| 19 Sep 2023 |
|  nael joined the room. | 23:42:39 |
| 21 Sep 2023 |
|  lunardigs 🌒 changed their display name from lunardigs (forwhatisthecometothisonehere?) to lunardigs (Jordachedude). | 23:35:14 |
| 23 Sep 2023 |
|  @harishkumaran:tchncs.de joined the room. | 22:18:43 |
| @harishkumaran:tchncs.de left the room. | 22:28:51 |
| 26 Sep 2023 |
| lunardigs 🌒 changed their display name from lunardigs (Jordachedude) to lunardigs (docnarcolongo). | 01:35:23 |
| 2 Oct 2023 |
|  @krishnamaithreya:matrix.org joined the room. | 08:07:02 |
| @krishnamaithreya:matrix.org left the room. | 08:08:01 |
|  @temp4096:matrix.org joined the room. | 11:31:31 |
| 3 Oct 2023 |
| @temp4096:matrix.org left the room. | 10:11:30 |
| 8 Oct 2023 |
|  lndn joined the room. | 05:30:08 |
| 10 Oct 2023 |
|  @partime:matrix.org joined the room. | 20:55:02 |
| @partime:matrix.org left the room. | 21:04:09 |
| @partime:matrix.org joined the room. | 21:04:35 |
| @partime:matrix.org left the room. | 21:06:10 |
| 23 Oct 2023 |
|  surfingturtle joined the room. | 09:28:51 |
| 2 Nov 2023 |
|  @falk:zeh.eu.com joined the room. | 13:33:25 |
| 3 Nov 2023 |
|  @blaster1920:matrix.org left the room. | 02:41:45 |
|  Zdeněk joined the room. | 07:29:09 |
| 5 Nov 2023 |
|  DO5RSW - Sebastian changed their display name from do5rsw to Sebastian - do5rsw. | 10:10:27 |
| DO5RSW - Sebastian changed their display name from Sebastian - do5rsw to DO5RSW - Sebastian. | 10:11:36 |
| 6 Nov 2023 |
|  PurpleLady joined the room. | 08:12:00 |
| PurpleLady | I'm more into homebrew computing and digital electronics at the moment, and more of the older stuff of that. | 08:13:58 |
| PurpleLady | I've been trying to work out a TRNG (true random number generator) using digital components. I know you can drive a transistor "backward" to the breakdown voltage and then amplify the noise this gives. From there, you can use things such as Schmitt triggers, ADCs, voltage levelers, etc. Of course, you'd need to have 9-12 volts to reverse bias the transistor so that shot noise can occur. If you were building a music synthesizer, that would be fine (assuming you want to be Eurorack compliant). But if you want to build something that a phone charger can power, that would be out. | 08:18:05 |
| PurpleLady | There are other things you can do if you want a TRNG. One would be to use ring oscillators. So get hex or octal inverter IC and wire an odd number of inverter channels through each other and back around in a "ring" formation. And since the IC in question would have an even number of channels, you can use the last channel as a buffer. It would remain outside of the ring. You can only use an odd number in the ring. Otherwise, you'd be creating a latch, not an oscillator. So if you use the remaining channel outside the "ring," yes, it would invert the output, but if that is the only load outside of the oscillation placed upon the circuit, it should be able to still sustain oscillation.
So then, you can XOR the outputs of 2+ ring oscillators (of different lengths, such as 5 and 7 channels). You'd have both the pseudorandom effect of beating 2 frequencies together and the drift/aliasment of that, but also the drift and skew of using unregulated oscillators. | 08:28:19 |
| PurpleLady | If you were making a Pierce oscillator for a system clock (instead of using an oscillator "can" of the frequency you need), you'd only use a single channel (only good at maybe up to 30-35 MHz if using DIP packages), but you'd use loading capacitors, resistors, and a crystal. This tank/loading circuitry is to slow the oscillator. | 08:32:22 |
| PurpleLady | Then once you get your random bits, you'd need a way to make them useful. So use a shift register to build bytes (or whatever size you need). After that, you could use a whitener of sorts. You could use the VN algorithm (10 = 1, 01 = 0, and skip 00 and 11, and make the comparisons non-overlapping). That is somewhat inefficient as it wastes some of your entropy, but it is a way to make a more balanced set.
But you don't need to do that. You could get another shift register (at least 1 more bit than you need) and XNOR the top 2 bits, feeding that back into the "serial-in" input. So that is a linear feedback shift register algorithm.
Then you can XOR/XNOR/ADD the parallel lines of the true random and the pseudo-random shift registers. | 08:39:19 |
| PurpleLady | Then you'd have a random unsigned or signed integer based on however you interpret the result. | 08:41:14 |
| @falk:zeh.eu.com left the room. | 08:42:32 |
| PurpleLady | Of course, instead of and LFSR, one could use a ROM and a counter. Put an even, scrambled table in the ROM with balanced periods. That would not be truly random, but it would be no worse than LFSR. That is just combinational logic generating the same table each time. So the counter chips drive the address lines and point to a different "random" number. Now, what if you want to use a seed? Just add a number to it for the entire period. thus the spacings are the same, but using different number.
Now, to make that seem more random, have a way to back up the state of your counters between powerings. So if you use NVRAM, the last address counter value and the addition offset can be backed up and restored for the next boot. So the user's behavior can add to the randomness. | 08:49:34 |
| PurpleLady | I'm sorry if I ran someone off. That was not my intent. | 08:49:51 |
