George Harker

Building a DIY signal injection transformer

Picture of georgeharker

To test switched mode power supply designs it is sometimes necessary to inject signal disturbances into the feedback loop to classify behavior. To do this you need a signal generator and oscilloscope or a network analyzer (expensive). You also need a way to decouple the signal generator from the circuit under test. One technique is to use a signal injection transformer. These too are unfortunately pretty expensive.

Following the guide here and eevblog posts here it is possible to make an injection transformer by hand.

I used the following parts:

The cores were taped together to make handling easier.

A long pair of wires (about 60x the height of the core stack were cut). These were twisted together and wrapped onto the coil (about 25 windings round the core which is about as much as can fit. Others note a different core which allows for better spacing- though more windings is unlikely helpful).

Care was taken to note the matching ends of the winding before pairing each loop back to the coax connector (so that the phase / connection to the signal pin is same orientation).

Testing was done using a siglent SSA1104x-e as the oscilloscope, and a siglent SDG2042x as the waveform generator. Connections to the oscilloscope were BNC terminated with 50ohm through terminators.

A bode plot of 10Hz to 10MHz was taken.

The results are -3dB at 28.6Hz rising to flat at approx 110kHz and remaining flat to just under 1MHz before rolling off smoothly to -3dB at 5.6MHz.

It would be nice if the HF end was a little higher gain further out. But this will serve my purposes pretty well. And it’s a lot cheaper than $560 for a picotest.

Note ideally at least device under test end should be fused. I didn’t. I won’t be using this on particularly high voltage / current. But… I should have.

Equipment

Picture of georgeharker

Soldering

Soldering Iron:

Hot Air Rework:

Reflow:

Tips:

Sleeves:

Solder:

Paste / Flux Dispenser

Flux:

Dispenser needles:

Desolder wick:

Tip cleaner:

Work holding:

Fume extraction:

Soldering mats:

Tools

Tweezers:

Snips:

Wire cutters:

Crimps:

Screwdrivers:

Knife

Files:

Calipers:

Lab Equipment

Oscilloscope:

FTDI and programmers:

Power supply:

Signal Generator:

Electronic load

Multimeter:

Optical

Wearable Magnifiers:

Microscope:

Wire and Components

Hookup / rework wire:

Cable construction:

Fixed hookup:

Caps:

Resistors:

Disposables

Gloves:

Painters tape for holding solder paste stencils etc:

Kapton tape:

Isopropyl Alcohol:

Storage:

Getting a T962A up and running

Picture of georgeharker

I strongly recommend reading 1bitsquared’s blog post and This hackaday post before deciding to buy a T962A oven.

The short version is that getting these to work reliably is a /lot/ of work.

Issues Include:

  • Poor temperature profile following
  • Uneven heating causing scorching in places even whilst other areas do not reflow
  • Lack of internal convection leading to uneven heating
  • Unshielded infra red bars causing hot spots
  • Poor default firmware
  • Masking tape internally that will stink when it burns
  • Lack of grounding

However good reflow ovens are expensive. And with some work you can get these functional – even for relatively large boards. This makes them a reasonable – albeit time consuming way to get a reflow oven at affordable cost.

Several other sites (listed above) have better guides on what first steps to take.

Required mods:

  • Fix the grounding – do this first
  • Fix the masking tape and replace with Kapton – also do this before plugging in or turning on the oven
  • Flash the firmware to unified engineering’s firmware: UnifiedEngineering T962 firmware

Additional fixes:

  • Add better thermocouple interface – I used this one https://github.com/UnifiedEngineering/T-962-improvements/wiki/Better-thermocouple-interface and followed the instructions there for wiring it up
    • More thermocouples allows us to average more places in the oven. To do this I hacked the firmware to allow calibration of pairs of sensors and average them all – see my GitHub fork for T962A firmware
    • Allow thermocouples to be nearer to the PCB – this is a tradeoff. You can mount directly on the board and deal with thermocouple wires and fixing the TC to the board all the time, or you can install additional thermocouples in tubes near the board. I did the latter using threaded metal tubes usually used for lighting fixtures
  • Add additional thermocouples:
    • More thermocouples allows us to average more places in the oven. To do this I hacked the firmware to allow calibration of pairs of sensors and average them all – see my GitHub fork for T962A firmware
    • Allow thermocouples to be nearer to the PCB – this is a tradeoff. You can mount directly on the board and deal with thermocouple wires and fixing the TC to the board all the time, or you can install additional thermocouples in tubes near the board. I did the latter using threaded metal tubes usually used for lighting fixtures
    • I also put them at slightly different heights to allow for better averaging – the existing TCs are near the top of the oven and will read hotter than the board
  • Seal the front of the oven with gas grill seals so that airflow from the front doesn’t drop the temperature at the front of the oven.
  • Measure and calibrate your thermocouples against various locations on the PCB. I use this sheet here which allows calculation of calibration mapping from the MAX TC probes to probes on a PCB which are what we ideally want. I measured several locations, and select two curves for calibration which get entered into the firmware

  • Develop and tweak your own paste profiles using the same link you can develop a curve based on manufacturer instructions for the paste.
  • I can now successfully reflow boards of about 8in x 8in in solder free paste which was certainly not possible before