Power supply simulations and reality don’t match!

[Kashmire]

I upgraded my Zen v4 power supply with a big 625VA Avel-Lindberg from PartsExpress. I choke-loaded the power supply using a 15-gauge 3mH Jantzen air-core inductor. The choke then feeds into the filter capacitor bank. I choke-load the transformer to lower the peak currents, improve power factor, and reduce the amount of noise the filter capacitors . By doing so, I have to choose the transformer voltage very carefully.

I simulated the circuit, using estimated DCR of the transformer and published specs of the choke. Based on the simulations, I chose a 45V RMS transformer and 3mH of inductance. This system should output 48VDC under 2A load.

When the parts arrived, the power supply output 54.5VDC under 2A load. I had to increase inductance to 6mH to reduce the voltage to 48.5VDC. This in itself isn’t bad; more inductance means lower peak currents and better power factor. What’s bad is the simulation didn’t match.

I went back to the simulation using measured values for the inductor and transformer parameters. The simulation predicts my configuration will output 42.6VDC, while the real thing is at 48.5VDC.

Any suggestions on how to get the simulation tuned?

Here’s the parameters:

Avel Transformer: No load 69.9V p-p (49.5V RMS), 0.25 Ohms copper loss
Jantzen choke (first try): 3.0mH / 0.64 ohms DCR
Jantzen choke (second try): 6.0mH / 0.91 Ohms DCR
Filter Capacitor: 2 x 14000uF, 0.1 Ohms ESR
Diode bridge: used 1N5401 diode model in SPICE

The diode bridge has a 0.47uF soldered directly across the DC terminals to absorb the commutation noise of the diodes. I’ve also included it in the simulation, but it has exceedingly little effect in reality or simulation.

[GRollins]

Simulations and reality don't match? Golly, what a surprise!
I'd suggest going with reality. As long as you get the rail voltage and current capability you want, I wouldn't bust a gut wondering why the software let you down. And yes, all things considered, the more inductance the better, so it looks like you're in good shape.
The time you're spending fretting over a silly piece of software would be much more profitably spent with a soldering iron. Your amp awaits...

Grey

[Kashmire]

I didn't use a transformer model. I used an AC voltage source with a series resistor. The voltage source simulates the no-load voltage of the secondary. The series resistor simulates the copper and core losses.

I measured these values by measuring the output voltage of the transformer at no load and with two different resistive loads. I could calculate the effective series resistance (ESR) of the transformer using two of the points, and use the third measurement to verify.

Even if my measurements are wrong, my ESR value is close. Even if I double or half the value in the simulation, it doesn't meaningfully change the results.

[rebojorge]

What I really want to know is..... How is the performance now with the upgrade?

[azira]

Instead of "simulations are crap" I think "the simulator is probably not modeling a part correctly" is a lot more productive and probably a more accurate assesment of Kash's situation.

The first thing I'd consider probably is if the air-core inductor. I am not an experct by anymeans on inductors but I know that iron-core and aircore are somewhat different.

The other thing I'd think about is your voltage source modeling. Using the series resistor is a good start but you're going to have some source and load reflected impedances as well that might change things too.

Sorry, my text books are packed in a box right now or I'd help you look it up.

I think seeing actual results follow from simulations and math is a great way to go. Good luck, keep us informed.
--
Danny

[classd4sure]

You're results aren't all that far off.

Modelling the transformer as a pure voltage source with a series resistance, kind of idealistic no?

It would be a project unto itself just to model the transformer realistically, and another to do the same for the rest of the circuit.

How about stray/parasitic capacitances of the part? Also stray /parasitic capacitances/inductances for every connection and length of wire? Including every non ideality of every other component?

That would all be required to achieve the most realistic simulation, is that amount of effort really worth it? Might be good experience, but since your circuit works I think it isn't.

I think what you've done is good, proven the thoery with simulation, built it, and it works, so it's a little off, no big deal it's expected. Now you can use real world test equipment to optimise it to your liking.

Incidently, I didn't see you mention exactly what simulator you are using? One simulator to the next will have differing results, some are pretty accurate and others not.

[Kashmire]

Hey, thnx for all the posts.

I think my modeling results are poor. My modeling results were 7 volts from my target specification of 48V, which is more than 10% error.

I’ve applied some sophisticated models to the transformer with no meaningful results. I am going to make more measurements to validate each component of the model. I will post my results.

An air core inductor works more like an “ideal” inductor, because an air-core inductor won’t saturate. A ferrite-core inductor has hysteresis and saturation, which can be difficult to model.

I think my transformer model is good – an AC voltage source representing the open-load voltage and a series resistance. It should get me a lot closer than the >10% error I’m experiencing.

I agree that we should encourage comparing simulations with real circuits on a regular basis. Too many of us (me included) spend time on a simulator without doing the real thing. I like to see math and reality converge …

I am using B2Spice. The evaluation works as good as the real thing, as long as you don’t use too many parts.

How does it sound? GREAT! My previous power supply was undersized – it was only 38V at the drain of the current source transistor. Having 44V at the current source changed things dramatically.

My previous power supply had 120mH of inductance (choke-regulated), which made it continuous conduction. I had to do this, because the transformer was so tiny, it couldn’t handle any peak currents. My new power supply uses 6mH choke-regulation on a 45V (rms) transformer, giving 48VDC at 2A. The peak currents in the transformer are kept below 10A. I use a 14,000uF 75V capacitor. I chose 75V so it the system goes out of choke-regulation, the capacitor won’t blow up (WARNING! When using choke-regulation, if the current drops, the voltage will rise to sqrt(2) * 45V = 64V).

Since I’ve got these 120mH chokes sitting around, I made a pi filter using another 14,000uF capacitor. The AC ripple dropped below three decimal points on my precision AC meter!