Another kit: All-discrete power supply

[c4757p]

Yes, it will be, though I can show what I did briefly here, in pictures at least.

Line regulation: Truly clean measurements, for a smooth plot, were difficult to pull out of the random noise of thermal changes in the environment. Shielding from air movement helped, but I still had to average over a long enough time that I wanted to prevent ambient temperature change from causing error. The board was placed into a controlled chamber (actually a PID+toaster reflow oven packed with ceramic tiles) and stabilized at 60°C. A very stable power supply (Power Designs 2005) was used to offset the -12V output to allow more measurement precision, and I used my Fluke 45 to output data to my PC, where it was low pass filtered to get clean measurements. I stepped over a range of input voltages, then recorded the output delta after the LPF settled.

Load regulation: The main reservoir capacitor was removed to allow AC on the input side. I put my main power supply in series with a function generator, to inject ripple at various frequencies. I kept the input amplitude to 1Vp-p, and measured both the input-side and output-side ripple on an oscilloscope to calculate the rejection. Sensitive, differential amplifiers were necessary, at least at low frequencies where the rejection was very high.

I kept the board in the "thermal chamber", now fully cooled back to ambient, for this test, as the metal case would shield against interference.

Note that the ripple rejection is actually better under 100 Hz than the DC line regulation. This is possible because a large component of the line regulation error is due to self heating, and so has a long (> 1 minute) time constant.

[Thor-Arne]

Nice, thanks. 

[szabolcsx]

Hi, I really appreciate your effort to design a power supply using only discrete components. I'm a beginner and there are some "dark spots" for me in schematics. May I ask when the documentation will be finished? The concrete part of schematics which isn't clear for me is the current sense amplifier. I tried to simulate using LTSpice but with no luck. I don't understand why it's connected to +2V.

[c4757p]

Hi, szabolcsx. Yeah, as you can probably see from this thread, electronics has been moving slowly for me lately. The documentation's a ways off as is just about everything else.

Okay, the current sense amplifier. First, the simple part: why is it connected to +2V? Because it needs "headroom" above the 0V positive output line. In order for it to work, Q23A has to pull its collector at least 1.3V (two transistor thresholds) down, so that both Q24 and Q19 turn on. If it were connected to 0V instead, the collector would drop below the base, forward-biasing the base-collector junction and preventing the transistor from turning on any more.

Now, for the actual operation: R47 is in the path of the output current, and acts as a sense resistor. The left side of that resistor sits at 0V, and the right side drops down by Ohm's law as the current through it increases.

A voltage divider between 0V (the left side of R47) and -12V allows you to set a threshold voltage, which is
connected to Q23A's base. The actual current sense voltage goes to Q23B's base. The emitters, which are joined together, will go to whichever voltage is lower, and then R44 acts as a current source to pass about 1mA through the two transistors.

Normally, the current sense voltage (on Q23B) will be higher. Q23B is then the only transistor whose base is above the base-emitter threshold, so only Q23B is active. When the current sense voltage falls to the threshold voltage, though, both transistors have equal base voltages, and they both turn on. If it falls even more, only Q23A will remain on.

If the balance shifts enough towards Q23A that there is about 650µA flowing through it, each of R42 and R43 will produce 0.65V, about a transistor threshold, turning on both Q24 and Q19. Q19 sources current to LED DS2, activating the 'limit' LED. Q24 pushes the FB+ line higher, which makes the main error amplifier think the output is too large. The output is then decreased until Q23A and Q23B are balanced.

Let me know if I haven't explained something well, as I don't know what point you're at in your learning.

Would you care to share your LTspice file with me?

[Kevin.D]

Hi there . Where are your latest LTspice simulation files  ,the ones  posted back on page 2 of this thread dont match with your later schematics .
Thanx

[c4757p]

Try the ones in Github a few posts up. But I haven't really been using SPICE extensively for this project, so I'm not sure how current they are.

[Kevin.D]

Latest git hub (may) ones Doesn't work , some .libs and symbols missing.

[szabolcsx]

Sorry for the late reply, I had exams. Here is the LTSpice file. I wanted to measure the current on high side.

[c4757p]

I thought you were trying to simulate the amplifier in this kit! I'm sorry, I've got too much going on right now to walk you through designing a different one. Perhaps you could try posting a new thread?

[c4757p]

Everyone - I know this project has been neglected - I came off a hell of a semester into a hell of a summer, and will probably have another hell of a year to follow it.

I am not abandoning this project, but as so many other things have been demanding my attention lately, I cannot promise anything. Frankly, it's been so long and I've had so much on my mind lately as to pretty much lose interest - though I still like this project and will still complete it eventually. Don't be surprised if it takes another year....... That's what happens when you start a complicated project and assume the rest of your life won't get in the way.

I've got another couple of things slowly coming along as well, which I'll post later in development rather than posting a long string of "I promise you guys, it's almost done again!". Wouldn't be surprised if school and life took precedence again.

[Jebnor]

c4757p, Thanks for all your good work.  I have learnt a lot reading this thread.  No need to apologize for life, we understand.

When you have the time and inclination to get back at this project, I'd like to see it completed.  No rush.