The mention of cascading made me recall something I was intending to post but forgot. I stated in a recent post here that I assumed the ASWM used a 6x9 multiswitch in the front end. The "prior art" in the DSWM patent shows RF5201 chips being used, and I couldn't quite figure out how that worked since Entropic's site showed they had only two inputs, but they'd need three to work with a multiswitch at the front end. The problem of course was that the prior art showed four inputs, and that wasn't even including the flex ports it left out! It just didn't make sense. It turns out the "routing of LNA signals to all 3 chips" shown in the prior art (figure 2) don't count as inputs - the chips use side channels to share inputs with each other. So 3 chips x 2 inputs each = 6 inputs, now it makes sense!
Following this design, the SWM16's rather large legacy loss makes perfect sense, since each LNB input must be split six ways to go to the six RF5201 chips (three on each "SWM8 half") The loss from a two way split between "halves" followed by a balanced three way split inside the RF5201s exactly matches the "slightly over 10db" legacy loss reported for the SWM16. While this isn't much of an issue for the SWM output since it is gets the benefit of the AGC boost, based on this I'd expect the SWM8 may have a few db greater margin for loss at the input since it is only split three ways. If someone had a splitter and a SWM8 and SWM16, and a method for attenuating the input in stages (maybe by just adding splitters to one polarity while trying to tune a channel on that polarity via the SWM output on each of them using identical receivers) I'd expect the SWM16 to lose the signal "one splitter's worth" earlier. The SWM32 is reported to have an output that is about level with its input, so it must have a ~16 db AGC amp at the front end - to make up for the twelve way loss it suffers - and thus should be much more tolerant of weak input than the SWM8 or SWM16.
It also makes sense why Entropic's site shows the apparently unnecessary RF5200 - which is a 3 input / 3 output version of the 2 input / 3 output RF5201. That has to be what the SWM5 used, since if it used RF5201s it would lack sufficient inputs with which to support flex ports. I googled it and though I could only find a couple rather crappy stock photos, it looks like the SWM5 had only two legacy ports, which fits my hypothesis perfectly. The one thing I'm not clear on is why the SWM8 doesn't support the HD satellites on its legacy outputs, since there is no reason why it shouldn't be able to access any satellite via the RF5201. It would appear Directv must have put bandpass filters on these outputs that only allow the 950 - 1450 MHz range to pass, for some reason.
I figured all this out when I was looking at the RF5201 on Entropic's site. When I looked for similar products, I saw the RF5218 (used in European CSS systems) After a bit of googling I found some very detailed information on the very similar RF5210, which is sold under the RF Magic brand, which Entropic owns. It was easy to see how the RF5210 works, and the prior art in the DSWM patent with the very similar RF5201 suddenly made perfect sense. It turns out there is actually a 6x4 multiswitch in each of the three RF5201 chips in the ASWM (3 outputs for SWM, 1 for legacy) rather than external 6x9 multiswitch at the front end I was assuming (which would have been wrong anyway, I forgot about the legacy ports - it would have required a 6x12 multswitch)
Here are the links to a couple sources with functional diagrams of the RF5210:
This brings up another point. The DSWM ASIC includes no legacy multiswitch, so there's an implementation choice for a DSWM module. Either integrate a legacy multiswitch in the module to allow it to support the 4 legacy + 2 flex ports as the SWM16 and SWM32 do, or use splitters at the front end to provide cascade outputs. Those "cascade ports" would not act as legacy ports, because they'd be polarity locked (i.e. the 99/101 even port would always output only that polarity and an attached receiver could not change it) You'd lose ~5db on each cascade from that splitter loss, so you couldn't cascade very far unless it included an amplifier.
The pictures of the DSWM13 shows the traditional six outputs, but Stuart stated "the DSWM13 doesn't have any legacy ports, despite what you see" so perhaps they're cascade only ports. There's also the possibility they do absolutely nothing in the DSWM13. That might be the case if, for example, they built it using the chassis they plan to share with a future DSWM23, which will have some sort of legacy/cascade output. Maybe I'm wrong, but I kind of get the impression that Directv is progressively discouraging cascading more and more, so I guess I wouldn't have been shocked if the DSWM13 didn't have those six outputs at all. Of course, the reverse could be true, and they will support cascading with the DSWM13, but the DSWM23 will leave those ports out. Oh well, once Stuart is allowed to give us more details we can learn what those ports on the DSWM13 are/do, and it will provide some direction for future speculation