1. Home
  2. Electronics Design
  3. Estimating maximum current of DDR3L memory banks

Estimating maximum current of DDR3L memory banks

Hi,

i want to estimate the maximum current that can be consumed of a DDR3L memory bank, consisting of 4 chips with 16 bits data bus width, so in total 64 bit. I need this maximum current for the power supply design (buck converter, 5V -> 1.35V).

I have read through the document TN-41-01: Calculating Memory System Power for DDR3, it shows how to calculate the system power in a very sophisticated manner. But there are a lot of variables i don't know at the moment, like memory access patterns etc.

What about using the highest value of the datasheet, which seems to be often I_DD5B, for example 250mA at a AS4C512M16D3L as the maximum current, so 4 chips consume at max. 1A in total (DRAM Controller of a SoC not included)?

And about the refresh, the datasheets note that a maximum of eight refresh commands can be posted, does this mean, the DRAM executes always only one refresh command at the same time, which refreshes one eight of the DRAM, probably one bank, and the next refresh command refreshes the next part of the dram?

Thanks in advance,

Steffen

Reply to
Steffen Koepf
Loading thread data ...

I recently measured some supply currents on a microZed board. It has a

7020 ZYNQ and two DDR3 ram chips. The Zed schematic and such are available online.

The board

formatting link

formatting link

runs the Linux that comes with the zed. Bootup takes about 1 second. Once it's running, some/most of the code is probably cached, so the dram isn't working very hard.

+1 core supply 0.32 amps for 1 sec, then 0.6A

+1.5 ddr supply 0.23 amps for 1 sec, then 0.36 amps

I'll try to get an Ethernet connection, and then have some guys flail it and play waveforms and things, try to exercize the dram more.

The TLV62130 switchers on the uZed are rated "up to three amps"

Let me know if you measure anything interesting.

I'm designing a new box that will use the 7020 and one DDR3 chip. "500 megabytes should be enough for anyone." My little switcher chips are good for 2 amps, maybe 2.5, so I should be OK.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com
Reply to
John Larkin

This is really stupid:

I'm planning to use a bunch of cute little TPS54302 switchers to make my supply voltages from a +24 source. I'll switch to +5, +3.3, +1.5, and +1.0. A separate Cuk will make -6, and a lot of linears will make

+22, +12, +4, +3, +2.5, -2.5, -3, -5, and maybe more.

The question was whether it's better to switch directly from 24 to

1.0, or to switch 24 to 5 and then 5 to 1.0. At 400 KHz, and 1/24 duty cycle, the switcher will be on for about 100 ns, going from +24 to 1.0.

So I hacked the eval board down to 1 volt output. But I can't get the loop compensation right: the output keeps oscillating. Envision an hour or two of fiddling. Finally, at the end of the day on Friday, I notice that my 1 ohm test load is a giant wirewound resistor. So I measure its inductance. 20 uH.

OK, fresh start on Monday.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

I always switch down in two stages, even from 12-16V. Lately I've been using 3.8V as my primary and either bucking or boosting from there. 3.8V because I could see intermittent dips to 4.5V and it's easier to just make the thing run that low than adding capacitance to hold the primary rail up. I try to synchronize all of the switchers, and generally run them at around 2MHz (so get really concerned about duty cycle - at both ends). I do use some cheap boosts that run at

3.5MHz, or so. I don't worry about them beating against the others much. The one thing to watch out for is coupling the primary and secondary switchers too tightly and make sure there is plenty of input capacitance on the secondaries.

Oops.

Reply to
krw

The TPS54302 is a sot-23/6 that's rated for 3 amps, realistically maybe 2.5. It runs at 400 KHz spread-spectrum, so I expect it to be very nice as regards radiated EMI.

We have had the debate, synchronize or not? I vote not, to keep the EMI things from summing up. The TPS can't be synchronized, so that's not an issue here.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics
Reply to
John Larkin

With a good layout, EMI shouldn't be a problem. I find it easier at high frequencies because the parts are so much smaller. Our EMI requirements are much tighter than class-B. The only time we have problems with power supplies is if someone botches the design badly (or some dumbass tries a 2-layer board).

Spread-spectrum is a no-no in our organization. Too much chance of beating causing beating in the radios or even audio. Synchronizing the switchers does several other nice things. You can set the frequency where you don't care as much about EMI and it's easier to track EMI down to the offending power supply, should there be an issue.

TI makes tons that are, though not in SOT23-6, obviously. They have a few new packages that further reduce EMI, as well.

Reply to
krw

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.