Graphic errors / instability with Indivision AGA MK3

Caution: Non registered users only see threads and messages in the currently selected language, which is determined by their browser Please create an account and log in to see all content by default. This is a limitation of the forum software.
Don't Panic. Please wash hands.
  • Hiya,

    Finally the ferrite beads have arrived, and I've had the time to try them... and they seem to have fixed the issues with both my 1D4 1200 boards.

    Interestingly, one of the boards had a resistor at E127R, but the other had what looks like a ferrite there already. Both were exhibiting the same graphical glitches, so I thought it was worth a shot to swap out both... and both boards now seem to be 100% glitch free. So a dodgy ferrite bead maybe? The original solder still looked good and there were no visible cracks or anything, so I think it was factory fitted.

    I don't know if it's possible to measure the old ferrite bead I took off in any way useful (it has 0 ohms resistance on my test meter as expected) but I still have it at the moment, so if anyone wants me to check anything about it please let me know. Something about the ferrite bead seems to have been equally guilty of causing the glitches, as the 27 ohm resistor which was on the other board.

    I now don't seem to need to turn on any of the CCK pullup or capacitance options in the Indivison tool either.

  • So a dodgy ferrite bead maybe?

    Or maybe just different specs. I don't know what the original specs are, but with ferrite material advancing all the time, it's very possible that the old one has a smaller impedance at 100MHz and possibly higher DC resistance, and as a direct result, a totally different frequency curve. Note that "resistance" is what you measure on a resistor, which ideally does not have any dependence on frequency. The ferrite bead has inductive properties, so you want to measure reactance, which is the frequency-dependent part. The sum of both is the impedance of the complete component. You normally find a graph in the datasheet that shows the individual shares of the contributing components.

    I don't know if it's possible to measure the old ferrite bead I took off in any way useful

    If you want to draw a frequency-impedance curve, you'll need either a component tester, or a function generator in addition to your oscilloscope. And there's a lot of mistakes you can make.

    In my opinion, it gets way too complex if you want to go that deep, as you're not looking at a simple point-to-point connection. There is multiple targets on the CCK signal, and each target has an equivalent circuit of resistor (for the trace), inductor (again, for the trace) and capacitor for the pin at the target. Parts of the signal will reflect at each target, contributing unwanted parts onto the source signal that end up at the other targets. Further, each target will have different switching characteristics, so you need to take that into account as well.

    My guess is that the ferrite bead will take care of the higher frequencies - which are mostly the reflected parts that you don't want anyway. It will "be" a high resistor for the reflected parts of the signal, reducing these effects.

    To make it short, it would be great if you posted the exact part number that you've used for the ferrite. I'll take a look at the data sheet and see if I can find one or two similar parts, so people have something to work with if they experience the same glitches.

  • The ones I installed were...

    MI0805K400R-10 Laird Ferrite Beads 40ohms 100MHz 1.5A Monolithic 0805 SMD

    They were the only ones Mouser had in stock at the time, in the right size, same Mhz and _roughly_ similar resistance to your recommendation. So I figured they were worth a try.

  • MI0805K400R-10 Laird Ferrite Beads 40ohms 100MHz 1.5A Monolithic 0805 SMD

    What a nice example for a good data sheet!…805k400r-10-datasheet.pdf

    Not only do you have all relevant information on a single page, they even tested impedance over different current loads, and also mentioned the test equipment used (bottom left). The 50mOhms figure appears to be unique, as I could not find an equivalent Bourns or MuRata part. Only the Chinese brand Meled has a near-equivalent with part# MLB2012-400 (that's rated max. 0.6A). Anyway, I normally start at 56R@100MHz, and the fact that you've been successfull with a lower value shows that it's unlikely to be *that* critical. The high current that you've chosen probably has no influence; I'd expect it to be only a few mA that pass through the part.

    So to anyone reading this: While this is now a known-good part, it's most likely not the only one. The same ballpark of impedance at 100MHz will most likely do the trick.