Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Rambus (RMBS) - Eagle or Penguin -- Ignore unavailable to you. Want to Upgrade?

To: Ali Chen who wrote (65371)	2/5/2001 1:57:34 PM
From: Ian Anderson	Read Replies (1) \| Respond to of 93625

Ali, You don't know what you are talking about! A RIMM does not have any 'Y' connections. The tracks run directly from driver to terminator, with the pins on the RDRAM presenting small lumped capacitances evenly distributed along the transmission line. This approximates very closely to a transmission line with a slightly lower characteristic impedance. The tracks are designed with a slightly higher than desired impedance, so that when the RAMBUS chips are present the desired impedance is achieved. I know this works because I designed just such a system in 1989 using fast SRAM. The only undesirable effect is that the propagation speed reduces slightly. In my design with 1.5pF per memory chip pin, this was around 90% of the expected speed. Relative to the speed of the memory core, this is not significant. FUD off somewhere else, and stop spreading disinformation here. Ian

To: Ali Chen who wrote (65371)	2/5/2001 2:13:51 PM
From: John Walliker	Read Replies (1) \| Respond to of 93625

Ali, Why are you trying to defend indefensible? I'm not. It isn't indefensible. You cannot match impedances in a simple "Y" connection as in Rambus RIMM. Period. This is a simple Physics. Period. Well, yes and no. Firstly, is is possible to match impedances at a Y junction, but only accurately at one frequency (and certain multiples). (Actually, a bandwidth of around 20% in practice.) For example the Wilkinson power combiner uses 1/4 wave sections of 71 ohm transmission line to match 50 ohm inputs and outputs in a Y configuration. This is not appropriate to the much wider bandwidths used in Rambus. However, in a Rambus system it is perfectly possible to match impedances up to an upper frequency limit. In RIMMs there is a short section of narrowed track just before and after each connection to a memory chip. This acts as an impedance matching element compensating for the extra capacitive loading caused by the device input/output and extra package parasitics. To keep these very small Rambus use Tessera uBGA packages. Rambus drivers have controlled rise time outputs so that the amount of energy at frequencies beyond the point where the compensation network becomes less effective is acceptably small. Now, even if in the almost hopeless case of 32 "Y" drops as in RIMMs, professionals found a way to handle the signals, what makes you think that it cannot be done for only 2 or 4 connection, as in DDR? The difference is that the Y branch is much longer and is loaded with a tree of multiple chip inputs (especially in the case of address lines). Hence the frequency at which accurate matching is possible is MUCH lower than for matching the tiny length of connection between the ball contact and the chip in a uBGA package in a Rambus system. Also, on data lines, the termination conditions vary according to which device is driving the bus, whereas with Rambus the constant current drivers present a high impedance while driving, so they do not upset the termination. Finally, I have never said that DDR systems would not work. Only that accurate impedance matching is not possible, something you obviously agree with. Your position in this discussion is a total loss. Nice try. Never mind. Sorry. Go get a beer, as Scumbria has suggested. Yes, excellent suggestion. John