A bit of context
If you are new here, I recommend reading the first part of this series. There I discussed the various design constraints and considerations and weighed the pros and cons of each one.
In this post I will discuss the first prototype I designed.
First prototype
For the first prototype, I wanted to test what in Part 1 I called Option #1, more specifically the Solid State Relay (SSR) output stage. The rest of the prototype was simplified a bit to allow me to test the prototype a bit more easily, ending up in a sort of fusion between Option #1 and Option #2.
I won’t show the full schematic because I ended up going for a completely different design (post coming soon) but I will try to explain what I learned during the process.
Power conversion
Although this block can be solved by throwing components at it, I wanted to keep the BOM cost reasonably low. This configuration features few parts, only a switching converter for the 3.3V rail and a negative linear regulator for the op-amp.
Although this circuit works well, I soon realized that it being limited to AC input (for which I do not have a protected power supply) makes prototyping much more stressful. An unfortunate truth about one-off projects is that development costs are never recovered, so in that sense sometimes the technically optimal solution is not the best choice.
Thermocouple amplifier
Instead of implementing a discrete instrumentation amplifier, I chose to use a single AD8293G160 part. Although the gain is a bit low, it requires no external precision parts (as it has a fixed gain) and is overall a very integrated component.
However, I can’t attest to the performance of this solution because, as will be discussed later, this prototype had a major flaw that prevented further development.
Power switch
The power switch was to be one of the main advantages of this design. This isolated Solid State Relay (SSR) only requires two back-to-back n-channel MOSFETs and an admittedly fairly uncommon photocell optocoupler. The MOSFETs can be selected for very low RDS(on), meaning superb performance with a very low BOM cost.
Now, at this point I should probably mention that when I designed this (almost three years ago) I had never had any formal electronics training at all; all I knew was random bits of lore passively obtained through watching EEVBlog and Marco Reps videos. Needless to say, I only calculated the most necessary resistor values and eyeballed the rest, so I hope the reader won’t judge me too hard for the upcoming amateur mistake.
Where it went wrong: MOSFET input capacitance
Any engineer with half a brain cell was probably frowning at my optimistic description about this circuit. The truth is that I had stolen it from here and I didn’t think much about it, even though because of the semiconductor shortage in 2021 and 2022 I had to swap the two components of the design.
One of the limitations of MOSFETs is the parasitic capacitance present between their pins. In the case of the the gate, if the driving source has a high series impedance, the created transient can significantly limit the switching speed and efficiency of the switch (more time in the linear region, more disspated power).
For this reason, high-frequency applications always use additional driving circuits to quickly turn the MOSFET on or off. The optocoupler specifies a typical open circuit voltage of 8.7V and a short circuit current of 14uA, meaning that its series resistance would be:
$$ R_{s} \approx \frac{8.7}{14 \times 10^{-6}} \approx 621\ k\Omega $$
As for the MOSFET, the specified CISS (the sum of CGD and CGS) is 4556 pF, meaning that the transient created would have a time constant of:
$$ \tau = RC = 621 \times 10^3 \times 4556 \times 10^{-12} \approx 3 \ m s $$
Using the standard criterion that stability is achieved after t = 4τ, then a full switch would take:
$$ t = 3 \times 4 = 12 \ m s $$
This is way too much for this design, as it doesn’t allow for zero-cross switching at 50 Hz and is too slow for a closed loop system.
Conclusions
I’m pretty sure this designed would have worked with carefully selected parts or a more complex driving circuit. However, many of the parts I wanted were either out of stock (remember, it’s 2021) and/or only available in packages with hidden pads. As you may have guessed from the fact that that I’m building a soldering station, I don’t have a fancy hot air tool.
Finally, I decided to abandon this concept and try Option #2 (as discussed in part 1).