How the uBITx works: A technical summary

uBITx 45MHz roofing filter (First IF stage)

Here’s a pretty complete summary of what’s going on within the uBitx, under its dual conversion IF system, as explained by Jerry KE7ER.

Actual frequencies are those used in the supplier’s original µBitx code:

Release version dated Dec 6, 2017 at:

uBITx 12 MHz Filter – 2nd IF stage

Local Oscillators

  • CLK0 (BFO) is fixed at 11996500 Hz, maybe 500Hz below the 12MHz filter’s 2000 Hz wide 3dB passband.
  • CLK1 (second local oscillator) is fixed at 56995000 hz for USB and 32995000 hz for LSB. You can find those three numbers in the file ubitx_20.ino at lines 166, 163, 164 respectively.


The VFO is used to select the operating frequency Fop according to these formulas.

For USB:

Fop  = VFO – (CLK1-BFO)   so:

VFO = Fop + (CLK1-BFO)   where CLK1 is around 45 MHz + 12 MHz

For LSB:

Fop  = VFO – (CLK1+BFO)     so:

VFO = Fop + (CLK1+BFO)      where CLK1 is around 45 MHz – 12 MHz

To receive a 7.2MHz LSB signal (where 7.2 MHz is the frequency of the suppressed carrier), the VFO gets set to:

7200000+(32995000+11996500) = 52191500 Hz.

That formula gives an exact result, not an approximation.


The BFO corresponds to the carrier frequency of the station being received or transmitted. The 12mhz filter is always above the BFO, so within the 12 MHz IF it allows through only the upper sideband.

The VFO is always above the 45 MHz first intermediate frequency, and so always inverts the sidebands.

A carrier at 7200000 Hz would get translated to:

VFO-Fop = 52191500 – 7200000 = 44991500 Hz

A lower sideband at 7198500 Hz would get translated to:

52191500 – 7198500 = 44993000 Hz

In this example we assume the lower sideband is generated from a single audio tone into the mic of 1500 Hz.   We have chosen 1500 hz because it will land in the middle of the 12 MHz filter’s passband, assuming the filter has a 3 dB passband that’s 2000 Hz wide and the BFO is 500 Hz below that passband. The actual range of frequencies passed will be 500 to 2500 Hz. Those assumptions of 2000 Hz and 500Hz might be off by a couple hundred Hz.

Likewise, a high side CLK1 of 56995000 Hz  for USB always flips the sidebands when translating to 12 MHz, however the low side CLK1 of 32995000 Hz we use to receive the 7.2 MHz LSB signal does not.

Our 7.2mhz carrier:

44991500 – 32995000 =  11996500 Hz (exactly equal to our BFO frequency)

Our 7.2mhz lower sideband:

44993000 – 32995000 = 11998000 Hz  (in the middle of the crystal filter passband).

That’s how the original uBitx code works.

There is a problem because where an LSB signal hits the 45MHz filter will be 4khz removed from where a USB signal hits it, resulting in different audio quality between the two.


VU2ZAP experiments with TDA2822 Audio Chip

Raj VU2ZAP has done some experimenting with TDA2822 chip failures.

FCI chips were sourced locally and plugged in. No failure was observed when using the chip with normal volume.  However, when he raised the volume to a high level the chip failed immediately!


1: Add a resistor in series with the 12V line maybe 10-15 ohms to drop the voltage.

2. Simple solution is to use only a 16 Ohm speaker or add a series 8-10 Ohm resistor with the 8 ohm speakers.

3. Cut the power track to the TDA and insert a 78L08 or 78L09 to reduce the voltage. There is a convenient track which can be cut.  Solder either an SMD or DIL regulator onto the board at this point. This track leads directly to the chip and filter cap (see photo below).

The center lead looks soldered, but is in fact floating.  However, the tab is soldered to the copper groundplane below. The white wire is part of Raj’s fix to prevent the filter relays from clicking during PTT.  Raj suggests the board as illustrated now needs a brush cleaning!

Raj observes that with this mod, audio is fine but distorts at high volume. The 9V voltage out does not drop at maximum volume.

Feedback on the solution from Jerry KE7ER

Jerry KE7ER says “This looks like a fine solution!   The fact that the 78L09 inherently limits current to something reasonable is a bonus.  And it fits in there perfectly!

Jerry suggest that it might be a little bit easier to just rip that trace out entirely. You could place the part up against the through-hole on one side and run a short wire to the other through-hole.

Gain should remain as it was at 12v.   Just that when you turn the volume up it distorts earlier.

If what you have on hand is the TO220 LM7809 (or any voltage from 5v to 9v), it will fit in there nicely as well.  Glue it face down to the board with pins hovering over  the trace that Raj has cut.   Or, if you don’t like glue, solder a short wire from tab to ground plane.

ST makes the 78L** parts in a SOT-89 package.   This is the same package as U2, the 78L05 for the IRF510 gate bias.

TI and Fairchild LM78L** parts only come in the SOIC-8 and TO-90 (also some dinky BGA that we don’t want to talk about).  Digikey and Mouser don’t sell ST’s SOT-89, so a bit harder to obtain. The SOIC-8 should do fine, solder the ground pins to the ground plane for heat transfer.   The TO92 could work, although it would likely heat up faster because the heat has nowhere to go.

The maximum available current draw from a 78L09 would limit the maximum audio level.    Apparently this provides enough audio for Raj, and limiting that current does protect the TDA2822.   If it gets to be too much current for the 78L09, it will shut down till it cools off.   Just what we want, if it’s enough audio power for your situation.

If you want more power and less protection, use the 1 Amp TO220 LM7809 (or 08 or 06 or 05) face down, flat on the board.   ST sells the 1 Amp L78** in the Dpak/TO252 package, about the same as the TO220.  TI sells the 500ma LM78M** in the Dpak/TO252 package, might be an ideal solution for moderate levels of current.

Any of the above could be made to work, and in fact any regulator providing between 5 and 9 volts.   This assumes you don’t have one of the clones made from factory floor sweepings that smokes at 7v.