DeeFuse comments

Results 13 comments of


                                            DeeFuse

sd_diskio_dma_rtos_template - SD_write with Scratch buffer not working as expected

1.27.0 produces the same file (the template hasn't been changed since 2020) [sd_diskio_dma_rtos_template_bspv1.c](https://github.com/STMicroelectronics/STM32CubeF4/blob/3d6be4bd406f275728e0a321cc371c62a3100533/Middlewares/Third_Party/FatFS/src/drivers/sd_diskio_dma_rtos_template_bspv1.c) This problem is related to #2

Replace timer based RNG seed with ADC reference based one

Unfortunately I got no AVR board to test the entropy. I can imagine that the result of the AD conversion will fluctuate in a range 0+X < n < 1023...

Replace timer based RNG seed with ADC reference based one

I had a look into the [hydro_hash_update](https://github.com/jedisct1/libhydrogen/blob/4a477649e679aff15480f88d8dc93913540bd70e/impl/hash.h#L2) function, it seems that the input is XORed with the state and then fed into the `gimli_core_u8` every gimli_RATE (16) bytes. I could...

Replace timer based RNG seed with ADC reference based one

I can't give a definite answer on the number of unbiased bits from the code alone. I'll look if i can find an AVR arduino in my stash to generate...

Replace timer based RNG seed with ADC reference based one

A basic test could be to simply create the average of all random uint8_t outputs of the generator function and observe if it converges towards 127. An other option would...

Replace timer based RNG seed with ADC reference based one

I'd say that the 0 readings are expected, ever 2nd itteration the ADC gets grounded to be able to recharge it. we could optimize the code to ground and recharge...

Replace timer based RNG seed with ADC reference based one

I'm doubt that 1 instruction is not enough to discharge the internal capacitor to have a sufficient long charge time and range for a good distribution > ADMUX = 0b01001111;...

Replace timer based RNG seed with ADC reference based one

I see that you removed the ADC [converisontime "randomisation"](https://github.com/jedisct1/libhydrogen/blob/878e59d1ca24d12ce07e695baa0fc831fbdb925f/impl/random/avr.h#L44). With that in place the randomnes of the readings could be increased.

Replace timer based RNG seed with ADC reference based one

I've tested the codes with a ATmega2561 in a comercial appliance I had at hand. Following porblems emerged: 1. The V_BG ADMUX settings are different (compared to the ATmega328) 2....

Init RNG seeding seemingly very predictable on Arduino Uno

The AVR random implementation relies on [TIMER1](https://github.com/jedisct1/libhydrogen/blob/4a477649e679aff15480f88d8dc93913540bd70e/impl/random/avr.h#L35) as a source of random data. This can lead to "predictable" or "similar" states after boot. A better source of randomnes would be...