Filtered Driver Amplifier for use with SDRs

  • A few weeks ago I was chatting to a Lime mini SDR owner who expressed some concern over out of band spurious products and the inability to generate enough power to drive a EP-AB003 Wifi amplifier.
    My WiFi amplifier requires 175mW of drive while a recently completed DJ0ABR amplifier needs just under 200mW for 5 Watts output. There are ebay modules that are good to 100mW but driving these harder just leads to poor quality uplink signals to QO-100.
    So another project was born.
    Pictured below is a 45 x 43mm prototype PCB that has 2 stages of amplification together with a 2.4GHz bandpass filter on the input. The PA device is the Qorvo TQP7M9101 which is a 5V 250mW linear amplifier.

    The board is mounted on a small heatsink. This is not really needed for heat dissipation, more to provide a support for the 0.8mm PCB. I anticipate 200mW out with good linearity. Now to populate the PCB, see how well it works and correct any board errors.

    73 David G0MRF

  • Great solution! Filtering is always a good idea, as SDR transmitters can have annoying spuriouses at the output. The Lime SDR mini at 100% drive level - about 0 dBm output - have a spurious level of -55 dB or below.

    This is actually not too bad. Main spuriouses are spaced 40 MHz, so filter bandwith must be chosen accordingly. There are also close-in products not shown on this 500 MHz sweep.

    73 Ole OZ2OE

  • Thanks Ole.
    This is the first time I've seen a spectrum plot for the lime. Scary number of unwanted products.
    I see from the plot that the marker is at -55dBm but the level of 2.4GHz is a little unclear. Is it around -22dBm on the scale? If so then those products are about 33dB below the wanted frequency.
    Sadly the filter choices are normally limited to whatever is available on various commercial markets. I've selected the SF2124E as it's fairly large and not too difficult to solder.

    Many thanks

    David G0MRF

  • The one thing about 'social distancing' is that it gives you a little more time for projects.
    I finished placing the 22 components on the board without discovering too many errors.
    Everything fits and when I connect the board to 12V it consumes 125mA, which is as expected.

    Next I need to find a power meter to do some linearity and P1dB tests. I also need to see if the 2 matching capacitors on the input to the Qorvo TQP7M9101 are optimised values for the FR4 PCB material.

  • Hi Dave - G0MRF

    Nice amplifier. And you're right, we all get more time for projects these days.

    Regarding my Spectrum Analyzer picture in message #2, output on 2.4 GHz is 0 dBm and spurious attenuation 55 dB. I used an SDR Console TX setting of 100% drive level and DC-offset of 200 kHz. Not recommended for QO-100 operation, I only used this setting for certain measurements.

    What you see on 2.4 GHz at -22 dBc is actually an unwanted spurious 800 kHz above the wanted 2,4 GHz. Using a narrower span of 2 MHz all the close-in spuriouses show up! - and the -22 dB spurious you noticed is actually 800 kHz above the wanted frequency. See picture:

    In order to reduce the close-in spurious level, just reduce drivelevel to say 90% - that will help a lot! To completely remove them, set DC offset to 0 Hz, instead of my selection of 200 kHz. All the 200 kHz sidebands will collapse down to 0 Hz, ie. the carrier frequency. (DC offset selection is found under: Transmit/Options/Radio Configuration/LimeSDR/Calibration).

    Now with 90% drive and 0 Hz DC offset the spectrum looks like this - and note span is now only 10 kHz:

    There is now a "spurious" on 0 Hz - the carrier frequency - about -25 dB relative to PEP. I would prefer this to be -50 dB down, but don't know if possible at all with LimeSDR mini? Anyway it is not audible and may even serve as a tuning aid.

    In order to measure the two-signal IMD products, I use the DC offset function to move any "SDR-spuriouses" away from the wanted signal. With DC offset of 200 kHz they disappear from the picture and there is just the true two-signal IMD products at 90% drive level:

    Now, this is nice! 3'order IMD of -55 dB. However spuriouses are still there, just on other frequences.

    73 Ole

  • Thanks Ole, that gives a very good picture of the performance of the Lime mini SDR. Not running it at 100% seems essential if you want good IMD performance.
    Meanwhile, tests on the 2 stage amplifier have left some questions about which of the 2 stages are the limiting factor in the P1dB (linearity) performance. I decided that the only way to answer this was to build and test the PA stage alone and measure it's performance. Then I can select the most appropriate amplifier for the first stage.
    Pictured below is the Qorvo TQP7M9101amplifier alone. It gives me the chance to swap the matching components while looking at gain and output power. I expected some loss of performance due to the 0.8mm FR4 PCB material, but it's not as bad as it could be.

    The PA device output is internally matched to 50 ohms. However, the data sheet recommends matching on the input. Qorvo have suggested circuits for 2110-2170 and 2500 - 2700 built on microwave laminate, but nothing specific for 2.4G or the much cheaper FR4 PCB material and so a little bit of exerimentation is needed. Initial results for gain look OK at 15.2dB using 'standard' capacitors rather than any expensive RF products. Also the 15.2dB does not account for the small amount of loss in the 5 cm length of UT-086. Qorvo measued a gain of 16.5dB at 2.5G, so I'm about 1.3dB low. Actually, given the FR4 and best guess matching, that's not too bad.

    Now to look at P1dB which needs work to calibrate out all the losses in patch leads between sig gen and power meter.

