Combined PFC AC/DC & DC/DC SMPS

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi SMPS fans:

For my subwoofer I will need a SMPS.

My dreams:
- AC Input voltage range: 120V – 277V , additional +/-10%
- Power Factor Correction: Yes, THD < 15% , at low loads 30%
- DC Input for car use.
- Power Rating: Suited for 1kW class D amp of subwoofer

Well, of course first I was thinking about a traditional push pull
car smps for the DC/DC part.
For a line isolated PFC AC/DC converter a flyback always looks charming. :clown:
No, not really… it is a :headshot:
After some short calculations a flyback
for 1kW clearly prooved to be a walk through hell ! Component loads will clearly exceed any reasonable limits.

My new idea seems to be much better.
The DC/DC part will still be a traditional car push pull smps.
The PFC AC/DC converter would now be a composite of:
- PFC boost converter (high efficiency, easy to handle, perfectly reliable)
- From the output rail of the PFC a half bridge is driving an additional primary
winding of the push pull transformer.

Of course you can only use AC input or car Battery. And the not used MosFets must be
turned off reliably….

Please refer to the attached schematic. It is just a principle drawing, but should help
to convey my idea. Only the components for the basic function are shown.
It is clear to me that I will need EMI filtering, may be some snubbers, etc….
Also it is clear to me that I will have to implement a safety isolation according
EN 60065 (170 pages full of safety requirements and lots of chances for misinterpretation…).
Don’t worry about that. I will manage that and in cases of unsolvable issues I will
simply post my future questions….
Also please don’t worry about the PFC boost. That’s the only section, where
I am 100% sure that I will get it working in a proper way (my everyday business).

My current question is, if you think that this combined topology can work in principle.
I do not see a flaw and it looks to promising to be true.

Looking forward to your comments!
Also additional proposals for the control ICs are welcome. The types which
I have proposed are just a first idea.

Cheers
Markus
 

Attachments

  • combipfc.zip
    13 KB · Views: 753
Hi Amp man !

Nice insights. Interesting link. Thanks for this!
It will give me further inspiration.
But their SMPS is a push pull forward.
Can't follow your perception that it offers a PFC, that
draws sinusodial current from the mains.

Nevertheless: Really cool that QSC offers their schematics.
Probably they dare to do this, because they know, that this is not enough for copying their amps.
 
ChocoHolic said:
Hi Amp man !

Nice insights. Interesting link. Thanks for this!
It will give me further inspiration.
But their SMPS is a push pull forward.
Can't follow your perception that it offers a PFC, that
draws sinusodial current from the mains.

Nevertheless: Really cool that QSC offers their schematics.
Probably they dare to do this, because they know, that this is not enough for copying their amps.


hey buddy,

this schematic is of their discontinued model PL9.0 which is obsolete 2 years ago from now. THis amp is suffering from a problem thats why they have to discontinue it.
The Power factor correction is clearly evident on PFC LOOPS page of this document.
I have access to this document by some other means, qsc doesnt give the schematics of thier regular production models.

The qsc amps are very easy for copy indeed.
try this link for more discontinued models schematics and see how easy are their schematics.
http://www.qscaudio.com/support/technical_support/schematics.htm
regards,
kanwar
 
Hi Kanwar!

...hm, I was less irritated by the control loops rather than by the
power electronics circuit...
At first galnce, it looked like there would be no possibility to transfer energy during that time when the mains sine wave is far away from
it's peak. But the clue seems to be the combination of L1 and the switched transformer T1. Quite interesting! Especially that you can
skip a lot of caps, compared to two separate stages.
But it does not look like beeing easy to handle.

@ Leolabs:
Up to know I did not build it. But I plan to do and because of that
I thought it might be a good idea to get some opinions first.
But here don't seem to be opinions.....

Bye
Markus
 
SMPS PFC Amp

Choco-

Sounds like an amp I wanted to do a while ago, minus the DC-DC section. I, too, wanted the output of the Active PFC to be a couple of series caps to provide a centerpoint for the half-bridge main power transformer. I have a pic of a 185W Active PFC I built about 5 years ago as my Senior Project for my undergraduate degree. I will post a pic of it when I get to the CPU where it is.

