Greetings!
The title says it all. Now that Charlie has a Pi-4, I want to make sure he has enough battery life to run longer than 10-15 minutes at a time.
One thought is a Lithium battery pack, like the ones used on cordless drills. Unfortunately, 12v drill batteries are now very difficult to find. I can get the 18v One+ Ryobi batteries - and I have chargers for them - but I fear that attaching 18v to Charlie will leave a smoking hole in the carpet.
Search the forum - I remember reading a Matt post that said over 14 is OK but the egpg.volt() will only return 14 for anything over that.
For the Pi4 Idle is 3.4 watts and my 8x Panasonic Eneloop Pro 2550mAh NiMH cells delivered about 23 watt-hours.
If you promise not to try machine learning while moving, you may be able to continue to use the eight-AA format, or perhaps try a 10 AA cell holder with the Eneloop Pro cells. (They cost a bunch initially, but because they deliver what they promise for twice the cycles of Chinese cells, the per hour cost ended up the same.)
Strongly suggest you create a load map e.g.:
Max startup battery current seen on meter: 430ma at 10.7v
Average idle battery current: 318mA (6/2019) at average 9.4v or 3 watts
Estimate Power Used in Test 22.3
Portion of Rated/Measured Battery Capacity Available 83%
Distance Sensor draws 30mA at 10.1v battery of Rated
Pan Servo draws 70-100mA at 10.1v battery
forward() “no-load” draws ~200mA extra (~430mA total) at 9.8v
forward() “loaded” draws ~330mA extra (~580mA total) at 9.5v
Connecting with remote desktop adds 10mA
Each “Eye” draws about 15mA at 9.5v
Speaking with espeak TTS draws about 8mA at 9.1v
Running glances draws ~40ma ave., 80ma peak, temp +3deg to 53C, (CPU at 5%)
Speaker (idle) draws ~1mA at 5v
GoPiGo3 PowerLED went “yellow” and quickly “Red” at around 7.4v?
Another data point: (but no motors and no multi-core video pattern recognition)
My philosophy for Carl is to manage load to find limits and stay out of trouble. Perhaps, if you want to try really pounding a Pi4 get another one and put it on your desk. The GoPiGo3 board will not supply enough 5v to drive the Pi4 to the max, but will serve up enough for exploring lots of topics.
You can also just add a usb-c powerbank to the Pi4 power connector and use the existing batteries to power the GoPiGo3, no?
I really want to post a picture of Charlie connected to a 12v car battery with huge-honkin’ jumper-cables as a goof. Not really connected, but to make it look like I’m powering Charlie off a gigantic DieHard!
That’s my question! What happens if I connect something to both power connectors?
More better goodness for longer. (Saw that in some forum post long, long ago …)
How do I go about creating the load-map. Perhaps yours is a sufficient guide that (at the very least) gives me relative magnitudes?
Put Charlie up on a pedestal, strap a multimeter between the battery and bot, and start plugging and unplugging, driving and not, sensing and not, speaking and not, etc.
But only when Carl gets off. Don’t you have him “up on a pedestal” already? 
Interesting observation:
I have Charlie sitting on my desk with a fresh set of batteries installed. With that, the battery voltage returns 9.6.
If I attach a True and Authorized Raspberry Pi-4 Power Supply to the USB type-C connector, the “measured” battery voltage jumps to 18.something!
Obviously, the battery voltage hasn’t jumped by that amount. I suspect I’m going to have to take a look at the GPG driver board schematic and propose a hardware change.
nice way of giving me a heart attack!!!
r u sure? usb-c allows it.
actually, I think Also remember matt saying something about that happening and to ignore it.
USB-C allows a lot of things, including video.
The Raspberry Pi supply itself spec’s at “5.1V (DC) 3.0A”
Even with the GoPiGo battery connector removed, it still reads 18+ volts.
Ouch!
Translation: The “battery voltage” is only trustworthy if nothing else supplying power is connected.
Update:
This thread:
. . . has a link to the voltage regulator data-sheet, and according to that, the absolute drop-dead max voltage is 45 volts for the garden-variety version.
Ergo, an 18v battery should work.
Ah, yes, but… but no biggie at 18v…
The DMP3017SFG power switch has a 30v breakdown voltage and who knows what the caps are rated at, but Matt didn’t seem to be too worried about being slightly over the 14v limit.
Here is one of Matt’s post on powering the GoPiGo3:
and a more specific “The absolute maximum voltage should never exceed 14.5v.”:
and this one also - he expressed concern about someone exceeding the 14.5v limit:
So it would appear the answer is:
I can accept that, but I wonder why?
I guess that GoPiGo schematic is coming out of the drawer again.
