Skip to content

VU3CER's Robust Class-D HF 'QRP' PA for 40-20-17-15-10 bands. Yes, IRF510 can work on 10m with a minimal design ;) Input is 8-10mW.

License

kholia/HF-PA-v5

Repository files navigation

HF-PA-v5

VU3CER's Robust Class-D HF PA for https://github.com/kholia/Easy-Digital-Beacons-v1 project for 40-20-17-15-10 bands.

For extra fun, use this PA with https://github.com/kholia/HF-Balcony-Antenna-System.

This design aspires to be the standard mW-to-QRP-Gallon HF PA in the ~600 INR (8 USD) design space. Well, it does much more than QRP now...

Design Constraints:

  • PCB is to be homebrewed.

  • Only one PCB layer (side) is available for routing work.

  • IRF510 MOSFET needs to be on an edge to allow usage of heatsinks.

  • Typical external linears (e.g. PA150 HF Linear Amp) need 2W minimum to function. So we need a minimum of 2W of RF output from our design. Even at 28 MHz.

  • Minimal part count is strongly preferred. Low cost, and ease of availability are strong driving and deciding factors.

    The cost of this entire HF-PA-v5 system is less than one RD16HHF1 MOSFET ;)

  • Reproducibility is a MUST! We focus a lot on this aspect by documenting each, and every detail possible.

  • Repairability in the field is a MUST!

    Dhiru's finding: Repairing homebrewed PCBs is so much easier, and fun than reworking commercial lead-free PTH PCBs.

  • No moving mechanical parts are allowed in the design - this rules out usage of (failure-prone) active cooling options like fans.

  • We do NOT like the robustness of QCX, QDX, uSDX PAs. Removing heat from TO-92 devices in a reliable way is not a trivial problem.

Note 0: All screenshots are clickable and zoomable.

Input: Few milliwatts is fine. Note: Si5351's output is around 8-10mW.

Output: See https://github.com/kholia/HF-PA-v4 for results.

Note 1: The output power is software controllable (via PWM grounding) for flexibility.

Note 2: The idea behind making these (and my other) design notes publicly available is to deliver REPRODUCIBLE, flexible, home-brewable, standardized, reasonable, and cost-effective solutions for some common problems. I include many references, simulations, and notes in my projects to enable you to do your own (better) designs - do share them ;)

Robustness Notes:

  • Ensure that the IRF510 is fastened securely to the heat sink.

    Use lock washers, and thermal epoxy adhesive to ensure good mechanical contact.

  • We recommend operating at 24v instead of >= 29v.

Please see the older https://github.com/kholia/HF-PA-v2 project for the power-supply (over-current protected), and LPF designs.

Schematic:

Schematic

PCB rendering (v5):

HF-PA-v5

Homebrew version of the same PCB:

HF-PA-v5 Homebrew Version

Note 3: The HF PA is connected to a 20m dipole over 55 meters of HLF-200 coax.

Note 4: The PCB is done using KiCad 6.x - a free, and open EDA software.

Note 5: This PA can be homebrewed from scratch in around 4 to 6 hours (one afternoon, and an evening).

Notes

[From QRP Labs] Si5351A outputs a 3.3v square wave. The power output of Si5351A is around 8-10mW (10 dBm). The BS170 based driver circuit amplifies this to ~200mW (23 dBm) for a gain of around 13 dB. The IRF510 finals provide a gain of around 14dB and produce around 5W of RF power.

Input drive matters the most in getting good RF output from IRF510. This is precisely what Si5351A lacks (which is perfectly fine), and why the BS170 driver stage is necessary.

Important: Don't exceed the IRF510's gate limit of 1-2W. Keep the output of the BS170 driver in check, and within the limits of IRF510.

For safety, keep the output of BS170 less than 22 or 23 Vpp by controlling the BS170 PA voltage (via the LM317 trimpot).

PA tuning process

  • Remove the BS170 and IRF510 MOSFETS from their sockets.

  • Set DC gate bias for BS170 to 0v. Now increase the bias (via trimpot) until the current consumption increases slightly (QRP Labs' process).

    You can also start by setting the gate bias to 2v directly ;)

    The final bias value can be around 2.1v to 2.2v in many cases.

  • For IRF510, set gate bias to 3v. This ensures that IRF510 isn't turned ON - double-check this by monitoring the power consumption.

  • Connect the MOSFET back, and measure RF output. Happy with the output power? - If yes, you are done else follow the next steps.

  • Remove the MOSFET, increase the gate bias voltage in very small increments (0.1v types).

  • Connect the MOSFET back, and measure power consumption without any RF input. Ensure that MOSFET remains OFF without any RF input.

  • Apply RF input, and measure RF output.

  • Repeat these steps as needed.

The QRP Labs' procedure for bias adjustment and this document seem to have a better description of this process.

PA BOM

  • 1 x IRF510 (Genuine Vishay parts are tried-and-tested) - 22 to 50 INR

    • The SEC (SEC Electronics Inc) branded IRF510(s) also work great!

    • The SEC IRF510 can perform even better than Vishay's version. This particular result will vary between different product batches.

    • IRF510s with International Rectifier branding are most likely fake or "new old stock" if you are very lucky!

      Note: IR and Infineon have stopped manufacturing IRF510 MOSFETs.

    • AliExpress IRF510(s) are likely to be fake

    • The IR IRF510 gave ~3W versus 5.2W+ from SEC IRF510 versus ~5W from Vishay's IRF510!

