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