On Wed, Mar 14, 2012 at 4:43 AM, Werner Almesberger <[email protected]>wrote:
> Next sheet, Ethernet. I'm puzzled about the transformer J4: > > First, the part indicated in the data sheet (Molex 48025-0002 [1]) > doesn't seem to be pin-compatible with the part indicated in the > BOM (ComWeal UDE RTF-114B8A1A [2]). > > Specifically, P1 is where R8 is, P2 is where R7 is, and so on, > while the internal connection of P1 corresponds to R1, that of > P2, to R2, and so on. This would mean that TD doesn't connect to > TD+/TD- but RD-/RD+, and RD+/RD- to TD-/TD+. > > If this is really the case, then what may save us here is the > auto-crossover feature of the chip. Not sure if the polarity > (and winding direction), which is also reversed, is important in > this case. > > The LEDs, on the other hand, are pin-compatible. > > Next, the internal connection of [1] and [2] looks different from > that of all the transformers Micrel recommend. I.e., Pulsec H1102 > [3], Bel Fuse S558-5999-U7 [4], Bel Fuse/Stewart SI-4600 [5], Bel > Stewart SI-50170 [6], Transpower HB726 [7], and LanKom LF-H41S > [8]. I couldn't find the YCL PT163020 (company dead) and the Delta > LF8505 (data sheet link leads to error page) they also recommend. > > What all of them have in common is: the parallel transformer is on > the cable side and the antiparallel transformer is on the host > side. [1] and [2], on the other hand, have only TD/TX that way, > but RD/RX is reversed. In addition, [1] and [2] have another coil > on the TX (cable) side. Are those two designs really equivalent ? > > Last but not least, Pulse give a circuit example on page 3 of [3]. > It differs in two respects from what we have: 1) the midpoints of > the coils are referenced to GND, not 3V3. 2) they don't go to the > rail directly but via a (poorly drawn) pair of 10 nF capacitors. > Wow ~ lots of researches ! I would speculate that they made specification sheets to differ from each well-known competitors. When I've sent [1] firstly datasheet to http://www.ude-corp.com/ , their sales and FAE have confirmed they are compatible. Yes, not only you felt puzzled but also I were told both are compatible then. One reason could be possible is no matter what circuits /or symbols drawn in block, a try to make them as look-alike as possible among competitors. Another, while producing rc1, we've tried functionally in success for 3 pcs of [2], see: http://lists.milkymist.org/pipermail/devel-milkymist.org/2010-June/000643.html in rc2, all used by [1], in rc3, all used by [2], so we might be wrong also be right after a cost down. > > [1] http://www.molex.com/pdm_docs/sd/480250002_sd.pdf > [2] > http://downloads.qi-hardware.com/hardware/milkymist_one/datasheet/Ethernet/RTF-114B8A1A(023-00).pdf > > The rest of the Ethernet sheet looks good. R31 has the small > issue that it doesn't have an explicit tolerance. (Like most of > the precision resistors of M1r4.) > applied temporarily to: http://downloads.qi-hardware.com/people/adam/m1/tmp/m1r4/Ethernet_20120316.pdf Thanks, - adam
_______________________________________________ http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org IRC: #milkymist@Freenode