  • Measuring linearity with questionable power metering ability has not been conclusive, So the board has been posted out for testing. ( Tnx Mike). Meanwhile I have ordered a batch of components and have begun making a few changes to the PCB. The major change is to the pad layout for the first amplifier. I want the layout to allow different devices to be fitted to the same board. Hopefully, it will accept either ERA or Gali devices from Mini-Circuits. This should allow any one of Qorvo PAs to be used. TQP7M9101 (250mW) TQP7M9102 (500mW) or the TQP7M9103 ( 1W ). All are available from Mouser. :)

  • Hello Ole OZ2OE,

    I will try out your tests on my Lime-Mini to see if my results are similar. How have you got the two tones set?

    When you mention 90% drive, is that with your SDR Console "Drive" configuration for the Lime set to the default spread. With mine like that the 90% equates to just over 6dB backed off from maximum.

    73 Chris

  • Hi Chris, G4SDG,

    It will be interesting to see your measurements on the LimeSDR Mini. The two-tone signal I'm using, is the default "Tone" button on SDR Console.

    The 90% drive level is based on default SDR Console setting - and it also gives me 6 dB back-off.

    Some time ago I measured LimeSDR Mini output vs. drive setting in % and came up with the following curve measured at 2.4 GHz and for levels between 50 to 100%. Blue line is output and orange shows stepsize. (see also DD1US homepage under "What's new" of 20'th Nov. 2019)

    Output increases smoothly most of the time, but there are jumps of 2 to 4 dB at certain %-settings - which can make output adjustment a bit jumpy! Especially from 87 to 88% where output changes by 4dB. I don't know if all LimeSDRs behave the same? - maybe you could look into this?

    73 Ole

  • PCB has now been changed to allow MAR or Gali style devices to be fitted in the first stage.
    Boards and components have been shipped......Now I just have to wait for the postman. At least I will not miss the deliveries. - I'm not going anywhere :(

    While I'm waiting, I've just changed the 10MHz TCXO in my upconverter. The ECS and Fox products both had noticable jumps as they corrected for temperature change.
    This device from Murata, £7.99 from Digikey, is 280ppb and seems very much better.

    73 David

  • New driver ampliifer PCB arrived. Only minor changes on this version to allow ERA or a Gali device for the first stage. The silk screen has the correct component designations and now makes sense. When a Gali 5 is used the overall gain of the board with a Murata bandpass filter is 25dB. An ERA2 gives around 2dB less. Testing is still ongoing, but Mike G0MJW reports good SSB into QO-100 using just this driver amp. - Of course a nice 2.4m dish helps. Thanks Mike. Should be good to go between a Pluto / Lime SDR and a 5W PA.

  • Now the part I'm not so keen on. Mechanical assembly.

    2 Members of my local radio club want a low cost way to get onto QO-100. Seen here are the modules for a 70cm to 2.4GHz upconverter.

    Top row: e-bay ADF4351 oscillator board. (1967.5MHz)
    Upconverter PCB with RMS30 mixer and PIC12F629 controller,
    Filtered Amplifier board
    EP-AB003 Wi-Fi Amplifier

    Bottom row is the same except for the PA which is the kit version of the 5W DJ0ABR design

    I've probably been a bit optimistic thinking that Wi-Fi amplifier is going to fit in the Hammond box :(

    73 David

  • Hello David,

    The filtered amp with good gain is a much needed item. I've dabbled with a few uplink options for QO-100 from on Day #1 an SG-Labs U/C, Spectrian amp and a 25 turn helix to the current set-up of a Lime-Mini, MHL21336 and "8W" Wifi amp to a 90cm off-set, and along the route I tried the "Easy-100" modular system with ADL5350 mixer and a handful of low power modules and WiFi amp. The MHL21336 works well but does get warm (OK - hot) and my version of the 5W DJ0ABR amplifier was a few dB short on expected gain. I have three "8W" WiFi amplifiers and none of them get produce more that 4 watts saturated and one of them was 4dB less gain than the other two. Overall, gain in the middle section has generally been a problem.

    Ole OZ2OE - I checked my Lime Mini and saw similar jumps in the output settings especially in the 80-90% area. Quite apart from this hardware problem I think that SDR Console would be easier to use if it had Up/Down buttons rather than a slider. Perhaps my old fingers are having trouble using the mouse with enough sensitivity!

    I've got a problem with my analyser so I haven't been able to check the output spectrum.

    73 Chris

  • Hello Chris.
    There are now a couple of good choices for 3 - 5W PAs. My 5W in the photo above was also short of gain but still produced 5W with a fraction under 200mW drive. I think most of them are better than that with an input requirement nearer 160mW.
    The filtered driver amplifier uses a "quarter Watt linear amplifier" and it can produce 250mW, but not with a linearity that would be acceptable on the transponder. However, as a driver for the amplifiers that need <200mW it is very good.
    I'm in the middle of producing an article for AMSAT-UK and as soon as the Gali5 amplifiers arrive I'll send off a batch of kits to the AMSAT-UK shop.
    Looking at the DJ0ABR board, I wondered if I could mount the driver amp on the PA PCB. (?) The input stripline looks to be just coupling the SMA socket to the input matching. Given the loss in FR4, it may even recover some of the gain.

    Thanks for the info on the MHL21336. Always good to hear other people's experiences. With the EP-AB003, I was considering bypassing the 2 TX/RX switching ICs, That would probably improve performace a little.



  • Finally, after a lot of postal delays the final part for the kits finally arrived this weekend. So, now available at the AMSAT-UK shop.
    Design is reliable with no signs of instability. Gain is around 25dB. Power output is 200mW with good linearity and 250mW near saturation.…it/p3815740_20451527.aspx

    Still in lockdown........what shall I do next?

    73 David