AVR Freaks

It looks like this is a new AVR device that has only the UPDI (Universal Programming Demented Interface) method of communicating with the factory-new IC.  UPDI may or may not have advantages over the older ISP method of programming.  No doubt the engineers who developed it could dazzle me with demonstrations of its manifest superiority over the previous methods of device interface.  However, it has one primal disadvantage that has afflicted the embedded-systems industry since the 1970s: it requires a very expensive programming device made only by the CPU company whose only use is to get the object code into the memory of the microcontroller.  And this expensive programming device will only be purchased by companies that will be buying this IC design by the tens-of-thousands.  Not by you.

Basically there are four ways to get compiled/assembled object/hex code into an AVR device.  One is the new and vastly superior UPDI method, which few people use because it is new and expensive.  Expensive means any device that costs 50 times more than the cost of the IC itself.  This is actually the oldest approach: Intel sold "bond-out" programmers for $15,000 back in 1978 for 8051-based devices. If you wanted to be a player in the new embedded-systems industry, you bought one, along with a highly-trained and expensive engineer to program and operate it.

The next was an in-circuit-programming approach.  The reset pin was activated and often a higher-than-normal voltage was applied to the IC. Then bytes were sent in a serial-based code to the device one-by-one until the entire program was loaded. The reset was de-activated and the program ran on the device from its internal non-volatile memory.  The serial-based code was unique to the device family.  Non-corporate users could build parallel-port based bit-banging devices to implement this approach.

Next was JTAG: an industry standard means of reading/writing the memory of the device, and "freezing" it in the middle of a program's operation to examine each pin's logic at the moment of the "freeze". It also required advanced training, expensive equipment, and expensive engineers.

Finally, around 2005-2010, came boot-loaders. The device has a small program that monitored the UART peripheral on power-up/reset.  Signals arriving on the UART within a timeframe were assumed to be object code.  The bootloader would put these UART bytes into the non-volatile memory.  At next reset, the bootloader timed out and ran the program that had been previously loaded.  The Arduino system works with this method.  It requires no training or expensive equipment, but the bootloader program has to have been installed previous to the delivery of the IC device to the user.  This is clearly the best approach for low-cost non-corporate developers.  However, it doesn't have the ability to freeze and examine the CPU during an actual program run. Which can be essential for debugging complex programs and is demanded by engineers with big corporate budgets that are making tens-of-thousands units of single design.

Yeah... bad news.

I would be embarrased a little if I was atmel. Now Microchip.

You are right. Serial programming didn't require high voltages. Until now...

Atmel-ICE ist a very versatile programmer for ARM and all AVR Devices. I'm using the pcb version and think 45€ I have paid for it absolutely worth it. UPDI needs only one pin for debugging and programming, this works very well and economical.

But not forget: Programming this new Mega MCU does not require such a pulse. Never :-)

It's not so much a matter of "thinking of", it's more that on the smaller parts the customer starts complaining if he can't use the programming pin as an IO pin.

On the larger parts, the value of you not shooting your foot exceeds the added value for the customer with better aim.

Erm, the Atmel ICE is not that expensive and will work.... I paid GBP40 odd for mine, not sure of the current price, obviously a lot more will be flying off the shelf now.