MCF52259 mini-flex-bus series resistance. Is it really necessary?

[drummer]

I am re-spinning my MCF52259 board for a third time now and I am thinking about dropping the 22 ohm series resistance on the mini-flex-bus pins.I originally based my design on the MCF52259 EVB schematic that used the bus for external flash. It had the little resistors. My board connects to 8 static ram chips. I use a multiplexer to control the chip selects.

[PabloBaena]

Hi Rick, I have same question for phyceiver resistors. I'm designing a MCF52259 board for my first time, and would like to drop the series resistor to the phy. What would you suggest?

Thanks,

[drummer]

I only use series resistance on the clock signals TXC (FEC_TXCLK) and MDC (IRQ5). The distance is quite short between the MC52259 and the PHY. My emissions were so hi on my prototype where I had no resistance I was suprised that aircraft weren't landing on my building. I didn't notice any problems with the PHY functionality. Those clock signals apear on the spectrum at 125MHZ. After re-spinning the board with the manufactures recommendation for placement, the aircraft now pass safely above me.

[PabloBaena]

Hi Rick, thanks for your prompt response!

[PabloBaena]

Hi Rick,

I wonder if you can give me some tips for my problem...I just received my own first board for the MCF52259, but BDM programming is not working, CW 10.3 says that there is no target connected.  As example for my design I followed the Tower circuit. Now I'm going nuts, can't find what's missing! Already mounted three boards without success. I connected the DSI, DSO, DSCLK, BKPT, TCLK, PST[3..0], DDATA[3..0]...pullup for reset, etc....

What would you suggest? Or maybe even share the simplest circuit for debugging...:smileyhappy:

Thanks...

[drummer]

Check the PSTCLK signal on pin 24. It shouldn't be over 80MHZ.

Choose your crystal and phase lock loop settings carefully.

Use 4.7 K or less for pull ups.

Pull up BKPT, DSCLK, DS0, DS1 and reset input.

Make sure there isn't any capacitance on the reset input.

[PabloBaena]

Thank you Rick ! I solved it with your tips.

[TomE]

> My board connects to 8 static ram chips.

That has to be a fairly long set of tracks.

Stan's point about ringing is a good one. You should look for it, but you need a very good high frequency oscilloscope to see this.

I worked on an MCF5329 and the bus on that only connected to two Flash chips and one SDRAM chip. Various boards were unreliable, corrupting memory and crashing and so on. The problem was caused by undershoot and overshoot on the bus lines. These get clamped by protection diodes in the chips on the bus, and those current spikes can seriously disturb their inner workings. These problems were intermittent and depended on chip brands and temperature. A VERY nasty problem.

A 3V transition on a long unterminated track with a fast driver will reflect and will attempt to generate voltages of -3V and +6V. Only the diodes will try to stop this. The "minimum and maximum" voltages on the data sheet are there for a reason. They REALLY mean it sometimes!

I fixed that problem by changing the CPU's Pin Drive Strength on those pins from its power-on-default of "High" to one of the lower ones. Only the RAM Clock pin had a series resistor. The Data, Address and Control lines didn't.

You can change the Drive Strength on the MCF52259. The PDSRL and PDSRH registers control this. BUT... You can change the drive-strength on the address bus on signals going from the CPU to memory and on Data Writes from the CPU, but all READS from the memory are driven from the memory chips, and they don't have any drive strength adjustments. So even if the drive strength fixes the signals sent from the CPU, series resistors are the only way to cut down the ones from the memory.

Overshoots and Undershoots also contribute to EM Interference. The resistors should help reduce this. Even if your boards don't have to pass emission tests, this should be something you worry about.

Tom

[drummer]

Just the act of putting my oscilloscope on the signal changes the signal. Its almost like a quantum paradox. I run extensive memory tests and see no problems with the 22 ohm resistor so why fix what ain't broken? Thanks for your help.

[TomE · Accepted solution]

> Just the act of putting my oscilloscope on the signal changes the signal.

For this sort of thing you need a better and more expensive oscilloscope and probe.

With these sort of design issues, no tests on your prototype boards will ever fail, but some of your first (or second...) production batch will. You can test one board, but you can't test across the full production variation of all the chips in use, the change in board capacitance between batches, the full temperature and voltage ranges of everything and the RAM chips you'll substitute from another manufacturer a few years down the track, or your supplier substitutes.

The only safe way is to obey all of the manufacturers hardware specifications, and the "min max voltages" are very hard to check and follow everywhere.

Tom

[scifi]

The resistors should reduce the amount of ringing. Whether or not ringing is a problem will depend on the layout of your board, I guess. You can try replacing the resistors on one of your boards with shorts and run some stress tests. This will give you some idea about whether it can run without them. Use an osciloscope to estimate overshoots/undershoots with and without the resistors.

Once I connected an LCD with a parallel interface to the mini-flexbus of the MCF52259, no resistors. It worked fine.