The control chip is OnSemi's MC33262 critical conduction PFC, run in the continuous-conduction mode, and the two output caps are 390uF, each rated at 250WVDC. The output is set for 385-390V.

The main MOSFET is an old Motorola MTW20N50E (20A, 500V), now discontinued. Any 500+ V N-Channel MOSFET from IR or others should do.

The input rectifiers are MUR8100s (8A, 1000V), ultrafast types. These should be ultrafast types as the current pulses being drawn through them are high frequency, modulated at the rectified line frequency of 120Hz (100Hz in Europe).

The flyback diode is also an MUR8100, although there is some ongoing discussion as to whether this diode should be just a fast or even standard-recovery type, depending on whether the circuit is run continuous- or discontinuous mode. I used an '8100, because that's what I had on hand.

I tested this circuit once, and promptly blew its input fuse upon power-up. Since I'm a little reluctant to go messing with AC mains voltages directly, it will be some time before I can get to powering it again, since I don't have a line-isolation transformer handy.

As for the PWM controller, I am using OnSemi’s MC33025. I have used this chip many times with good results. This is pin-for-pin compatible with Unitrode’s UC3825, and is the next-generation of the venerable SG3525, with higher speed oscillator (2MHZ), wider bandwidth error amp, voltage- or current-mode operation, ease of paralleling, and high-current totem-pole outputs(+/-2A each), for driving MOSFETs that have lots of gate capacitance. I'm not slamming the '3525, I still use them, but for those wanting to do an SMPS with all of the things they wish they had in the '3525, they're all in this chip.

If OnSemi was really on the ball, they would combine a dual-output PWM chip with an on-board PFC section to synchronize the two sections and avoid beat frequencies. In case there is such a chip today, I am not aware of it.

Anyway, I digress.

For the gate driver IC, I am using the old Motorola MPIC2113, their version of IR's IR2113, which is a high- and low-side MOSFET driver, rated at 600V, and is specifically designed for driving MOSFETs in half- or full-bridge configurations.

Since I have a few of the MTW20N50Es still laying around, I will use them for the PWM Section, as well.

Since I'm just starting out a new family, it may be some time before I actually get to do this, but I've been gathering the parts for all of this for some time now.

Hope this gives you some ideas that you're not the only one thinking along these lines.

Best of luck,

Steve
 
Hi Steve,
that's exactly the discussion I am seeking for.

Only point which seems strange to me that you were running your boost in continuos conduction mode. This causes full hard switching.
Especially when the MOSFET is turned on then it is switching into the conductive boost diode. This is giving quite impressive currentpeaks in the diode and in the MosFets you will see the high current peaks and still high voltage at the same time.

Usual you run these boost converters in the critical conduction mode. Exactly at border between continuos and discontinous mode.
Means: Right after demagnetization of the choke, the switch is turned on again. If you use nice timed zero sloping detection, then you can switch on lossless. L6561 and L6562 are providing this. Also there are several similar chips. Furtheron their criteria for MosFet turn off is the current in the MosFet. This makes the system very rugged against disturbances and also against saturating chokes.

I will have a closer look to the UC3825. If I remember right, then I have some of them somewhere in my mess IC assortment....

By
Markus
 
BTW:
My high DC rail of 480 is a result from my 277V/AC wish....
Please note that your 390V are already the lowest value which you can handle at 230V mains. If the difference between mains maximum and the DC rail is coming to small then the zero detector of most
control chips start to make nonsense.... Usually no defects happen, but
control is a little bit "disorganized" :D

OK, we should stop the boost topic here, otherwise the moderators will close this thread..... :judge:
 
Topic Transfer

A better idea would be to transfer this thread to the Power Supply Design section, as a good PSU is ABSOLUTELY critical to the good performance of any amp. :bigeyes:

As for the MC33262, I must have misspoke myself, as this chip is a critical conduction chip, how could I run it continuous? Duh!
:xeye: I need to go back and re-read the datasheets and applications notes for this chip.

