Signal integrity with multiple devices.

I am attempting to learn more about signal integrity for a microcontroller design I am working on. I have done simple designs with just one device attached. Now I want to add more devices. I have done spice simulations and see the increase in noise and ringing as more devices are added. What I am not sure about is how to correct it.

Simple series termination would not seem to be sufficient. However when I look at the few microcontroller designs that have schematics available, that is all that seems to be used. I have seen nothing that uses AC termination for example.

I have tried to find information on designing for multiple items attached to a bus, but it seems to be hard to come by just searching on google. I find plenty of stuff for one or two devices, but what about three or more? Should I use tees or buffers or what. What about stability when component values are varied.

Any recommendations on sites or books that can help me understand how to deal with the >3 case would be good.

Reply to
db
Loading thread data ...

I presume we are talking about a databus system here. If this is for an RS485 style network then perhaps you should look at the application notes on Linear Technology's site. They are sometimes quite useful.

If it is any other type of bus come back with more details and you'll get a number of others responding.

--
********************************************************************
Paul E. Bennett ....................
 Click to see the full signature
Reply to
Paul E. Bennett

have

it.

Yes this is for a databus. Basically there were between 4-6 devices on the bus. Specifically the microcontroller(Sharp 79520), sdram,sram , flash, fpga,usb, compact flash and maybe a cpld.

I have pulled values from the manufacturers sheet for pin characteristics and calculated for the impedence of the trace.

So far I have tried series termination and schottky diode termination in the spice simulation. The results look pretty poor. The terminated end usually looks pretty good, but the middle would almost certainly generate double clocking.

I did find a sheet from phillips (AN246) with a pretty good coverage of different termination issues in a multiport situation and what the simulations look like. It is good to see some confirmation of some of the waveforms I saw in my simulations with what is in the Philips app note.

I guess one specific question I would have is what connection topology should I be trying to use. So far I have simulated a bus with the components stubbed in and also in parallel. Should I try some configuration of tees or a star topology.

Reply to
db
[%X]

[%X]

So this is a serial device bus similar to IIC.

Essentially keep the traces parallel and ensure they visit each device in turn (no spurs or T's). You should also attempt to keep each trace very similar in length. Terminate the end with a suitable termination network and you should minimise the reflections along the bus. There are probably app notes with various companies that do these type of devices. Does the Philips AppNote say anything useful about techniques?

--
********************************************************************
Paul E. Bennett ....................
 Click to see the full signature
Reply to
Paul E. Bennett

I have been able to find several now that have been helpful and have been able to get better results from the simulations. One mistake I was makeing was not accounting for the capacitance of the attached devices and accounting for how that affects the impedence of the line.

The best notes I found were Phillips AN246 Transmission lines and terminations with Philips advanced logic families

Motorola AN1061 Reflecting on transmission line effects. AN1051 Transmission line effects in pcb applications.

The boards will be quite small. I am shooting for the 3x4 inch size, so I may not need terminations at all. However I want to be real sure. I do not want to spend the money to build these boards and then discover they don't work because of sigint problems.

I still have a more simulation to do, but the problems seem a little more manageable now. Based on what I have simulated so far AC termination seems to provide the best results. Parallel was good also, but I do not want the power drain. I saw a lot of ringing with diode termination. Series seemed to be the worst, but that does not seem to be recommended except in point to point situations.

Reply to
db

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.