks for the explanation.
jgmdesign 
Yes - impedance matching! Not for transmission line effects but so that the ISP signals can override other devices connected to SPI when programming the flash.
Why do you have 4.7k ohm resistors in series with the SPI signals?
I myself use 220 ohm resistors in series, but Hey Ho, whatever works.
JIm
gchapman Greater than 220 ohms is recommended though if the attached SPI device is de-selected and not in tri-state then 4K7 may work for an Atmel-ICE.
Shared Use of SPI Programming Lines | AVR® Microcontroller Hardware Design Considerations (AVR042)
...
Typically, the resistor value R can be of 330Ω.
...
These typical values are used to limit the input current to 10 mA for a supply voltage (VCC) of 3.3V. It may vary depending on the programmer/debugger used and the requirements of specific hardware design.
...
Pull-up Resistors | AVRISP mkII
... i.e. the resistor should have a value of 820Ω or more.
...
edit : typo
jgmdesign I was not saying my 220 ohms were used for pullup. But for isolation of downstream devices from the ISP when used.
Jim
I have a (stupid I'm sure( question on this ISP connector pinout.
I have the Atmel ICE package, the ATMega328 Explained board (and a couple others).
Looking at the schematic and the pinouts, I traced everything out and it seems to match. The problem is the connector itself that comes with the ICE.
(we called them "Berg" connectors back in the day).
The connector block itself has a small, hard to see, arrow denoting pin on. This doesn't match anything.
The connectors pin on would (looking down on to the connector from the top) connect to pin 5, or 2, depending on which way you plugged it in.
Plus, I notice that the Number One wire, the red stripe, actually connects to "3" (again, looking down), which is next to the pin one (would be on the left).
I'd like to get this clarified before I plug this thing in. Without a key to ensure you plug this in only one way, it's a might confusing.
The PCB's, all match up, pin one being in the lower left, or upper right corner, depending on board orientation. (which matches the drawing here)
I can post photo's if that would help clarify things, but that little arrow on the connector may not show up as it's very hard to see.
EDIT..... never mind, I just looked at the ICD documentation, (don't ask me why I didn't think of this before....) and Atmel ignores the pin one marker on the connector. That pin that is marked as number one, is indeed number 5 by Atmel configuration.
Sorry.
awneil 
Notes if you want to use a debugger:
Nice.
You might consider adding an LED WITH a series resistor, so the user can see the results of their work.
JC
Can I add another photo, from a 2015 Thread:
The microcontroller has its own Internal clock.
If this is used then no external clock crystal is required.
The internal (RC Oscillator) isn't very accurate, but can be calibrated.
Shown below are other options for providing the clock for the microcontroller.
#1 Ceramic Resonator Usually 3 leads, the center being ground.
It usually has capacitors built into it.
It is inexpensive, and more accurate than the uC's internal clock.
The one shown is 20 MHz.
#2 SMD "Crystal" Also called an Xtal for short.
This is a two lead device. The one shown is 14 MHz.
External capacitors are required.
Crystals use active circuitry inside the chip to make an oscillator, (clock).
#3 "Clock Crystal" These are, by definition, 32.768 kHz.
They are often used in watches, Real Time Clocks, and as a reference clock source.
They, like the other devices shown here, come in different shapes and sizes.
#4 & #5 Are both "Crystals" These are two different packages for 14 MHz crystals.
They are very accurate compare to the other ceramic resonator and the uC's internal RC Oscillator.
They both require external capacitors.
#6 External Oscillator This is a chip which has its own V+ and Ground pins, and outputs a square wave.
This one puts out a 25 MHz clock (square wave) at 3 V.
It uses a By-Pass, (decoupling), cap, 0.1uF, on its V+.
It doesn't otherwise use any capacitors.
The Wiki Page below has info on other clock sources, including those with their own temperature and voltage controller built in.
Wiki has a page on Crystals, seen here, with lots more info.
Note that on AVR Mega and Tiny microcontrollers one has to use the "Fuses" to tell the micro what is being used for the clock source to make the uC run.
On the Xmegas one can set the clock source via software.
Some uC's also have a Fuse to divide the clock, (whatever the source), by 8.
This slows the clock down a lot.
Some uC's have an internal 32 kHz clock as well as the internal clock that runs at several MHz.
awneil Great summary, Doc.
See also Tip #4 in my signature, below.
The internal (RC Oscillator) isn't very accurate
Indeed.
