charlton91
Posts: 12
Joined: Wed Mar 30, 2016 10:22 am

Compute module boot time

Wed Mar 30, 2016 3:02 pm

Hi,

We have manufactured our own base board for the Compute module and are seeing long boot time prior to the splash screen being shown.

Our base board is connecting the Official raspberry pi 7" display to the compute along with a LAN9514 and an audio chip.

If I plug one of the programmed compute modules into the Compute IO board i get the the splash screen with the colours within 2 - 3 seconds however if I use our own base board this can take approximately 15 seconds.
Once the splash screen appears then the boot time is roughly the same using the IO board as using our PCB.

From comparing the schematics for the pcb's I cannot see anything glaringly obvious.

Any help with this would be greatly appreciated.

stuartclark
Posts: 42
Joined: Wed Nov 07, 2012 12:41 pm

Re: Compute module boot time

Wed Mar 30, 2016 3:42 pm

If you put a scope on the power lines is it raising quickly and to the right levels?

charlton91
Posts: 12
Joined: Wed Mar 30, 2016 10:22 am

Re: Compute module boot time

Wed Mar 30, 2016 4:02 pm

All of the voltage regulators are getting to the correct voltage within 1 ms.

MrBool
Posts: 103
Joined: Sat Jul 05, 2014 9:51 am

Re: Compute module boot time

Wed Mar 30, 2016 5:50 pm

I had similar problem
viewtopic.php?f=98&t=127503&p=852959#p852959
I could not find solution.

MichaR
Posts: 16
Joined: Thu Mar 10, 2016 5:49 am
Location: Berlin / Germany

Re: Compute module boot time

Thu Mar 31, 2016 4:22 am

Hi,

did you try to connect the RUN pin directly to 3.3V ? For me that fixed the boot delay. It's strange, but I had to connect it without any pull-up resistor. Even more strange is, that the RUN pin has 3.3V if you left it unconnected, but the CM has a big boot delay.

Micha

MrBool
Posts: 103
Joined: Sat Jul 05, 2014 9:51 am

Re: Compute module boot time

Thu Mar 31, 2016 8:02 am

MichaR wrote:Hi,

did you try to connect the RUN pin directly to 3.3V ? For me that fixed the boot delay. It's strange, but I had to connect it without any pull-up resistor. Even more strange is, that the RUN pin has 3.3V if you left it unconnected, but the CM has a big boot delay.
This is really strange, because RUN pin is pulled up with 10k resistor on CM.

Very important is also power sequencing. From
https://www.raspberrypi.org/documentati ... gnguide.md

"Supplies must be synchronised to come up at exactly the same time. Alternatively, they should be staggered so that the highest voltage comes up first, then the remaining voltages in descending order. This is to avoid forward biasing internal (on-chip) diodes between supplies, and causing latch-up."

charlton91
Posts: 12
Joined: Wed Mar 30, 2016 10:22 am

Re: Compute module boot time

Thu Mar 31, 2016 8:38 am

My power supplies will come up as specified, "highest first then in descending order".
They are daisy chained together with 3 voltage regulators, 12V to 5V, 5V to 3.3V, 3.3V to 1.8V.

I have also checked that we have all of the correct voltages as per the compute module hardware design guide.

PatronP
Posts: 8
Joined: Thu Mar 31, 2016 7:04 am
Location: Sweden

Re: Compute module boot time

Thu Mar 31, 2016 9:03 am

I’m having the exact same problem on our custom carrier board and trying to find a solution for this. I have tried sequencing the different power and having the RUN pin connected to +3.3V without any luck.

rln
Posts: 174
Joined: Wed Apr 09, 2014 1:43 pm
Location: Sweden
Contact: Website

Re: Compute module boot time

Thu Mar 31, 2016 9:29 am

Check the GPIO47 and ensure it's not in "alternate" boot mode, as stated in the I/O board schematic.
Author of the robust Nard distro http://www.nard.se

PatronP
Posts: 8
Joined: Thu Mar 31, 2016 7:04 am
Location: Sweden

Re: Compute module boot time

Thu Mar 31, 2016 10:27 am

It makes no difference if the GPIO47 is high or low. The module will boot the same way and with a ~15s delay.

rln
Posts: 174
Joined: Wed Apr 09, 2014 1:43 pm
Location: Sweden
Contact: Website

Re: Compute module boot time

Thu Mar 31, 2016 11:38 am

What about checking with an oscilloscope or logic analyzer, on the module signals EMMC_CLK and EMMC_CMD?
Author of the robust Nard distro http://www.nard.se

PatronP
Posts: 8
Joined: Thu Mar 31, 2016 7:04 am
Location: Sweden

Re: Compute module boot time

Thu Mar 31, 2016 12:38 pm

I’m getting movement on both CLK and CMD right away after power on but it halts after some time (~1s) and then ~10s later more action for ~250ms and then nothing for 2s and after that the module starts. At least this is what I can see and make guesses from the oscilloscope picture.

rln
Posts: 174
Joined: Wed Apr 09, 2014 1:43 pm
Location: Sweden
Contact: Website

Re: Compute module boot time

Thu Mar 31, 2016 1:03 pm

Then you at least know that the CPU powers up as it should. Likely it's a SW issue... You may perhaps want to try the "next" experimental firmware bootloader to see if that helps.
Author of the robust Nard distro http://www.nard.se

PatronP
Posts: 8
Joined: Thu Mar 31, 2016 7:04 am
Location: Sweden

Re: Compute module boot time

Thu Mar 31, 2016 2:10 pm

The thing is that the module boots just fine when inserted in the RPi devboard. It looks like one or more signals aren’t as they should be at boot.
I’ll get the same boot delay when toggling the RUN signal.