BTW, there is somewhere buried in Motorola's pages, a schematic for a 450W PFC Ckt. I just need to find it.

Also, why the need for 277VAC input? :confused: Isn't this usually reserved for hi-power lighting? I realize that is also the phase-to-neutral voltage for 480V 3-F systems. Are you planning a 3-F input for your PFC? :confused:
 
Hi- relatively new around here.

I have tried a similar idea using a flyback and push-pull. It was for a lower powered amplifier, but still had one primary offline and one from battery. No PFC.

ChocoHolic- one thing you should consider is the voltage that you're going to have on the half-bridge primary when the push-pull circuits are energized. As an example, the turns ratio I had was such that when using the push-pull, it was creating a stepped-up voltage on the other primary due to a turns ratio that was very large. For instance, I think I had 48 turns on the flyback primary, and only 4 on the battery primary. I had almost 150V on the the flyback primary when I was using the battery mode- and it was charging my AC side input caps! Of course, if the FET's on your half-bridge side are of suitable voltage, and you're holding them in off state when not used, this shouldn't be a potential problem. It's just something I hadn't considered when I began the project, and something you should look out for. It may not even be an issue for you depending on the turns ratio.

Another thing to watch for is the increased leakage inductance you're going to have with the added primary there. Depending on your transformer winding arrangement, and whether you have AC or DC input, you may have an unused primary in between your used primary and the decreased coupling could couse some larger leakage inductance spikes. It depends on how the transformer is wound.

Just some thoughts, Matt.
 
Hi Matt,
yes, I was already thinking about the back energizing by the unused
primaries. But it seems to be no issue.

I am lucky that I do not have a flyback.
My boost PFC is coming fine from theory and matching to some components in assortment.
For 1 kW drawn from a 120V mains I will need 8.3 A.
Means 11.8 A peak. The HF current peak in the PFC inductor comes to
the double value. Say 24 A. Regarding the switch, I still have some
nice Cool Mos SPP17N80… 2 in parallel will do the job.
I have some ETD44 on hand. With a litz wire 150x0.1 I will get a nice 4 layer winding with 64 turns. Rdc around 0.1 Ohm.
ETD44 offers 173mm^2 magnetic cross section. I am ending up
at Bmax = (24A x 150uH) / (173mm^2 x 64 ) = 325mT .
That's pretty fine. From data sheet curves I expect around 5W losses in the ferrite.
In the copper I am expecting much higher losses.
From experience you can double…..tripple the losses which you would expect by Irms and Rdc! This is due to proximity effect and due to the eddy currents induced by the ringing field of the gap. Factor 2..3 is fine if you use 0.1 stranded wire and winding package is not to big. So in the copper I would expect about 50W.

Don't panic! Altogether 55W losses at 1kW is fine, if we take into consideration that the power crest factor of normal music is around 15..20 !
Even with a power crest factor of 10 we would end up in about 5W
average losses in the ETD44.
The design must be able to handle the 1kW electrically fine, but thermal average load
is not that critical. (BTW I will implement a thermal limiter, not only shut down. If one of the critical components is getting to hot, then the audio signal will attenuated.)

OK. Now the push pull stage:
Back energizing by the unused primaries will charge the unused DC supply rail.
The path is given through the unused primary and the reverse freewheeling diodes of the unused MOSFETS. I powered by 12V, then DC rail of the halfbridge will be charge to full voltage around 450V…500V, according the turns ratio of the transformer. If there is coming some uncontrolled additional energy due to leakage effects of the transformer this additional
energy will not overload the DC rail, because the symmetrizing resistors of the rail caps will
are consuming energy. For the other way round, if powered from AC, then the DC-Input will be charged to 12V…15V according the transformation ratio. Probably I should also add some
energy dissapating resistor in parallel to the heavy 2F cap.
Trafo: I have some perfect high power torroids from the VAC. Material is Vitroperm 500F, which is saturating at 1.1T. Low HF losses (20kHz/0.2T/ only1.4W per kg, and at 100kHz only 35W/kg). Additioanl it offers high permeability, which make it easier to get a good coupling. My torroid 63mm x 50mm x 25mm offers AL=19uH !!!
Ae is 124mm^2.
Operating at 50kHz and using only +/- 0.6T is resulting in:
N_HB = 16 turns for the primary from the halfbridge.
N_PP1 = N_PP2 = 1 turn for each primary of the 12V push pull.
N_S1 = N_S2 = 3 turns for each secondary.