The main upshot of this is that it really isn't a good choice of you want to use the UART - because UART comms does rely upon an accurate & stable clock
Crystals use active circuitry inside the chip to make an oscillator
Might be worth clarifying that the "chip" in question here is the microcontroller itself.
Indeed, an excellent clock summary. Worth its weight in nooby questions!
In terms of performance, ceramic resonators range in accuracy from twice (or so) the limit for UART communication (which is around 2%) to half, or so, of the required UART accuracy. So, you CAN use a resonator for UART clocking if you watch out for the accuracy spec. To accurately clock a UART, a frequency source that is 2% or better is really needed.
All of the others tend to be good enough for UART clocking, over a wide range of temperatures. Clock (e.g. 32.768KHz) crystals are rarely used for direct UART clocking but, in most AVRs, can be used to calibrate a higher frequency but lower accuracy internal clock.
All rely on the active half of the oscillator (which is inside the MCU) to oscillate EXCEPT for the external oscillator (#6). External oscillators tend to use a lot of power, although there must be low power or micro-power devices out there. They also tend to be relatively expensive compared to a bare crystal. They keep generating a clock signal even if the micro is unpowered.
Jim
awneil the limit for UART communication (which is around 2%)
note that this is the final, overall error on the UART baud rate; the clock source contributes to this, but other things (like not-quite-exact division ratios) also contribute.
So you don't want to be starting out with a clock source having 2% error!
External oscillators tend to use a lot of power, although there must be low power or micro-power devices out there
Indeed there are; eg,
They also tend to be relatively expensive compared to a bare crystal
but they are also available in temperature-controlled versions (TCXO) - to give even higher accuracy.
EDIT
Perhaps from #20 on should be split out to a separate tutorial ... ?
I use a standard Arduino Nano module board with an 8-pin wide header pin strip soldered across the tops of the VIA holes for A2 to GND.
Then this module plugs into an Arduino-style 8-pin socket on the protoboard.
This gives Vcc,Gnd, Reset, I2C, two digital and two analog I/O in a plug-in space-saving module.
The list of "clocks devices" in Doc's #20 post is wonderful. However, it should be stated in term of Atmel's strange and EXTREMELY confusing naming of clocks devices in the fuse selections. They keep mixing in the word "oscillator" all over the place.Whoever came up with the names should be required to eat a bowl of burned chips for breakfast.
awneil I tried to flesh out the distinction between crystal and oscillator in #22, so that it would be all together, in one place, with Doc's excellent photo.
Jim
I know this is an old thread, but people keep linking to it.
I don't understand the significance of separate caps on pins 7, 20, and also on the reset pin. But "close to pins" implies that this is one of those things where actual-electronics differ from abstract/theoretical electronics.
Secondary question: Are there circumstances where those aren't necessary, or could conceivably be a problem? I don't understand exactly what they do, so I'm not sure whether there's ever a circumstance where you wouldn't want them.
I don't understand exactly what they do,
All your questions are answered in APP NOTE AVR042 AVR hardware considerations!
Section 1.1 says:
Read the complete description in the App Note.
Those caps are critical for proper operation, trust those of us that have spend years and lots of dollars to acquire the education and years of experience proving the theory works in thousands of commercial products.
You can drive on the wrong side of the road and get away with it for while, but some day you will meet a larger truck coming around a corner and you will wish you had listened to your driving instructor!
Follow the rules, and you will have fewer surprises and more working circuits!
Good luck with your project.
Jim
The caps on 7 & 20 are bypass caps, they ensure the mcu doesn't suffer from brownouts, these are mandatory.
The one on reset is optional, only required if your experiencing random resets due to power supply noise.
due to power supply noise.
Possibly caused by missing decoupling caps!
Thanks, that explanation helps a lot. I think what I still don't understand is what "too far away from the microcontroller" means in this context. Is that distance? How many other things are in the circuit, or connected to it between the cap and the MCU? Can the caps be too large?
"too far away from the microcontroller" ..... Is that distance? How many other things are in the circuit, or connected to it between the cap and the MCU? Can the caps be too large?
Yes. super ideal would be soldering the capacitors right on top of the chip leads, but as that is not possible you want them to be as close to the supply pins as you can.
There should be no other things in between the cap and the MCU pin......
Can the caps be too large?
again yes......
in the datasheets there should be preferred capacitance values. Like a 100nF on each pin. Sometimes also a larger cap is specified that has to be close to the MCU ( in addition to the 100nF capas) but not right on top of the supply pins.