      Note: The gate threshold voltage can vary even between two samples picked from the same manufacturer!

    • The Cgs ("gate capacitance") of a fake IRF510 is typically more than 1000pF on a LCR-T4 meter. This value is around ~400pF for genuine IRF510 parts on a LCR-T4 meter.

    • Some of the fake IRF510(s) cause a short in the power supply chain resulting in bad things happening (TM).

      I was not aware of this problem a few years ago, and was wondering why the power supply fuse was blowing in a uBITX I was repairing. I didn't have the LCR-T4 meter back then.

    • Ensure proper sourcing (trusted supply chain, etc) to get genuine parts

  • All resistors should be MFR!

  • BS170 MOSFET - 9 to 15 INR

  • 3 x 4.7k Preset (Potentiometer) RM-065 - 20 INR

    • It is easier to get 5k (502) trimpots
  • 1 x 3.3k 0.25W MFR - 2 INR

  • 1 x 4.7k 0.25W MFR - 2 INR

  • 1 x 220 0.25W MFR - 2 INR

  • 1 x 10 Ohm 0.25W MFR - 2 INR

  • [Optional] 2 x 10k Ohm - 0.25W Metal Film Resistor - 5 INR

  • A very large heat sink - 50 to 175 INR - larger heatsink is better!

  • Regulator heat sink - https://etstore.in/index.php/product/aluminium-u-type-heat-sink/ - < 5 INR

  • 1 x LM317 (TO-220 package) Voltage Regulator - 18 INR

  • 1 x 1N4001 (or 1N4007) diode for regulator protection - 2 INR

  • 2 x SMA connectors (Vertical PCB mount) - 42 INR

  • 1 x 10uF 50v Electrolytic Capacitor - 3 INR

  • 4 x 100nF (104) 50v MLCC - 20 INR

  • 1 x 100nF (104) 100v MLCC - 20 INR

  • 1 x FT37-43 (Black Toroid, Fair-Rite 5943000201) - less than 25 INR (handles <= 350mW only)

  • 1 x FT50-43 (Larger Black Toroid, Fair-Rite 5943000301) - less than 32 INR

  • Misc: Relimate Connectors, Glass Epoxy Copper Clad Board (PCB), Various Consumables (~100 INR), SMA cables for testing, RF adapter (SMA Male to SO-239), M3 nuts and screws, TO-220 insulator kit

  • Tools: LCR-T4 Meter (for measuring things), RF Power Meter

  • Optional power supply: LM50-20B24 Mornsun SMPS - 24V 2.2A - 52.8W AC/DC SMPS - 700 INR

  • Optional PPTC fuse

  • PA Total Cost: Less than 600 INR

Power Supply + LPF BOM

See the following projects:

Note: For LPF, use C0G 100v rated caps!

BOM + Equipment Sources

Note for international buyers: The BOM can be carefully sourced via Mouser and DigiKey.

O(Current consumption)

The ESP8266 usually draws a peak current of about 250mA after it wakes up and then more or less constantly draws around 70mA before it goes back to sleep.

DS3231 -> 5mA

Si5351 -> ~50 to 100mA

PA -> Less than 1 A (rms)

Upper bound -> 1.4A maximum.

Tips

MOSFETs are sensitive with regards to ESD and high soldering temperatures. Use best practices when dealing with MOSFETs and perhaps always ;)

TO-220 Socket TO-220 Socket

By using these TO-220 Sockets, we can easily support any MOSFET pinout. This technique allows usage of off-pcb heat sinks with a lot of flexibility. It also avoids common ESD + MOSFET problems as well. The gate bias can be easily checked by safely removing the MOSFET from the TO-220 socket. Also, this socket idea allows us to safely eliminate the LDO from the design.

Note: Keep the length of the TO-220 Sockets as short as possible.

Also, use a SIP socket (aka 40×1 Round Header Female Berg Strip) for the BS170 MOSFET.

A burnt 78(L)05 regulator can easily eat up 4-5W of power by itself. If there is no RF output but power is being consumed then check your linear regulator.

Mods

If you have 1-to-2 W of RF input drive available, then remove the BS170 driver stage altogether. The rest of the circuit works as it is.

Future ideas

  • Use a PPTC fuse in the PA voltage line (oscillations are a real problem, and can be nasty) - Sandeep Lohia (VU3SXT).

    [Alternate] For testing purposes, I am using a current-limited power supply.

  • Add a Type K Thermocouple based temperature monitoring to ensure safety.

  • Alternatively, use ACS712 to monitor the current consumption to ensure safety.

  • Better still, use a power supply (module) with constant current limiting.

  • Add foldback current limiting - see QRP Labs docs.

  • Add negative feedback - see QRP Labs 10W HF Linear Power Amplifier design.

References

See References.md.

Credits

This design uses ideas, and support from Hans Summers (G0UPL, QRP Labs), VU3SXT (Sandeep Lohia), VU2ASH, Mr. Bhatnagar (VU2SPF), AC7LX, and PY2OHH.

Also see https://github.com/kholia/HF-PA-v3#credits and https://github.com/kholia/HF-PA-v2#credits.

About

VU3CER's Robust Class-D HF 'QRP' PA for 40-20-17-15-10 bands. Yes, IRF510 can work on 10m with a minimal design ;) Input is 8-10mW.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages