SNA – A Disappointing Discovery!

4 Apr

I spent a good deal of time yesterday getting the SNA hardware rebuilt on a board that fits inside an Altoids tin. I hoped that on Monday, I could release an alpha version of the hardware, firmware and software designs so that others could reproduce it.

Today, while cleaning up what I thought were final details, I decided to check the flatness of the DDS module from 3-30 Mhz. Oh no!!! The output level of these cheap Chinese modules seems to vary by almost 10 db, dropping off severely as frequency increases! I’m not sure where the rolloff is occurring, but it seems to be in the lowpass filter. Now, I know that the 9850 chip is much better than this, as I’ve used it before. The last time was in the first incarnation of the SNA, using a DDS-60 board from AMQRP.

So now it’s back to square-one for a while to make some investigations of a good Analog Devices chip vs what came in the cheap Chinese DDS module from Ebay – maybe it’s a knockoff chip? This could take a while. Meanwhile, I’m going to compensate for it using the SNAP! software’s “normalize” feature. This feature takes out generator level variations so the displayed output is flat. I’m sure it can correct the display flatness, but it will result in lowered dynamic range as the RF level drops off.

As Gilda Radner used to say on Saturday Night Live, “It’s always something!”



8 Responses to “SNA – A Disappointing Discovery!”

  1. Mark April 4, 2015 at 8:50 am #

    I suspect the output filter is not designed for 50R. Normally the ebay boards need the LPFs removing before you get uniform output across the frequency span.

  2. DuWayne April 4, 2015 at 8:56 am #

    When I was first working on my SNA Jr. I tried both the filtered and unfiltered outputs of the 9850 module. The unfiltered seemed to not drop off as fast, so I used that as my output. There are 2 methods that you could easily use, both require a calibration sweep. If the drop off is linear you could just compute the db./khz. and multiply each frequency point by that value. Other method is what is used by the NAT-SNA group, do a sweep of the frequency range, normalize the values of the points to a +- offset. Then when doing the the actual sweep add the offset for each point. I updated my Blog with the schematic, board layout and code for the SNA Jr. at if you would like to take a look at it

  3. Peder April 4, 2015 at 11:07 am #

    Hello there.
    This solution might be worth trying:
    It is an automatic level control for an AD9851 DDS consisting of some opamps
    I have not yet tried this soulution, but it sure looks interesting.

    Best of luck
    Peder, SM0GNS

    • w3jdr April 4, 2015 at 11:14 am #

      Thanks for the referral Pedar, but I prefer simple designs that work correctly over bandaids that add parts and complexity. I’ll get there.


  4. kd7ura Mike April 6, 2015 at 1:18 pm #

    I read somewhere that the chinese ebay units have a mismatched output filter , was designed for another chip and somewhere outhere are a few mods to flaten out the filter responce or bypass it all together, cut the trace near the first cap and rebuild a better one.

    • kd7ura Mike April 6, 2015 at 1:19 pm #
      check this link for a bit of details about the ebay ad9850 boards

      • w3jdr April 6, 2015 at 1:27 pm #

        Hi Mike

        From what I can see, I’d expect 3-6 dB of rolloff over the 3-30 MHz range just due to sampling theory, so my hunch is that the remainder is in the filter. As to ‘mismatch’, there’s really no such thing in this case. The output of the DDS is a current source, which is theoretically infinite. The termination resistor you put at the output turns the current into a voltage and sets the source impedance for the filter. The filter on these boards has a 200 ohm source resistor and a 200 ohm termination resistor, then it goes into a high impedance comparator to square it up. If you take your output from the termination resistor, then any further loading (eg, 50 ohms) actually causes a mismatch to an already terminated filter, but LTSpice modeling indicates that the effect on frequency response is minimal. I suspect low-Q 0403 SMD inductors.

        In any case, I don’t want other builders to have to fool around with massive mods to the stock DDS module, so I’m pursuing a very simple workaround (see my previous post).


      • w3jdr April 6, 2015 at 1:29 pm #

        Yes, I’ve seen this article Mike. I take some of it with a grain of salt.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: