I'm heading to my hometown for a month, and while I want to leave the feeder up while I'm gone, I don't think it's ready for that just yet. There are some tweaks I need to make to the PCB and the sensors to make it more reliable and I don't feel good about leaving it unattended for so long since I've only had V3 up for a few weeks so far. 

Electronics

Building analog helped me better understand what's happening and breadboarding made things easy to adjust. While my next step will be to use a microcontroller (likely a PIC16F1709) I think this was a good step for me to take. It forced me to keep things simple and not get too distracted by "nice-to-haves".

This also helped me focus on solvng problems I have rather than try to solve too many future problems. The Morsings spent some time coding up a CV object differentiator to tell trash from natural debris and found they did not need it. To me this is somethings you would do if the birds were putting pebbles, twigs and other non-litter bits in the feeder, not before they have brought you anything at all. What they did is impressive, but I only want to make what I need and don't really have the time or resources to go down any rabbit holes.

Power

My current circuit and motor use about 850mA when the feeder is dispensing (max power consumption). It can be run by anything tha can supply 5-6 volts and 1 amp. Since the feeder is only activated when the PIR sensor is triggered, idle power consumption when no birds are present is probably around 20mA (not including the optional keep-alive pulse of ~150mA).

One improvement I will try to work into the existing PCB is for the powerbank keep-alive to only work while the feeder is in idle mode.

I originally designed this around 6V and think that since moving to 5V the drop sensor, which draws about 200mA with 4 IR LEDs, has become a little unreliable. I'm not sure what it is. I probably need to put slightly higher value resistors  on the IR LEDs (each has a 100Ω resistor connecting it to ground) since the actiual voltage going into them is closer to 4.5V due to transistor forward voltage. Some more testing and tweaking is required here.

Mechanical

Vibration dispensors worked best for me. Augers are ok if you are using a strong motor like a drill and don't mind wildly variable amounts dispensed, crushing the food as you dispense, and eventually turning things like peanuts into peanut butter. Hans Forsberg uses a 3D-printed vibration dispenser and the Morsings use a drill/auger setup. Since I am building small and low-power I used a very simple vibration dispenser I came up with (that certainly has been done before - but I have not yet seen examples) and can build out of jar lids or PVC end-caps.

Enclosure

The enclosure should be durable, light-tight, stand up to the weather and allow for access to the electronics. It's also nice if you can mount it on something to keep it off the ground and away from non-avians. My latest version is made of PVC pipe and is working great, but I miss the ability to access the insides easily like I could with the first two models. Not so important once you dial everything in, but being able to tweak things without disassembly is great while you are tweaking and experimenting.

Feed

I have been using shelled peanuts, but the problem with these is that almost all birds love them - especially pigeons. You could also use small cat or dog food, which corvids like but pigeons can't eat easily. I have thought about using peanuts in the shell for the same reason, but that will require a scale-up of the dispense mechanism and more testing - I'll get there.

Peanuts and cat/dog food will contain small crumbs. I had some issues with the feeder dispensing a tiny particle of peanut or peanut skin, apparently not dispensing anything to the birds. The dispense sensor would see something go by and turn off the dispenser. One of the reasons I tuned the sensors - now it's less likely to happen. I also put the nuts in a jar and shake the smaller particles to the bottom before filling the feeder so this is less of a problem.

Conditioning the Birds

My feeder has two on states - "normal" and "timer". The timer mode is just like normal mode, but will dispense a peanut every ~50 minutes automatically. This is good for getting the birds used to the machine and to see it as a food source. It's best to have some kind of platform where the birds can knock things into the deposit hole by accident - this will help them make the connection between litter and food being dispensed.

My conditioning regimen so far is as follows:

1. Put food on the platform, add some caps/butts/etc. and turn on timer mode - 2-4 weeks to get birds used to eating from the machine and to give them an opportunity to knock something into the deposit hole. It also helps to put some food under or in the caps so the birds will shove them around and knock them into the drop hole.
2. No food on the platform in timer mode, caps and butts on the platform - 1 month. By now the birds should see it as a regular food source and will start seeing a relationship betwen the non-food items on the platform and food being dispensed.
3. Normal mode with caps and butts on the platform - 1 month
4. Normal mode with caps and butts in a magazine - this is as far as I've gotten, expecting this to go on for a month or two
5. Normal mode with caps and butts on the balcony floor
6. Normal mode with caps and butts on the adjacent roof
7. Normal mode - no planted caps/butts