kevin gilmore

3 oz copper might be better for the power supply boards, harder to destroy lands that way. although with that much capacitance, copper is gonna fly no matter what if you do a hard short

for the V2 attenuator, there was one missing wire, will fix soon. fixed http://gilmore.chem.northwestern.edu/boards/attenuatorsmtv2.zip http://gilmore.chem.northwestern.edu/attenuatorsmtv2.pdf

only for the bipolar version do you need 10k, otherwise anything works. schematic, nope never did one, 2 x pcf8574 driving 2 x uln2803 latching relays.

all the boards I have ever had made were pcbnet, no issues yet.

oops, forgot to mention i will pay for my boards.

for multiamp, you really want the 10k version.

finally cleaning up, i have one extra set of the original attenuators, part of the buy that kerry did, professionaly assembled, and i tested them last night, so if anyone is interested i can sell that pair. Price is whatever kerry charged, which might take me a while to look up.

ray is still trying to sell his business. I think he is down to about $1M by now. I know for sure 3 people/companies he has offered it to, and all were smart enough to turn it down. pretty sure his 15 minutes of mil-spec fame are over. Very over.

if you are comcast you definitely want the 6121 or 6141, as of last week in chicago they upgraded the firmware, so everything is fully ipv6 compatible. Also getting 90mbit now. and the big netgear for the wireless or something similar so that ipv6 works correctly too.

10 boards plus 3 transformers plus 5U case per side 1 x hv ground reference 2 x hv900 1 x driver 1 x output

174k will be fine

cascode current sources at 1 amp, not so easy parts are not available

everyone wanted tiny. they got tiny. i could do singles sic diodes or schottky dioes, board has to get a bit bigger. and the noise is already 140+ db down. i can't measure that at home, my led lights over the bench generate 7mv into a single turn of wire. and the switcher in the burglar alarm, and the high frequency switcher in the whole house UPS... and the switcher in the variable frequency drive for the AC... and at work, just barely can i measure. the stuff that does -160 to -170db seems few and far between. And in the stupid price category.

the noise spikes are 120hz (100hz for you) and due completely to the switching noise of the input diodes. every power supply, even tube rectified ones do this. The idea is to remove the spike noise, and its not easy. caps across the diodes, etc, just does not work. and by the way, so far, I can't even get close to these numbers with a shunt regulated supply, the noise rips right thru the main current source. 7815 is rated at 90 microvolts of noise (ti datasheet). which is great except that in the typical configuration the diode noise rips right thru, and is usually in the millivolt range. and the voltage stability is +/-4% over commercial temperature range. plus line and load regulation. S22, same thing, the capacitance multiplier at the input is useless. Some companies in the stupidly overpriced 3 terminal replacement business call this a gyrator. Nope, not a gyrator. yes it can regulate at 3V, but you don't want to do that. What you might want to do is use that voltage to calculate line voltage -15% for the absolute minimum voltage. Then make sure all the components are rated for line voltage +15% its all about heat in the pass transistors. if you want to go down to 8 volts, you have to change the series zener to 4.7v that zener is there to make sure the thing starts up under load. and change the reference to 5v here is the original design that was modified by ucc, then modified by me. http://waltjung.org/PDFs/Improved_PN_Regs.pdf

http://gilmore.chem.northwestern.edu/kgsshvcarbonproductionv5.pdf so for original servo, adjust the amplifier for an output voltage around zero then put the jumper in for servo v1. for the new servo, adjust the amplifier for an output voltage around +20v then put the jumper in for servo v2 never put both jumpers in at the same time.

can't really zero the input dc. some of the that340 are better than others. if you can get the lsk170/lsj79 parts, those sub directly. and have much lower offset currents. on the 8 boards I have done, all adjust to less than 1mv.

exactly, and a few parts removed http://gilmore.chem.northwestern.edu/kgsshvcarbonv5.jpg this guarantees that its gonna work, but its gonna work anyway! reduces the GB price!

Kerry is a bit nervous that the v3 servo for the carbon may not work, so I did a v4 board which has both servo's on it, use one or the other. Much safer for the group buy I think. http://gilmore.chem.northwestern.edu/kgsshvcarbonv4.jpg

those are bipolar inputs, with a bit of offset current, so you really need a 10k input pot, or 10k resistors to ground. other than that I listened to my latest set of boards on the new power supply, and there is definitely a noticeable difference in noise with iem's.

yes, buy a bunch. but it is more money than an open frame ps.

I guess that the next thing I have to master is micro engraving. I do have the tools...

if you do the 2 complementary singles, then yes, otherwise no. also the bom lists the on board pass transistors, not the off board ones I did more testing, and at 500ma, the minimum dropout voltage seems to be closer to 3v. But at 5v, the pass transistors do just fine with a small heatsink as is. 5V, 500ma ... 2.5 watts

not an error. from the krell power supply, 10 ohm resistors from iec input ground to power suppy ground. not likely that anyone including me is going to use it though.

ground down the middle. there is actually a benefit of dual supplies. the noise spikes what little of them exist cancel.

2 watt transformer... no 22 watt, just barely 30 watt transformer is about right one of the mounting holes is 25th off of square. current posted boards (dual and singles) have the mounting holes one size smaller, the silk screen fixed, and 1 resistor each moved by 25th. done measurements this morning in a screen room is about 800picovolts of noise. and voltage stability exceeding 6.5 digit meter. anything better is going to require stupidly expensive test gear.