summaryrefslogtreecommitdiffstats
path: root/ps_notes
blob: 349ac116b319568b2a237093dbf302d2b8cb4f6e (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52

ST232CN w/ capacitors:

5.05v in (from USB via maple), 14.92 out unloaded

with circuit, 14.4v or so. remeasuring later, 13.3-13.7v (with a little can
capacitor)


---------------------------------------------------

Maxim MAX233CPP

Sat Nov 10 21:29:48 CET 2012:

with a 4.7kohm resistor, measure 16.99 volts. noise on output is a combination
of ~100mV p-p triangle waves at 168 kz, plus extra 120mV p-p bursts at peaks,
plus random sort-of-digital digressions of about 100mV on a time scale of 5ms
or so. large amplification of USB input noise? in this configuration, USB
voltage only wanders by 20mV or so, but maybe current varries.

without any load, output is 19.28v; noise is ~50mV p-p triangle waves with
100mV peak transients, plus same old macro scale wanderings.

swapping input cap to 0.1uF and adding a 0.1uF output cap gives 40mV total
noise with transients gone

running entropy.cpp and doing crude stats in python...

no negative resistor, INPUT_PULLDOWN, delay(1):  percent: 26.562500
no negative resistor, INPUT_PULLDOWN, delay(10): percent: 30.078858
no negative resistor, INPUT_FLOATING, delay(10): percent: 31.715096
no negative resistor, INPUT_PULLUP,   delay(10): percent: 32.595183

while probing un-amplified output, percent: 42.660407

switched to a design where the amplifier is fed from the 0/+5v USB power
instead of -10/+10 output. got:

with negative resistor, INPUT_PULLUP,   delay(10), percent: 50.102838
with negative resistor, INPUT_PULLUP,   delay(1),  percent: 52.064632
with negative resistor, INPUT_FLOATING, delay(1),  percent: 48.503847
with negative resistor, INPUT_PULLDOWN, delay(1),  percent: 45.382451


I think that this circuit can not be used for more than about one raw bit per
milisecond (meaning about 125 bits/second in the end assuming a 1/4 whitening
and 50% testing redundancy?).

for raw 32 kilobits/sec (eg, like entropykey?) need to sample with a period of
32 microseconds or so