javaraspi
Posts: 15
Joined: Fri Jan 25, 2019 8:32 pm

Power Supply Sequence

Sat Apr 27, 2019 9:41 pm

Hello everyone.

The datasheet states power supply voltages should come up from higher to lowes in order to avoid latchup. I´ve seen some designs that simply put 2 or 3 LDOs directly tied to the 5V input.

Does anyone hahave implemented the powerchain this way & experienced latch up? .

Thanks for your comments.

hippy
Posts: 5118
Joined: Fri Sep 09, 2011 10:34 pm
Location: UK

Re: Power Supply Sequence

Sun Apr 28, 2019 1:52 pm

javaraspi wrote:
Sat Apr 27, 2019 9:41 pm
Does anyone hahave implemented the powerchain this way & experienced latch up?
I have experience of it happening in non-Pi circuits. I doubt there would be the recommendation to sequence supplies if there was not a risk of it being problematic.

javaraspi
Posts: 15
Joined: Fri Jan 25, 2019 8:32 pm

Re: Power Supply Sequence

Sun Apr 28, 2019 4:07 pm

Thanks Hippy for you comment.

The datasheet also says

"Alternatively supplies can be synchronised to come up at exactly the same time as
long as at no point a lower voltage supply rail voltage exceeds a higher voltage supply rail voltage"

I´ve seen a couple of baseboards (WaveShare Compute Module IO is one of them) that simply takes 5V and connects directly to the following LDOs:

1117-3.3V
1117-2.5V
1117-1.8V (This one is connected to the 1117-3.3V output)

Thanks again.
Javier.

aBUGSworstnightmare
Posts: 1275
Joined: Tue Jun 30, 2015 1:35 pm

Re: Power Supply Sequence

Sun Apr 28, 2019 4:48 pm

Always start with the highest voltage. Using a big 5V DC/DC and generate the other voltages out of it is the simplest way.
But I would recommend to use switching DC/DC instead of LDOs!

hippy
Posts: 5118
Joined: Fri Sep 09, 2011 10:34 pm
Location: UK

Re: Power Supply Sequence

Sun Apr 28, 2019 6:03 pm

javaraspi wrote:
Sun Apr 28, 2019 4:07 pm
I´ve seen a couple of baseboards (WaveShare Compute Module IO is one of them) that simply takes 5V and connects directly to the following LDOs:

1117-3.3V
1117-2.5V
1117-1.8V (This one is connected to the 1117-3.3V output)
What it boils down to is all higher voltages must always be greater or equal to any lower voltages or latch-up may occur.

If the unasked question is, can you connect all your LDO's to 5V, or have a cascade of regulators as someone else has; the answer would be, yes, if it it meets the requirements, no, if it doesn't.

User avatar
Burngate
Posts: 5868
Joined: Thu Sep 29, 2011 4:34 pm
Location: Berkshire UK Tralfamadore
Contact: Website

Re: Power Supply Sequence

Mon Apr 29, 2019 10:08 am

javaraspi wrote:
Sun Apr 28, 2019 4:07 pm
I´ve seen a couple of baseboards (WaveShare Compute Module IO is one of them) that simply takes 5V and connects directly to the following LDOs:

1117-3.3V
1117-2.5V
1117-1.8V (This one is connected to the 1117-3.3V output)
The earliest Pis had almost exactly that - the 2.5v regulator was also connected to the 3v3 output.

Later Pis - e.g. B+, 2B, 3B - use a single switching regulator chip for the 3v3 & 1v8, with the second output delayed by an RC circuit on its enable input.

hippy
Posts: 5118
Joined: Fri Sep 09, 2011 10:34 pm
Location: UK

Re: Power Supply Sequence

Mon Apr 29, 2019 11:16 am

Burngate wrote:
Mon Apr 29, 2019 10:08 am
javaraspi wrote:
Sun Apr 28, 2019 4:07 pm
I´ve seen a couple of baseboards (WaveShare Compute Module IO is one of them) that simply takes 5V and connects directly to the following LDOs:

1117-3.3V
1117-2.5V
1117-1.8V (This one is connected to the 1117-3.3V output)
The earliest Pis had almost exactly that - the 2.5v regulator was also connected to the 3v3 output.
The Pi B 1.0 appears to have -

Code: Select all

5V0 >---.-------------------------> 5V0
        `---> LDO ---.------------> 3V3
                     }---> LDO ---> 2V5 *
                     `---> LDO ---> 1V8 *
I am assuming the validity of that rests on the output voltage of a regulator never exceeding its input voltage. So 3V3 comes up later than 5V, 2V5 comes up later than 3V3 and 5V, as does the 1V8.

The issue is, could 1V8 come up before 2V5, which would invalidate the requirements ? That could potentially happen if 2V5 were more heavily loaded than 1V8, and would of course depend on what the characteristics of each regulator were.

I do note the 2V5 LDO is a different product type to the 3V3 and 1V8 LDO's. That could be because they chose a 2V5 regulator which had a fast come-up time to avoid the 1V8 coming up before the 2V5.

On the Waveshare board the issue would seem to be, could 2V5 come up before the 3V3 ... ?

Code: Select all

5V0 >---.-------------------------> 5V0
        }---> LDO ---.------------> 3V3 *
        `---> LDO ---|------------> 2V5 *
                     `---> LDO ---> 1V8
Given all LDO's there are the same product type, 1V8 coming up before 2V5 would seem unlikely, but potentially possible.

One would have to give WaveShare the benefit of the doubt and assume they verified things were as intended and checked the outcome actually met requirements.

aBUGSworstnightmare
Posts: 1275
Joined: Tue Jun 30, 2015 1:35 pm

Re: Power Supply Sequence

Mon Apr 29, 2019 1:17 pm

Some regulators have an Enable input.

javaraspi
Posts: 15
Joined: Fri Jan 25, 2019 8:32 pm

Re: Power Supply Sequence

Wed May 08, 2019 12:39 pm

Excellent!
Thanks all for your inputs!

Return to “Compute Module”