More problems… more solutions…
So, I was just going over the main power supply board, looking to solve something that on paper should work, but in reality is less than ideal. The output of my monster transformer is somewhere in the neighbourhood of 42.6V with no load. That’s a tad high considering the maximum voltage differential of the LM317 is 40V and the current output seriously suffers when it’s that high. I had tried breadboarding an earlier version of the circuit to get a total fail, the regulator refused to even bother to regulate. So, I figured an easy way to drop some voltage would be to use a bunch of series power diodes to drop 0.7V each, so I inline-soldered 6 of them to drop 4.2V (when under load) down to 38.4V. Looks nice and bodgy, i’ll post a photo at some point, looks laughable.
Though this approach sort of works, it is far from elegant, ugly, and I’m not entirely sure reliable. 38.4V vs. 42.6V is not that great a difference and still dangerously close to that maximum voltage differential. Not only would I like more padding for safety, but for more reliable operation that dissipates less heat.
The general calculation for linear regulators and power dissipation is given by: ( VIN – VOUT ) * IOUT. Even with passing most of my current via a pass transistor, the LM317 will dissipate an unacceptable amount of power. I read somewhere that the absolute maximum dissipation is on the order of 15W and at full load and minimum voltage, I calculate that it will dissipate 9.3W. Though this is under the limit, the wide differential voltage could potentially cause problems. Also, I just don’t like it. Yes, it’s my first power supply – but I want it to be a good one I will want to use for years.
Zeners to the rescue
A little googling around revealed that I could use a transistor, Zener diode and resistor to act as a voltage regulator. In this case it would be a pre-regulator and would drop the voltage down to about 30V which would solve the above issue. I found this circuit from Elliot Sound Products, a site I’ve been liking for several years now as most of his projects are for audio. The entire site is full of not only useful circuits but in-depth explanations, I highly recommend it. Turns out this technique is used all the time in audio amplifier applications for this exact purpose.
The mechanism of action is quite simple – use a zener to bias the transistor and the voltage drops by the zener voltage plus 0.6V. So a 12V zener would knock down my voltage to pretty much exactly 30V. Nice and tidy
I looked back at my schematic and didn’t like the idea that I would have to add yet another transistor pair and more circuitry. It then occurred to me that I’m already kind of doing the same thing with my current limiting darlington transistor which drops the voltage in response to hitting the set current limit. If I bias this transistor properly, I can essentially achieve the same effect without adding in another transistor and keep my circuit mostly as it was. Brilliant. Of course, I’ll need to simulate and then do a test build to make sure it won’t catch fire.
By my (inexperienced) reasoning, I believe I can combine both functions by using a 12V zener to bias the transistor from the positive rail, dropping it down to 30V by default, and keep my other transistor which will open to ground if the current limit is reached. Hopefully, this will still drop the base, when the current limit is exceeded, to 0V and thus zero current, shutting it down.
I have to point out here – I really have very little clue what the hell I’m doing, so if you try this I’m not in any way responsible for what may happen to you. Do your research and use caution. I’m learning out loud. If you have any comments on the above please post them, I can use all the help I can get. For myself, I’m not trying this until I simulate it first.
I’ll post schematics and simulation analysis when I have it.
Opamps and something I didn’t think of…
This process also brought to light something I completely forgot about. Using simulation, the opamps were omitting their power rails as is often done on schematics assuming rail to rail operation. In reality, these opamps would have a supply voltage, and the output of said opamps is limited by this. Under ideal conditions, the opamp would be rail to rail and be able to handle the full 80 odd volts (since I have to power them before all the regulation happens to ensure operation). I’ve never seen an opamp that can handle 80V rail to rail, though it wouldn’t surprise me if these exist … and are expensive as hell.
Really, I have no need of such precision and such a wide voltage swing in my outputs. The opamps are used to manage the set voltages for the current limit and voltage regulation. Using the idea from Dave Jone’s excellent power supply videos, I use a 1Ω current shunt, and a bunch of opamps to compare a measured current to a set voltage and keep the two in line allowing me to set the maximum current allowable. For the voltage setting, the opamps are used to buffer, split and invert the signal so that one voltage setting can be used for both the positive and negative regulators so they track each other and avoid the messy issue of using dual-gang pots (also, I’ve never seen a 10-turn dual gang, again, wouldn’t surprise me if they exist but man that would be expensive).
Last night, before I slept, I chewed it over a little and figured out how I can solve this supply voltage issue. The requirements are that I need to have a 3V voltage swing for the current limiters and a 24V voltage swing for the voltage setting. The former is quite easy – I can use just a couple of standard LM3900 quad opamps since they can easy output a 5V swing. If I supply them using say ±5-12V that would provide an adequate swing for both positive and negative current limiting (whose desired values are 0-3V for the positive, -3-0V for the negative) and prevent going anywhere near either rail since these are not rail-rail opamps. I can use two LM3900s (one for positive, one for negative) and supply them both with at least ±5V using zener diodes. Lovely solution!
The voltage setting is a bit more difficult. In shopping about I had a hard time finding opamps with a supply voltage greater than 44V rail-rail, which is just shy of the 48V r-r I would need to set the voltages from 0-24V (and invert to negative for the negative regulator). Of course, they beauty of op-amps is that they are analog calculation machines, in essence.
The circuit as I have it uses 4 opamps which act as buffers and one inverting amplifier. It would be a simple matter to power the positive ones from 0-24V (the LM3900 has, I believe a ±16V supply limit, giving me 36V if I were to single-supply power them) and the negative ones from -24-0V. The one inverting amplifier would, according to my reasoning, have to straddle the negative and positive voltage rails in order to invert the signal. Here’s the neat bit: to overcome having to convert a maximum of 24V to -24V (and use the associated > 48V power supply) I can simply convert the buffers leading to the inverter into variable gain amplifiers to divide the voltage in half, then invert it, then multiply it by two. This way the inverting amplifier can run on half it’s supply voltage (24V, -12V – +12V) and stay well within it’s limits. I’m not sure how much precision I would lose in this conversion but it would afford me greater control of the symmetry between the rails as well if I add in a couple of trim pots on the gain of the divide-by-2 and multiply-by-2 amps. The supply voltage for the all the amps can be easily accomplished using zener diodes in various configurations.
I know some of you are screaming at my already to use a bloody microcontroller and stop fiddling about with messy opamps. Really – I want to do it this way. I feel that in my self-education, I want to use opamps effectively and get a good understanding of analog electronics before pushing on to digital. This is why I am trying to keep the power supply entirely analog, which I have done so far apart from the meters.