Tried the "next" versions and that didn't help either.

6by9
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 7135
Joined: Wed Dec 04, 2013 11:27 am
Location: ZZ9 Plural Z Alpha, aka just outside Cambridge.

Re: Compute module boot time

Thu Mar 31, 2016 3:05 pm

Really long shot: Are GPIOs 0&1 pulled down on your board? Or at least not pulled up.
The Compute Module shouldn't probe HATs, but if it did and the I2C lines are pulled down then it'll clock stretch and take longer to time out. Possibly also applicable to the I2C going to the display, or HDMI.
There's not a great deal that goes on before the splash screen, but HAT and display probing would be two things.
Software Engineer at Raspberry Pi Trading. Views expressed are still personal views.
I'm not interested in doing contracts for bespoke functionality - please don't ask.

PatronP
Posts: 8
Joined: Thu Mar 31, 2016 7:04 am
Location: Sweden

Re: Compute module boot time

Thu Mar 31, 2016 3:40 pm

We are using GPIO 0 and 1 as I2C and have devices on 0x1C, 0x50 and 0x68. And as on all I2C buses there are pull up resistors.
I do see some action on the I2C bus before console printout, but I have not yet decoded fully what is going on. It looks like address 0x45.
Anyway, there is no one answering and all that probing takes less than 2ms.

Is there any way to verify that this could be the problem, can I disable the I2C probing?

6by9
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 7135
Joined: Wed Dec 04, 2013 11:27 am
Location: ZZ9 Plural Z Alpha, aka just outside Cambridge.

Re: Compute module boot time

Thu Mar 31, 2016 3:52 pm

PatronP wrote:Is there any way to verify that this could be the problem, can I disable the I2C probing?
Sorry, I'd read Charlton91's original post, not noticing you'd taken his thread. He'd said:
We have manufactured our own base board for the Compute module and are seeing long boot time prior to the splash screen being shown.

Our base board is connecting the Official raspberry pi 7" display to the compute along with a LAN9514 and an audio chip.
Hence my reference to the screen.

In config.txt

Code: Select all

force_eeprom_read=0
ignore_lcd=1
ignore_hdmi_edid=1
should disable HAT probe, display probe, and HDMI EDID respectively. As I said, both the first two should be disabled by default, but it was just a daft thought that struck me.

edit Add EDID bypass too.
edit2 I'll confirm that 0x45 is the LCD display probe address. That would imply that you have DISPLAY_SDA and DISPLAY_SCL defined in your dt-blob.bin which the default blob doesn't. I wonder if you've something else quirky in there. Could you post the pins_cm section?
Software Engineer at Raspberry Pi Trading. Views expressed are still personal views.
I'm not interested in doing contracts for bespoke functionality - please don't ask.

mbuchmann
Posts: 1
Joined: Thu Mar 31, 2016 3:42 pm

Re: Compute module boot time

Thu Mar 31, 2016 6:02 pm

Might be same problem as this: viewtopic.php?f=98&t=115095

PatronP
Posts: 8
Joined: Thu Mar 31, 2016 7:04 am
Location: Sweden

Re: Compute module boot time

Fri Apr 01, 2016 7:08 am

mbuchmann wrote:Might be same problem as this: viewtopic.php?f=98&t=115095
Yes that did it, Thank You!
Adding hdmi_edid_file=1 to the config.txt fixed the boot delay for me and we are getting an instant boot after RUN toggle or power on.

As info, we are using our own display via HDMI to parallel bus but without a EDID memory.

MrBool
Posts: 103
Joined: Sat Jul 05, 2014 9:51 am

Re: Compute module boot time

Sat Apr 02, 2016 7:58 am

mbuchmann wrote:Might be same problem as this: viewtopic.php?f=98&t=115095
Thanks a lot mbuchmann! Adding hdmi_edid_file=1 to the config.txt fixed the boot delay for me also.

BTW. Is there a documentation available for configuration with all available options?

charlton91
Posts: 12
Joined: Wed Mar 30, 2016 10:22 am

Re: Compute module boot time

Mon Apr 04, 2016 10:54 am

Thanks for all of the input from everyone on this issue.

I found that adding hdmi_edid_file=1 to my config.txt did resolve the issue.

However I have now changed my hardware design as per the previous link to add a pull up on the HDMI SDA line.

Once i added this wire i no longer needed to have the entry in config.txt

Again thanks for all of the help.

Return to “Compute Module”