Gregory's Server
 40 comments
solder paste before i installed the chip. this gives you a good idea of how finicky-small these things are to work with. what is it for? it's supposed to replace these chips on a very large circuit board that is expensive to respin. i want to make sure this circuit change works. oh yes! it works! this board is the MOnSter 6502 and i'm trying to figure out how to replace the really expensive transistor array devices (4x NMOS) with something cheaper.  @tubetime Are you planning on a second revision or is the Monster 6502 just expensive to keep operational? @bojanland it runs code. it's (more or less) the same CPU. It was in the Apple II, the C64 (the 6510 variant), the Atari 2600, the Atari 400/800/1200, and a bunch of other computers. whatever is driving the bus is very slow. the new analog switch closes quickly and loads the bus into accumulator bit 0, but the bus is still settling. here are 175 more adapter boards. gotta break them free first before soldering them down.  @RueNahcMohr hct4066 I'll have you know. also these are very cheap, 20 cents each another meaningless scope photo to some, but this contains clues as to why the MOnSter 6502 tops out at 100KHz or so. notice how, in the center of the screen, when channel 1 (the clock signal) falls, there is a corresponding drop in channel 3 (the data bus) this is an effect of the precharge mosfet. the clock signal goes to the gate, the drain goes to 5V, and the source goes to the data bus bit. the idea is that when the clock goes high, the bit gets charged up (to the clock pulse voltage minus the threshold voltage). when the clock goes low, the mosfet turns off, and the data bit remains charged. in practice, due to the gate to source capacitance (Cgs) the falling edge of the clock couples into the data bus bit. years ago I added bus capacitance which mitigates this somewhat. I've come up with a better solution. I've replaced the precharge mosfet with an analog switch chip -- one of those little circuit boards. and look: the glitch is gone, and we've got a clean signal now. the rising edge looks much cleaner as well. (we're looking at channel 2, purple)  @tubetime Was gonna say cuz that rising edge on yellow looks hella analog 😃 That said I like how the scope is labelling the voltages on the right edge.My Rigol doesn't do that, it seems like a reasonable upgrade. @tubetime oh god yeah just saw context now. building a whole CPU out of $6 matched MOSFETs is an expesive business. My first guess would have been SOT23 2n7000 and I guess that'd work for gates but body diode might scupper pass transistor stuff. 74hc4066 switches area good shout, txb0108 family might work as well.  @tubetime Diodes Inc makes 2x BSS138 in SOT363: https://www.diodes.com/assets/Datasheets/ds30203.pdf @hennichodernich I need transistors without the parasitic body diode from source to drain  @petrillic just a flat metal plate with a few of these boards taped down around it to get the height right. nothing really good or fancy. @tubetime similar so good to know I’m in the right ballpark. Have thought of trying to print custom 3d jigs but… @tubetime Also (not suggesting changes, just pondering out loud) I wonder if a board like this could survive PnP if it were fully routed on three sides and left hanging off of a single v-score on the remaining edge...  |
it started on a panelized array with mouse bites so each board can be broken out separately.