Barebones Arduino failing above certain voltage despite using voltage regulator

Hello

I have a barebones Arduino setup on an etched PCB. The device it ultimately belongs to runs on 15V. Therefore I am powering the Arduino board with 15V into a 7805 voltage regulator to give the ATMega328 a stable 5V supply. With Vin at 15V Vout is sure enough 5V. However the 'Arduino' doesn't behave properly even running just 'Blink' as a test sketch. The LED flashes dimly and rapidly.

However, if I drop Vin down to 5V (which I naively thought meant Vout would be 5V but it is in fact 3.4V) then all is well, the damn thing does indeed blink! I can take Vin to 6.5V (which does give a Vout of 5V) but beyond that it fails.

What's going on?

I have attached schematic/board layout. The sketch it will ultiemately run controls a pair of 595 shift registers and a MIDI in hence additional stuff. I'm testing blink by holding an LED against GND and pin 15 of the ATMega (Arduino pin 9 iirc)

Thanks
Jim

pic.png

schem.png

You shouldn't be building PCBs of circuits you have not validated (verified) on the bench. That is just foolish. For one thing your 7805 has no filter caps on the input or output. Is it a TO92 package ?

There are also no supply decoupling caps on the ATmega chip either. Decoupling caps are not an option.

raschemmel:
You shouldn't be building PCBs of circuits you have not validated (verified) on the bench. That is just foolish. For one thing your 7805 has no filter caps on the input or output. Is it a TO92 package ?

Whilst foolishness is well within my repertoire of expertise the circuit has indeed been verified on stripboard, where it worked as expected. No it is in fact a TO220 package. I can etch board in about an hour, I've got pretty good at it

However, the stripboard version was powered by a good quality bench supply, the etched one is not. Could this account for the correct functioning of the former?

Ah yes filter caps on the voltage regulator. My bad. I've looked over some older projects and I did indeed implement such things. I will refresh my knowledge and return should further problems arise unless you can spot any additional howlers for which I would be most grateful

No caps... been said...

The reason for failure is that you used the wrong pinout for the 78l05. Your output should be input ... and that explains the incorrect behavior. The results you got are what you would get when a regulator is in backwards

Top View... Flat side up... Output pin is left side. You have that as input. The TO-220 package is mirror of the TO-92 package and hat likely tripped you up

It appears from the silkscreen that there is a heat sink, but... is there a heat sink?

Don't connect AREF to 5V. Leave AREF open.

polymorph:
It appears from the silkscreen that there is a heat sink, but... is there a heat sink?

Nope, where is the silkscreen showing a heatsink?

In the pic of the board layout, there isn't a 7805 (which is a TO-220 package), there's a 78L05 in a TO-92 package...?

You do not have decoupling caps installed. You must have these - 0.1uf cap between power and ground, as close to the pins as possible, one for each Vcc or AVcc pin, otherwise you will have unpredictable failures.

You have AREF tied to Vcc; you're not supposed to do that on the ATmega328p - are you sure you're supposed to do it with the chip you're using? Check the datasheet (pin descriptions section).

You must have caps on both sides of the 7805.

You're making those boards yourself?! You must be going insane with all that drilling. I found it was less work to use SMD parts on than deal with the misery of drilling home-made boards. That's what made me start favoring SMDs...

I don't see anything that looks like a heatsink - I think that thing in the lower right is some kind of connector...

Ah! I took too quick a look, I was looking at the bottom right. A 78L05 is dissipating twice the power that the rest of the circuit requires when it is on 15V input.

And a TO-92 package isn't worth much when it comes to dissipation.

Definitely need decoupling on both Vcc and AVcc, 0.1uF ceramic for each, plus any capacitance needed
by the regulator. The ground-plane is broken across the middle of the chip where it is most needed -
that could be bridged with a short piece of wire to improve grounding.

I still don't see how you can say this circuit was tested on the bench. How long did you test it for , 5 minutes ?. Did you check the regulator to see if it was hot after an hour ? Nobody with any electronics experience would try to use a 7805 without caps on the input and output unless they just couldn't get their hands on any, but designing a pcb without the caps is just plain inexperience.

raschemmel:
I still don't see how you can say this circuit was tested on the bench. How long did you test it for , 5 minutes ?. Did you check the regulator to see if it was hot after an hour ? Nobody with any electronics experience would try to use a 7805 without caps on the input and output unless they just couldn't get their hands on any, but designing a pcb without the caps is just plain inexperience.

I am VERY inexperienced, but that's not a crime I hope! Thank you for your help and all the help above also

I couldn't tell if you are using a heatsink. Was that already explained somewhere ? When you use a 7805 with a 15V input it will have to dissipate 8V (the optimum input voltage is 7V so 15-7=8V.
If the load is 1 A, 8V @ 1A is 8W. You need a heatsink.

@Pollymorph,

Did I calculate that correctly ?

I'm not "Pollymorph", but 15V - 5V = 10V.

I'm not "Pollymorph", but 15V - 5V = 10V.

Duh.
I was trying to calculate the excess heat dissipated due to the fact that he is using a 15V input instead of a 7V input for his 5V regulator.

(15V - 5V) * current vs. (7V - 5v) * current

(15V - 5V) * current vs. (7V - 5v) * current

CORRECT

8ILOAD vs 2ILOAD

CONCLUSION: Using 15V input results in 4 times the power dissipation (8/2=4)

Please don't fight, we can't all be "Pollymorph".

Snerk.

ha, ha.