Blocking stage: temperature issues
The generation of the Blocking is achieved, but the rise and fall times are not as sharp as expected. That was due to the decrease of conductance in the driver stage because of a lower working temperature.
Why this was not detected before?*
Several parameters take part in correctly building the pulse up. Primary
inductance and its parasitics are big contributors. Nevertheless, there
are others like a fast driver stage. To achieve this, a fast-charge of
the gate of the MOS connected to the transformer's primary inductor is
required. Improving the charge network in the gate, either by raising
the source voltage or lowering the serial resistance to the gate.
So as to reduce the BOM, it was decided to remove the original 12V source and leave it with a 3V3 one, because BASH103 is glue-logic. Then, the gate resistance was lowered and a BAS16 diode was added in parallel to one of the two serial gate resistors. The task of this diode was to speed up the rise time due to its lower dynamic resistance compared to the resistor it has in parallel, specially at the beginning of the switching. The inadvertent "problem" was in that diode.
While profiling the board, soldering and desoldering was common and big copper traces were present. Then, a higher temperature in that diode was present, too. This higher temperature produced a lower dynamic resistance in the switching stage giving faster rise times than the ones when the temperature is lower.
CONCLUSION*
If possible, the diode should not be placed. A simple profile with
resistors should be more temperature-independent. It should be kept in
mind that the values of the resistors should not be too low to avoid
ringing. A reasonable solution if faster switching times are required
(which IS NOT the case of this board), would be increasing the source
voltage to be applied to the charge network. Then, gate resistance range
is improved.
It would be good*, however, using a supply of 5V in IC3 just in
case a replacement of the transformers is having a poorer parasitic
inductance than the Coilcraft 1:1 coupled inductors used in this design.
Then, the slower rising time due to that eventual parasitic could be
compensated with a proper value of the gate resistance.
CHANGES**
Referring to the sheet 12, page 17 of EDA-02446-V1-0, the components to
be changed are:
> D69
> It should be reversed to improve the rise time and not the fall
time.
> It should be not mounted by default.
> Due to the very low voltage that IC3 is set to,there's no excessive
margin for BSH103 to rapidly conduct. That's why it is important to use
a Schottky diode, instead. 15MQ040 is recommended (low forward
voltage and good performance against temperature).
> R171 R172
> Should be both changed to 47R.
Having slower rise edges is THAT important?*
Considering that Blocking is intended for distances of hundreds of
meters and that the frequencies of rise and fall will be attenuated, it
is not that important. By making the switching slower, reduction of the
EMI is achieved.