May be, I will double all number of turns in order to get smaller increments for stepping up or down the output voltage…
Further this would allow any operating frequency above
human perception (say 20kHz).
Also I am still thinking about how to distribute 1 turn over the entire torroid…..

Bye
Markus
 
PFC Pic

Here's the PFC Pic I promised.....
 

Attachments

  • active pfc 2 small.jpg
    active pfc 2 small.jpg
    94.4 KB · Views: 1,290
Hi Steve!
In fact I was not aware that we really have a PSU sub forum.
....well....
Thanks for the link.
Really nice set up your SMPS. Do you really always make a PCB? Even you will only built one or two samples?
I am becoming a fan of 3D-P2P (3 Dimensional Pin to Pin) set up.
3D-P2P offers best opportunities for short wires, small loops, heat disappation & low parasitic effects, - if well done.
Your MC33262 / MC34262 is exactly the ON-Semi version of the L6562, which I had proposed... The most traditional and reliable way for critical conduction / transition mode PFC boost, I guess :)

FAN4822 is a nice variation for optimized softswitching. But it does not overcome the turn OFF losses of the switch.
The turn ON is already nearly perfect in a proper designed MC33262 / MC34262 / L6562 design. After demagnetization of the choke, the voltage at the drain of the switch is not only jumping to the value of the input voltage, but due to unavoidable capacitances there is a resonant down sloping and you can manage to turn on the switch
at nearly perfect ZVS and ZCS (both !) during most of the time of the AC mains wave. Only when the mains is close to max. (theoretical condition: Udcrail vs. ACmains < ACmains vs. ground)
During that times the down sloping will not come down to ground
and the switch will be turned on without ZVS. ZCS is still vaild, except charge for parasitic capacitances. I practice I observed that the turn ON losses are neglectible compared to the turn OFF losses.
Turn off losses can be optimized by fast turn OFF ( but take care for EMI ! ). A snubber cap across the switch can help here, but must be taken into consideration when turning the switch ON.......

FAN4822 could provide perfect turn ON soft switching during the entire AC cycle. But in my application this small advantage, will not justify 3 additional power components. Especially as it does not overcome the issues with the turn OFF losses.

....will go now and dig around here for the push pull converter and how to regulate it most fortunate. Regulation possibilties for these push pull converters without additional choke seem to be quite limited.....

Bye
Markus
 
pc boards

Markus,

Almost all the time I try and do a pc board. If it's done right, there's plenty of room for error on it. I just try not to re-solder components ten or twenty times when swapping in and out different values of components because the repeated application of soldering iron heat will make the copper come away from the board. I have done probably 100-200 pc boards, and once I got the process down, the only time-consuming part is doing the artwork. Most of the boards I've done are the direct-etch types. I have done a couple of photo-sensitive boards, and the results are very good, but the time and involvement with the extra chemicals is a big pain in the a$$, if you know what I mean. Besides, with little kids around, I like to keep as few chemicals on hand as possible.

Anyway, getting back to the pc board, I guess having lots of Radio Shacks around is a good thing, as I am always using their direct transfer rub-ons, and these work very well.

As of late, I have employed generous use of ground planes for all audio and high-frequency circuits. Two reasons: First, noise problems are greatly reduced; and Second, you eat alot of less copper off the board, and your etchant (FeCL3 or whatever you use) lasts alot longer. A win-win situation here!.

