[Open-graphics] DDR Vtt termination ?
"Ing. Daniel Rozsnyó"
daniel at rozsnyo.com
Tue Oct 16 09:35:19 EDT 2007
Sebastien Bourdeauducq wrote:
> On Tuesday 16 October 2007 13:57:54 Ing. Daniel Rozsnyó wrote:
>> Sebastien Bourdeauducq wrote:
>>> Hi,
>>>
>>> I've had a look at the board schematics from Transversal Technology, and
>>> I wonder why termination resistors that connect the DDRAM signal lines to
>>> a 1.25V (Vtt) supply are not present. From what can be read in various
>>> sources, DDRAM uses SSTL-2 signaling which requires that resistor (see
>>> http://download.micron.com/pdf/presentations/dram/plat7justin.pdf p.8 for
>>> instance).
>> The presentation is about multiple DDR DIMM's, not single chips.
>
> That's the same, DIMM DDR memory modules contain only the chips plus the
> series resistor in the 20-30 ohm range, and a serial EEPROM to store
> information about the chips used. See an example schematics at
> www.elixir-memory.com/products/file/Elixir_
> DDR_512MB_B%20Die_TSOP_SODIMM_R1.0%20-%20160806.pdf
> p.5 ; and a quick visual inspection of a memory module from a PC shows the
> same.
Not exactly the same - motherboards (and your initial link to a
presentation) contain multiple slots for DIMM modules. That setup
includes connectors and signal integrity could/will be a problem.
>>> Is that because the PCB traces are short enough that you can get rid of
>>> that resistor and still get good signal integrity ?
>> I think this is because point-to-point links.
>
> What do you call a "point to point" link ?
> See the SSTL-2 spec at http://download.micron.com/pdf/misc/sstl_2spec.pdf.
> There is nothing about that "point to point" thing.
P2P for me is a `direct chip to chip trace without branching`, in this
case also for 1 driver and 1 receiver per line (not counting address
lines now - they are a mystery even after reading the above pdfs, that
the termination is same for load of 2 (data) and 16 (addr) when using 2
modules with 8 chips..).
>> And the chips already
>> contain some termination circuitry.
>
> Are you sure about this point ? I've always read the opposite
Well, that is valid only for DDR2
(http://en.wikipedia.org/wiki/DDR2_SDRAM : "..DDR2's bus frequency is
boosted by electrical interface improvements, on-die termination .. ").
Not sure if they talk about the CLK pair only or all the signals.
More information about the Open-graphics
mailing list