Proposal for a more realistic first project - 150W offline forward.

[KX36]

Hi All,

Please allow me to introduce myself. My name's Matt, I'm a scientist by trade and I've been into hobby electronics for about 3 years now, having spent about the last 2 years reading and SPICE simulating various SMPS and DC-DC converters. So I'm just starting to feel comfortable with building one for the first time. I'm glad I've found somewhere where amateurs are discussing this sort of thing, last time I came to this site I believe the SMPS section was decidedly empty.

I've been reading around here at all the "I have no electronics knowledge, please give me a 1kW phase shifted full bridge converter schematic so that I can blow myself up" (paraphrasing) threads and I find these both mildly amusing and quite scary. So I'd like to propose a more realistic first project and I'd like some input from the more experienced members. I'm a member at AX84.com and I like their approach to useful beginner projects, which I think we could emulate.

I'd like to build a bench power supply for general low voltage use. The sort of thing you could easily buy but the point is to learn something.

It should be:
-Offline input (an isolating topology) or a post-regulator for an AT/ATX PSU.
-universal input voltages (i.e. somethign like 90-250V ac)
-Adjustable output something like 5-30V dc, 5A.
-Efficiency is secondary to simplicity, and cost should be reasonable considering the lack of economies of scale.
-Should be a through-hole design which people could more easily make by hand than SMD, although the trade off is increased parasitics and EMI.
-I'd like to make it with discrete components rather than controller ICs for the sake of education, but I accept that this goes against the simplicity case and adds unnecessary parasitics (still a TL431 is often used as a discrete secondary side error amplifier and that seems to work OK) any input on this?

I'm thinking this is probably in the range of a forward converter. I have read that the linear transfer function of buck-derived topologies makes them more suited to adjustable out/in voltages than boost or buch-boost derived topologies. I have also read that voltage-mode control is better for this than current-mode control, (especially with feedforward of the input voltage to the PWM ramp voltage) but I don't know really which is more appropriate.

I've also put a lot of thought into running it off the 5V or 12V output of an AT/ATX power supply with a buck/forward converter post-regulator (depending on whether the required voltage output range is above 12V) People frequently hack these into "lab bench power supplies" but in really rubbish ways and I thought we could do better and give it a variable output and current limiting. Even if a forward converter is used, isolation is not necessary and this would remove the complexity of an optoisolator (which I'd like to avoid if possible, optos need to be tested for stability as their bandwidth etc widely varies between devides and amateurs don't have the tech for this), and the input voltage would be a steady low voltage, the universal input voltage range is taken care of by the ATX PSU.

I think the output current would probably have to come down from 5A if a forward converter was used after an AXT PSU as the peak input current on the input is much higher than the average output current. This could be partially accomodated by adjusting the Nr:Np for a higher duty cycle and therefore lower Ns:Np, so lower peak primary current for a given load current. I was wondering if there is likely to be any problem with the pulsatile current draw on the ATX PSU, whether this would be a problem in various synchronisation states with the ATX PSU's switching, or if it might work with some ATX PSUs but not others.

If the PSU is a standalone offline one, how realistic is it to have a wide input and output voltage range? Everything I've read so far has talked about a fixed output voltage and duty cycle adjustment to account for the input voltage variation, but if both are variable, that seems to be quite a wide Vout:Vin range to try to accomodate. Any comments on this?

There's a lot for you to chew over, hopefully some useful discussion will come of this.

Cheers,
Matt

 

[KX36]

Just realised this should probably be in the "diysmps" section. Bit weird there's a "Power Supplies" section and a "diysmps" section... anyway, sorry for that.

 

[nigelwright]

Hi All,
I've been reading around here at all the "I have no electronics knowledge, please give me a 1kW phase shifted full bridge converter schematic so that I can blow myself up" (paraphrasing) threads and I find these both mildly amusing and quite scary. So I'd like to propose a more realistic first project and I'd like some input from the more experienced members. I'm a member at AX84.com and I like their approach to useful beginner projects, which I think we could emulate.
Cheers,
Matt

I had a go at using an L6565 circuit but replacing the l6565 with a PIC.
I blew up a couple of mosfets and fuses and blew out the house circuit breaker a couple of times before i got to grips with it.
A 40 watt lamp in series with the mains helped calm things down a bit and later a 100 watt lamp as i started to take more power.
I did a flyback SMPS. Looking at waveforms on the scope helped me understand how it was working. The transformer is charged up magnetically by the primary pulse and power is taken from the secondary as the magnetioc field is discharged.

 

[KX36]

I had a go at using an L6565 circuit but replacing the l6565 with a PIC.

Yeah, I think I saw you write about this before, but I really don't understand why you'd replace a SMPS control IC with a digital microcontroller. I would have thought that would just worsen the transient response and complicate the frequency compensation which normally would be on the error amplifier. I also don't know why you'd bother using a quasi-resonant ZVS control IC (presumably your PIC is immitating the L6565?) for a flyback converter. Flyback converters are very limited in the power they can deliver because their "transformers" store their energy in the magnetising inductance as the primary and secondary conduction are out of sync. This is inefficient and needs a gapped, bigger transformer than other topologies, so why try to do ZVS on that, adding complexity and headaches for the sake of a minor increase in efficiency on a low power SMPS?

Anyway, that's kind of gone off on a tangent, hopefully we can get back to my suggestion of an educational beginners' project forward converter.

Matt

 

[Kanwar]

Infineon has some great combo ICs for PFC front end + Forward converters, better have a look at them, they are simple & easy to do.

 

[KX36]

Thanks, I've had a look at their ICE1CS02 controller. Certainly looks interesting and quite possibly a good candidate for the basis of a first SMPS project. IIRC, it has no internal error amp in the PWM section, so relies on an external TL431/opto for that.

