I'm working on a purpose-built controller based on a BeagleBone Black and custom cape. The software driver seems to be robust - but the hardware is not. I'm breadboarding the few parts required on a cape, and I can see that I'm having timing/noise issues. Does it seem reasonable/unreasonable that this should work at 50MHz input clock frequency on the breadboard? I've read that breadboards have high loading (capacitance - not sure about inductance) and in the past I have had good designs fail on a breadboard but work fine on the final PCB. 50MHz is fast, but not that fast... From the specs, the design should work, but it's down to +/- 5nSec tolerances and I'm wondering if I need to look for a less timing-sensitive design approach, even when on the final PCB. The circuit is so simple, yet I don't get a reliable, clean 2-bit countdown. I'm stronger on the theoretical/mathematics side of design - any advice/input from experienced, hands-on PCB/hardware designers is appreciated.
Edit: I just noticed the attached schematic shows the two unused inputs as floating, but they are actually tied to ground.
In my experience, most breadboards top out at 2Mhz
ags0000 wrote:I'm working on a purpose-built controller based on a BeagleBone Black and custom cape. The software driver seems to be robust - but the hardware is not. I'm breadboarding the few parts required on a cape, and I can see that I'm having timing/noise issues. Does it seem reasonable/unreasonable that this should work at 50MHz input clock frequency on the breadboard? I've read that breadboards have high loading (capacitance - not sure about inductance) and in the past I have had good designs fail on a breadboard but work fine on the final PCB. 50MHz is fast, but not that fast... From the specs, the design should work, but it's down to +/- 5nSec tolerances and I'm wondering if I need to look for a less timing-sensitive design approach, even when on the final PCB. The circuit is so simple, yet I don't get a reliable, clean 2-bit countdown. I'm stronger on the theoretical/mathematics side of design - any advice/input from experienced, hands-on PCB/hardware designers is appreciated.Edit: I just noticed the attached schematic shows the two unused inputs as floating, but they are actually tied to ground.
On a production board your traces can be short - but may still need the resistors noted below.
I take it that breadboard means long (6in+) wires.
If so, you can have transmission line issues associated with reflection, etc.
For the breadboard, try inserting a 100 ohm resistor inline with each of the high-speed lines.
If the device you are connecting to (as in a micro maybe) - check if the IO pins have slew-rate control. If the do, engage them to slow the edges down.
Most Ethernet MACs that use RMII run at 50mhz and require inline resistors. Look into those designs to get an in-depth technical discussion on the subject.
Hope that helps.
Joe
I must be lucky to be getting any functionality with my prototype on the breadboard.
Thing is that I cobbled together another version on just a perfboard, and it still had noise issues. It was ugly to be sure, but I hoped it would have proven the viability of the design before paying for a full PCB.
That is why big companies rely heavily on board level simulation packages. And we DIYers spend money on test PCBs knowing that there will need to be a revision or two to get it right.
True, we spend as much time on the board simulations (SI and PI) as we do on the actual layout. If your willing to wait a few weeks for a board, there are several low cost PCB pool companies out there. One suggestion is https://pcbshopper.com/ , I've not used them myself but others have. Some vendors are JLCPCB, AllPCB, DirtyPCBs and OshPark. A coworker just did a board with Oshpark, $30 for three 4-layer boards. Granted they were small boards. I've not done a breadboard for years because the PCB fab is so much cheaper. But for me, time is money and I'd rather spend my time on other things as well.
HyperLynx DRC Free Edition for electrical design rules is being offered by Mentor, but only if your design package can output ODB++ or IPC2581 formats (not just gerbers).
All good points. I've used Oshpark before and have been happy. Even at their (not the absolute cheapest) prices, turn-around time is quick and I think it's worth having a real physical board to try rather than spending (wasting, in this particular case) time on breadboard/perfboard prototypes and/or SI/PI simulations. Not that I have a simulator/field solver on my personal machine.

