Hacker News Reader

Comment by failwhaleshark | Hacker News Reader

As an alternative for high-throughput TRNG entropy sourcing: Why not just sample thermal noise like the Simtec Electronics Entropy Key# did? Doesn't it just need a fast-enough ADC? (How expensive is a ~ 10 Gb/s ADC?)

# I'm kinda annoyed they didn't open-source the design before shutting down. :(

View on HN

> How expensive is a ~ 10 Gb/s ADC?

$1000. Much cheaper to use many slower ADCs in parallel. Or put your entropy directly into a Schmidt trigger, etc. since you don't care about any of the specs that make ADCs so complicated.

Where do you get that price from? We have been trying to source high speed ADCs for some experiments and it's virtually impossible to get anything above 5 GS/s and prices are more than an order of magnitude higher than what you state.

RFSoC is an fpga with high speed ADCs and DACs integrated, 8 5Gsps ADC, and 16 10 GSPS DACs in a single device, and less than 1000 bucks IIRC, though not sure on volume (i.e. If you can buy single parts)

https://www.analog.com/en/parametricsearch/10826#/sort=1746,...

That is the price for just the IC, at 1000 quantity. Depending on the experiment that might not be a relevant number.

You don't need an ADC, you need a fast threshold circuit with regular sampling that will turn the noise into a bitstream, and a shift + latch register to accumulate large enough words from these bits.

The biggest problem of such a design might be correlation with the variations of the power supply voltage.

Ah-ha, hack enginerding FTW. Overenginerding costs $$$.

I wonder about metastable sequential logic (D flip-flop sampling 3 NOT gates in a loop) XORed thermal noise sampling from a resistor XORed with the D flip-flop's clock signal. Violating setup-and-hold times of digital logic would also probably help generate some fun randomness.

Reply on news.ycombinator.com