The problem with TL431/opto for this, as far as I can tell from this article: http://switchingpowermagazine.com/downloads/15%20Designing%20with%20the%20TL431.pdf, is that the stability of the whole system is dependant on the appropriate compensation, which is itself dependant on the AC characteristics of the opto, something which is highly variable between one part and another and something wihch people can't really test at home. If this problem is significant, it raises the possibility that some people making the same kit project build as someone else may have a highly problematic SMPS where others may not.

However, the only alternative I know of for offline converters is a secondary side controller and a pulse transformer back to a gate driving IC and all the extra bother and expense those things cause, I don't have the experience to say whether it's likely to be more successful than the apparently pot-luck simple TL431 approach, or worth the extra bother and expense.

 

[MicrosiM]

KX35

Welcome to diysmps, Please check attached and see if it helps

Please update me

 

[KX36]

MicrosiM, thanks for the welcome and thanks for the schematics. They are indeed helpful, have you built these? I hope you will see the following as constructive discussion for my sake rather than criticism, as that's not my intention.

It's always useful to see full schematics, and yours incorporate some features in ways I have thought about but not seen before such as how you interface the DACs with the error amplifiers through their non-inverting input. I had a discussion with someone about doing that before for a programmable power supply, but was advised instead to use a digital potentiometer for the bottom resistor of the potential divider into the error amplifier as changing that does not affect the stability of the loop whereas changing the reference voltage might. DAC or digital pot doesn't really make much difference as far as interfacing with the microcontroller, but obviously it's not so easy if you have differential amplifiers instead of just a potential divider.

The other thing I noticed from those is the way you sense the output current and voltage and then have 2 opamp differential amplifiers, one with the difference of the voltages across the current sense resistor (one of which is ground), one with the difference of voltages at the output. It seems a bit strange to me how you've done that. If I'm reading it correctly, the current opamp shouldn't need to be a differential topology if one of the inputs is ground anyway. Normally, I'd expect to see the ground on the other side of the current sense resistor, then the output voltage and sensed current would be referenced to ground and neither would need a differential amplifier (current would be a negative voltage, but you can easily invert that with an opamp). If you actually want to measure the load current rather than the inductor current which includes ripple and capacitor charging, you could have the current sense resistor between the output filter capacitor and the load (simply, on the low side; or with a differential opamp such as a specialist "high side current sense amplifier" on the high side). Incidentally, I think as it is at the moment, the current sense resistor would be adding to the output capacitor's ESR to lower the zero frequency of the power stage, although it shouldn't be as much of an issue in your current mode controlled power supply as it might have been in a voltage mode controlled one where there is also the LC double pole to think about.

The UC3825 is one of the controllers I've been considering, so it's useful to see it being used on the secondary side. I am leaning towards a secondary side control with pulse transformer feedback to a UCC3960, which I quite like the idea of. It gets around the problems of giving a housekeeping supply to a secondary side controller by providing the PWM signal itself until it gets some feedback from a secondary side master controller through a pulse transformer, which avoids the opto. IIRC, it can limit the peak primary current even when acting as a slave, so there should be need to feed back primary current to the secondary side controller through a transformer. Haven't chosen a secondary side controller yet but it doesn't need to be too fancy. Probably need a low voltage controller though so that it can be supplied by the output voltage while the output voltage can be quite low. I had a quick look at the UCC28250. It's quite fancy and up to date and can run off 5V. Downside is it's only available in surface mount. Considering a forward converter with feedforward voltage mode control if no active PFC preregulation (feed forward from a rectified secondary voltage (might needs its own winding) rather than the primary, which I haven't seen done before).

Thanks again,
Matt

 

[noz-25]

MicrosiM please,is it OK to make this one SMPS but only with main transformer and rectifier behind to get fixed voltage,i need 48V 5A!? Maybe i ask some dump question but i am beginer.Also is it possible to use bipolar transistor and not MOSFET?(I don't have MOSFET,i order it and it need min. 20days until they arive,but i have couple pcs BLD128 and D13009) Thank You!

I am talking about Your shematic in rar.

 

[norazmi]

bipolar use different topology and you cant just change with same circuit that running mosfet. but bipolar is less efficiency and more loss on heat, so better prefer mosfet

 

[Alvaro]

bipolar use different topology and you cant just change with same circuit that running mosfet. but bipolar is less efficiency and more loss on heat, so better prefer mosfet

salam pak cik norazmi,anda punya layout 1,2kva dari detex dgn gambar nilai komponen yg lebih jelas?

 

[KX36]

Hi again everyone,

Sorry it's been so long since I gave an update, life has been getting in the way. (Went abroad, been donig long hours at work, applying for other jobs and buying a house...) I actually wrote a long reply in about Feb, but my login timed out and I lost it and this is my first real chance to get back to SMPS.

In summary, I finished a design I have significant confidence in January, it's all simulated, seems to work well, but to get the universal input voltage, variable output voltage, and wide load range I had to make things significantly more complicated. I will still build it up and troubleshoot it and hopefully eventially release the design with some theory docs, but I'm afraid to say it's now too complex for a first project.

Any comunity first project will have to have tighter limitations and be less useful as a device in itself.

I will probably design a series of simple, relatively low power, non-isolated buck or forward converters with discrete components for error amplifiers, PWM comparitors etc. There would be different projects in feedforward voltage mode, peak current mode and average current mode etc, which can have lots of test points for troubleshooting and waveform study. I would also write theory documentation, how to calculate the component values, how to design the transformer and inductor, how to deal with the parasitics etc.

Regards
Matt Helyar.