[Cryptech Tech] Dieharder test of Cryptech RNG data

[Павел Шатов]

On 08.04.2015 10:04, Benedikt Stockebrand wrote:

Hello, Benedikt.

> Hi Pavel and list,
>
> Павел Шатов <meisterpaul1 at yandex.ru> writes:
>
>> I also don't quite understand why charge pumps are used? As far as I
>> know, they have very limited output current, while during avalanche
>> breakdown large current flows through transistor. Is this okay?
>
> there are current limiting resistors in place, but other than that
> Fredrik opted for the charge pump because it seems to generate less
> ripple than a boost converter.
>
> Unlike hime I personally prefer the boost converter for a number of
> reasons, but this is one of the reasons why I like modular designs where
> these things can be individually modified or replaced.

We actually had a discussion with Fredrik yesterday and decided to 
replace charge pump with buck-boost (SEPIC) converter, namely LTM8045, 
http://cds.linear.com/docs/en/datasheet/8045fb.pdf

What do you think about it? Will it suit us? According to the datasheet, 
it has very low voltage ripple (4mV rms), which I think is very good. We 
also wanted at first to place an LDO after this converter to filter this 
ripple, but now I think it is not necessary actually.

>> What signal is fed into FPGA, btw? TP19 ("AMPLIFIED")? Is it okay to
>> feed analog signal into digital input buffer?
>
> I don't know about FPGAs, but when I built my stuff I eventually took a
> very close look at the MCU datasheets, concerning voltage thresholds and
> such.
>
> Going Fredrik's way and using Schmitt triggers is more modular than
> relying on the specs of the particular MCU or FPGA, but it adds some
> extra complexity.

Well, as far as I know from my experience with MCUs, they often have 
Schmitt trigger input buffers. What type of MCU input buffer are you 
currently using in your design? FPGAs only have LVTTL and LVCMOS inputs, 
so we deciced to add an external Schmitt trigger.

>
> Cheers,
>
>      Benedikt
>