The cross-regulation mechanism in a multi-output flyback converter is extraordinary difficult to model and predict. The transformer multiple leakage inductances play a role in the way the primary current splits between the windings when the main switch opens. The clamp level on the drain also plays a role as it sets the speed at which the primary leakage inductance gets reset.
You will find a thorough treatment of the subject in Modeling of Cross-Regulation in Multiple-Output Flyback Converters, published during APEC 1999 by the CoPEC professors, B. Erickson and D. Maksimovic (accent on the c). Another good source is the seminar held by these gentlemen during the conference, same title. They describe the cantilever model, which includes all the inductances found in a multi-output transformer:
Needless to point out that extracting these leakage inductances is an extremely tedious task. But once done, simulations really match with what could be measured on the prototype in the end. Without these parasitic inductances properly modeled, it is illusory to try to predict cross-regulation via simulation.
Nevertheless, I have used my auto-toggling current-mode model to build an averaged two-output flyback converter featuring weighted regulated outputs. It means that there are now two resistances sensing the 5 V and 12 V outputs, and a weight is assigned, depending on the wanted precision. Here, I have assigned 70% to the 5 V and 30% to the 12 V. Diodes drops are important there for the bias point. Leakage terms are not included and have, at a first-order level, no impact on the small-signal response:
Once compensated, the crossover frequency is 2 kHz, with a good phase margin:
I can now use this compensator to model a cycle-by-cycle circuit and look at the regulated outputs, especially when one of them is subjected to a load step:
In this picture, you can clearly see the two resistances separately sensing the 5- and 12-V rails, but ending up at the REF pin of the TL431. I also did add leakage inductances in the secondary side, but more would be required to obtain the complete picture. If I now step the 5-V output, regulation is acceptable but the 12-V output rises a bit during this event:
If you compare the turns ratio for the 12-V output with the value selected in the averaged model, I have tweaked it down. And it is exactly how it will go with the prototype: you check the voltages and various drops then decide to adjust the turns ratios, or even rebuild a new transformer with a different windings layout. Yes, this is quite a long process in the end, especially if you want well-regulated outputs. That is why many folks prefer to add a downstream dc-dc or linear regulator in the end ^_^
Another option I remember exploring years ago, was the single-input dual-output (SIDO) converter. See here for instance. Two output switches kind of multiplex the primary current and actively distribute it to the output capacitors. Regulation is improved but efficiency can suffer and cost also. Power Integrations recently released an integrated version of this old principle (see their Innomux).
These new LTspice files will be uploaded in my ZIP file which contains a myriad of free ready-made switching templates.
