Interference from SDRAM

[stephen168]

I am debuging the low sensitivity of the WiFi from our product. There is a SDRAM inside the product and the SW is running inside the SDRAM. The DSP and SDRAM are put inside the shielding can, it worked very well. Recently, we changed the SDRAM from brand A to brand B, the sensitivity of WiFi is degraded by 10dB. Provided that the SW is same, the clock rate is same. When I swap the old SDRAM to it, the sensitivty is as good as before.
According to the data sheet, they are nearly the same. The current consumption from brand B(poorer sens) SDRAM is 20% higher by measurement.
Does anyone know what make it so different for interference?

 

[VEBill]

Is the RAM on those little daughter cards like used in a PC, or just RAM chips? If it is the cards, then perhaps the layout of the traces is suboptimal from an EMI point of view. Another likely possibility is that the timing has changed.


(Just yesterday, my coworker told me that Windows XP caused less television interference than Windows ME. He lives in a rural area and has very weak over the air TV signals.)

 

[stephen168]

It is a RAM chip.

 

[VEBill]

I understand that different chips can force the memory controller to change the timing. I know that this happens with PC ram sticks.

Maybe the first step would be to use a spectrum analyzer to sweep units with brand A and with brand B. Then line-up the spikes with the known clock frequencies, look for the change, and see it an explanation falls out.

Another option would be to simply strike brand B from the parts list and call it a day. This is not an uncommon solution.

The 'net is full of stories about bugs that cropped up due to subtle hardware differences. The March 20, 2008 edition of EDN (just arrived) has an article at the back called "The case of the bad memory chip" where the start-up uninitialized contents of DRAM chips varied and uncovered a subtle SW bug.

Stand-by to see if anyone else has some other suggestions.

 

[Higgler]

The 20% more power drawn may hint that it'd be a more powerful antenna radiator.
I know that having a transmitter antenna and a receiver antenna inside the same shielded can makes energy couple much more to each other. Orientation of the two antennas can really change the coupling.

From an antenna coupling point of view, if two antennas are aligned (!!) or misaligned (!-), there is a great difference in RF coupling (20-40 dB). Maybe one of the SDRams has a different direction of wires/current flow than the other.

Having them in separate shielded can's may solve the problem, or adding some microwave absorber between them inside the shielded can will help enough.

kch

 

[VEBill]

On a dB scale, 20% is next to nothing. My gut tells me that something else is the primary factor.

 

[biff44]

I would look long and hard at the interface between the sdram and dsp chip. If there is more current flowing, then there will be more RF leakage. You want as little high speed current flowing in the printed wiring board ground plane as possible, so that it does not look like an antenna.

You say that the design is the same except for the new SDRAM, but is it really? Did your layout guy change ANYTHING? A cut in the ground plane, re-routing some traces, etc, could make a huge difference. You are trying to avoid something called "ground loops".

You might be able to slow down the edges of the digital signals flowing between the two chips, which would dramatically change the ammount of power floating around at microwave frequencies. Maybe some 20 ohm series resistors in the digital path.