ESP32-P4 NOR flash LDO voltage

[thekahnman@gmail.com]

Hi all,

I'm working on a board using the ESP32-P4, and I'm considering the use of a 1.8V NOR flash chip.
It looks like the internal Flash LDO is switchable between 1.8V and 3.3V, settable via eFuse. OK, all makes sense.
Except - if I make a board with 1.8V flash, how do I safely power it on for the first time to burn the eFuse without blowing up the flash chip?
The docs say that GPIO45 is a strapping pin used to set the flash voltage, but I can't find any mention of that in the datasheet or the TRM.

And all of the existing dev kits break out GPIO45 to external headers. It seems like if it were really a strapping pin for the flash voltage, there would be some warning saying not to use it for general tasks.

Am I missing something here?

[JohnnyZ]

You could power it without flash attached, put it in download mode, and burn fuse

[dicethrow]

From what I've seen in esp-idf, LDO 2 defaults to 1.9v, and powers the internal PSRAM - so if the combined current draw, and the 1.9v voltage level are OK, I recon that would work. What flash IC are you thinking?

Relevant kconfig options appear to be ESP_LDO_VOLTAGE_PSRAM_1900_MV, ESP_LDO_CHAN_PSRAM_DOMAIN

[thekahnman@gmail.com]

You could power it without flash attached, put it in download mode, and burn fuse

Yes, but I want to make a PCB where the flash is attached before having been programmed.
My solution at this point is to include a cut/solder jumper to ensure that the first time the ESP turns on, it starts in joint download boot mode, then I can change the flash voltage over USJ.
I'd like to know if joint download mode wouldn't still send 3.3V to flash on first poweron...

From what I've seen in esp-idf, LDO 2 defaults to 1.9v, and powers the internal PSRAM - so if the combined current draw, and the 1.9v voltage level are OK, I recon that would work. What flash IC are you thinking?

Relevant kconfig options appear to be ESP_LDO_VOLTAGE_PSRAM_1900_MV, ESP_LDO_CHAN_PSRAM_DOMAIN

Unless the flash efuse is burned, the ESP will still attempt to use 3.3V logic for the QSPI lines, right?
Good suggestion though, appreciated.
I'm also curious - if I get the efuse set properly, is there still a chance the firmware application will attempt to set the flash voltage at runtime? I see there's user-mode control over LDO1, but I don't know if a generic firmware build would automatically try to assert the flash voltage from sdkconfig without being explicitly told to.
I'm thinking of using this flash... https://www.lcsc.com/datasheet/C6604692.pdf

[Sprite]

Hm, that's actually a very good question... previous iterations of the ESP32 turned on the flash LDO on bootup, even in download mode, iirc. I'll need to ask my colleagues if they know how you're supposed to use 1.8V flash on the P4.

[thekahnman@gmail.com]

thanks for looking!

BTW, I measured the flash voltage on an ESP32-P4-EYE in joint download mode, and i'm still seeing 3.3V on LDO1.

so I guess I will spin my board with a DNP resistor between LDO1 and the flash chip... hope it will not still try to put 3.3V logic on the QSPI pins?

[Sprite]

Yeah, seems you're right, there is no safe way to use 1.8V flash without modifications. Using a zero ohm link to isolate the flash may actually not be enough: the IO lines are still at 3.3V and some lines (e.g. /CS) may have that voltage in download mode as well.

[thekahnman@gmail.com]

OK, thanks.
Is this the sort of thing that's likely to get corrected in a silicon revision?
It seems like 1.8V flash is markedly more efficient, would be nice to be able to use it for battery-powered applications.

[thekahnman@gmail.com]

Seems like it might be possible to put the jumper between pins 71 and 30, if LDO1 is upstream from the entire flash power domain?

[Sprite]

Seems like it might be possible to put the jumper between pins 71 and 30, if LDO1 is upstream from the entire flash power domain?

Yeah, looks like you're correct. That saves a bunch of extra jumpers.