SM1000 Part 5 – Mysterious Triangle Waves

Getting close to Beta now.

I have just spent a week tracking down a mysterious 256 Hz low level (30mVpp) tone that ended up being a software configuration error in the DAC initialisation code! However before that I thought it was a noise issue caused by PCB layout, or the wrong sort of bypass capacitor. I also popped another uC looking for it, costing a few days while I had the dud uC removed (thanks Matt) and I loaded a fresh one. Think I better stick to software…..

To get to the bottom of the problem I partially loaded a “minimal” 3rd SM1000. Just enough parts to make the uC run the dac_ut unit test program, power supply bypass, STLINK, crystal oscillator, boot mode resistors. Only about 1 hrs soldering, quite remarkable really.

Here is the mysterious triangle wave:

Michael Wild DL2F2 in Germany has built his own SM1000, using a PCB I sent him and the parts off our Digikey BOM. That’s a pretty exciting confirmation of the design. Having a another unit running has been very useful when making comparative tests. Here is Michael’s SM1000:

Michael, Rick KA8BMA, and Matt VK5ZM have also been very helpful in suggestions when I have been debugging. Thanks guys! Matt has also been explaining to us how ceramic caps can lose up to 80% of their C as their working voltage is approached. So best to use tantalum or electrolytic capacitors for noise sensitive applications like power supply and analog rail filtering (e.g. C25, C28, C29, C10, C12 on the SM1000). I could hear the noise drop when I soldered an electrolytic capacitor across C25 (input to the SM1000 switching power supply).

Right now I’m trying to reduce some remaining noise sources and soon hope to test over the air to see how the SM1000 works with large RF fields nearby, and also ensure it works OK with a couple of HF radio models. Rick is busy updating the schematic and PCB with a bunch of small changes we have picked up working on the prototypes.

We have a quote for the Qty 100 beta run, and are still on target to ship SM1000s in 2014. As soon as we have completed testing the prototypes we will kick off the Beta run and I’ll start taking pre-orders. Lots of work required on the software, but I figure that can wait until the Beta hardware is getting made. I want to take the open source approach of release early and release often, and that means getting betas into your hands ASAP.

2 thoughts on “SM1000 Part 5 – Mysterious Triangle Waves”

  1. Excellent work there David..

    I can’t wait to see the beta units being released.

    All the best, Walter. K5WH

  2. Hi David
    Nice work- a few points about the circuit-

    0) The triangle wave noise- bear in mind the way to make a triangle wave is start off with a square wave and stick it into an integrator- that might provide some hints to where that waveform is coming from…

    1) Don’t forget snubbers on the switched mode supply on board. Across the buck diode(whether internal sync or discrete) (I assume it is a buck converter) helps alot. the stray inductance of the inductor tends to ring with the stray capacitances when the lower switch/diode turns off. Sure, reduces efficiency a couple of %- bu that’s not a big deal here…. Also if there is a boost capacitor,(likely) a 2 ohm resistor in series with can also improve EMR. Important to read chip datasheets regarding layotu and go for absolute physical minimum distance and get the current paths right. There is about a 20dB noise difference in layout performance…

    2) The comment of avoiding ceramics. I think this is not the full story. Ceramics are essential for good performance in modern fast switched modes.

    I would recommend X7R/ X7S / X5S dielectric – sounds like your guys have been using Z5U or Y5V ceramics if they are seeing temperature dependence. There is really no justification these days for using the high K , high Tc caps like Y5V.

    However they can be too good in the ESR department , so an electrolytic in parallel has sufficiently poor ESR as to dampen any ringing…. I use standard non low ESR 105 degC electros for the job, or some average tantalums. if using tantalum on the input, be sure to derate voltage 2x.

    3) Be sure to consider conducted noise paths out of your PC into the antenna system. A even mode (common) choke is usually the recommended medicine however there MUST be a low common mode impedance on the load side for a common mode choke to function effectively.

    73
    glen english VK1XX

Comments are closed.