PSU design that I've been working on


Thread Starter


Hello, I am getting started with digital electronics, and I've decided to design a PSU, which is a common project for beginners. However, I am no like most other people. So I, naturally, decided to take on a much to hard of a project and get myself in way too deep -- kind of.

Okay, okay, okay, ill stop being so cryptic. I decided to design a lab power supply with rather difficult parameters that I'll list at the bottom. I have a decent amount of experience making logic-level and control circuitry (ie. an improved micro-controller makerboard, a sensor, security cameras and their controllers -- over bluetooth, etc.). However, I have only done a small amount of practical power electronics designs.

I can include the schematics -- matrices, models, diagrams, and sketches -- that I have made so far if you wish, but currently, my question is this:
The PSU has 2-channel output, should I isolate the two at the AC-DC transformer, or should I isolate the two with a DC-DC converter after it's rectified and filtered?

I have started to come up with schematics for both if anyone finds my description confusing. Also, two last things. Would it matter -- since I am designing a PSU with floating outputs -- to try and keep the neutral reference close to mains ground or if that is irrelevant? And does anyone know an alternative to the LT3080 that would prevent me from needing to chain them in parallel or is that the best solution? I believe I'm asking the right questions, but I would be very grateful.

Planned optimal Specifications:

- Controlled by a uC and optical rotary-encoders
- Be able to save at least three pre-sets
- Independent floating dual outputs with a shared mains earth grounding point
- Constant current & voltage limit adjust (mode controlled by switch and uC)
- Full range (able to reach both 0 volts and 0 Amperes)
- Be able to reach equivalent negative voltages with the rotary encoder by switching a relay after 0 volts and following an absolute-value curve
- Be able to reach a full range of negative voltages without having to re-connect the load
- Over and under current and voltage protection
- Only one Main -> PSU connection
- Minimum noise and high accuracy (linear)
Secondary (flexible specifications if difficult to reach):
- Maintain all systems -- including the PSU-Main -- isolated
- EMI filters, DC-DC converters, etc.
- Include readouts for both current load and limit setting (either current or voltage for both outputs)
- The current (not amperage current, but the time-one) limiting mode will be displayed on the limiting readout
- If convenient, prefer to use a DAC for the control from the uC to the CV/CI circuitry
Tertiary (conceptual ideas to be explored in later versions):
- Have an expansion "slot" where a card with three more LT3080s could be connecting, increasing it from 2 in parallel to 5 in parallel
- This way, with the two outputs in series, the output would be 40 volts 10 amperes max
- Include USB monitoring and control
- Able to monitor noise, percent error, analyze load and line regulation, and maybe monitor temperatures
- Be able to replace its CV/CI PCB with a computer-controlled function generator PCB interchangeably (or as an addition)
- I know this one is ambitious, that's why it's in the tertiary section, it's an idea I'm looking at)
- Possibly include K-Type thermocouples for temperature measurements (over USB) and emergency shut-off
- Use all incremental controls, so that when it is turned on, all the values are reset to zero and modes are reset to the current default profile
- All controls should be interfaced via USB-C (either built-in support with the uC or a USB to UART)
- The rotary-encoders should have a 3-throw switch, one for "fine", one for "course", and one for "disconnected" adjustments

PS. I wrote most of this longer than a week ago and I very possibly left quite a bit out by accident. I was taken away because of school, however, I am back now and hoping to get back to work. I have all my notes still, so reply to this or message me.

My first thought is that if you require minimum noise then you should not include any switch mode parts.
I have already made the choice to use only linear regulation (despite the reduced efficiency), and to use bypass caps on both ends of the transformer.