About ten years ago, I actually completed an AC-Mains SMPS-powered audio amp. It was 40-50W/ch, and the half-bridge converter put out around 200W. Outputs were +/-33V, and the controller was SG3525, with a simple driver transformer for the two IRF840s. I did the power supply in modules- Input line filter, PWM controller, power switch and driver section, main power transformer, and lastly, output rectifier/inductor/filter section. Because the SG3525 was on the primary side, I used an optoisolator (MCT271- replaced by MOC8102) to sample the output voltages and send the error signal across the galvanic barrier, back to the (+) input of the '3525's error amp. Regulation, line- and load-response were very good.

The circuit ran very well, little-to-no noise, and the audio was very good. The amp didn't really pick up any noise, for it being not shielded at all. Unfortunately, it was disassembled long ago, and the parts used for other projects, so all I have are the memories if it.

My next amp will be a much better effort, witrh a PFC'ed SMPS powering it.

When I get around to powering back up my PFC shown on this thread, I will start looking at the switching waveforms to optimize switching and conduction losses.

Steve
 
Discontinuous-mode boost PFC converters are only practical at low power levels since at high powers it involves very high peak currents that cause high copper and switch loses and require a huge EMI filter to isolate the mains line from the high dI/dt of the inductor

For >1KW I would recommend continuous mode since turn-on losses may be reduced to a minimum by means of magnetic non-disipative snubbers

Also ZVS and ZCS are not practical at high power levels since they require extra power switches and inductors and they impose limitations on maximum duty cycle, causing poor current rise slope after zero crossing

This is my prototype, it uses a L4981A control IC from ST working at 45Khz to reduce switching losses. It also contains : Additional components to achieve consistent current limiting without entering chaos-mode, boosted CT discharge to achieve nearly 100% duty cycle [this dramatically reduces THD near zero crossing], buffered gate drive in order to avoid blowing the control IC everytime I blow an IGBT, a potentiometer for fine output voltaje adjustment and a jumper to select between 230V and 450V output [230V is safer for testing new features]

The input diode bridge is placed in an external heatsink and the EMI filter is not shown. I've successfully tested it up to 1.7KW continuous with proper cooling at 450V output and 160V AC input, however the board has space for additional components in order to get up to 4KW theoretical maximum output. Currently I'm using SKP20N60 IGBTs and MUR860 diodes combined with a coupled inductor that works as a magnetic snubber and contols the dI/dt at turn-on [and also at turn-off as a side effect]

Total output capacitance is 990uF 500V, quite high to reduce ripple voltage. The boost inductor is wound on a E65 core with 2mm gap, measures approx. 390uH and starts saturating at approx. 19A [I have no datasheets for the core]

The project has been frozen for months since I can't afford bigger isolation transformers to test above 1.7KW and I'd have also to buy additional electric heaters to dissipate such a high power

Some pictures :

pfc2k_0.jpg

Note that the control circuit is mounted on a breadboard to allow for further experimentation, only the critical switching sections are on the PCB

pfc2k_1.jpg


pfc2k_2.jpg


pfc2k_3.jpg


pfc2k_4.jpg

This image shows the magnetic snubber in detail, it's just a coupled gapped inductor that stores energy during turn-on and returns that energy to the output capacitors after turn-off. Primary inductance measures aprox 10uH and it saturares above 30A. It's designed to get 45A/us turn-on slope for 450V output. Primary to secondary turn-ratio is 1:3 [the secondary-side rectifiers have to withstand >1500V during switch turn-on so four MUR460 are used in series]. This could be called soft-switching
 
Hi Eva!
That's great! I will dig a little bit in the data sheet and application notes of L4981A.

In fact I do not need 1.7kW continuos power!!!
1kW short time 200W...300W avverage would be enough for me.
..1.7kW and still not enough. What are you planning to do?????

You are right the transition mode is not perfect for high power.
But it is somehow nice that I would have the required components on hand right (except boost diode).
Well I would need a MosFet driver for two or three paralleled 17N80,
...unbelievable ... I have some UC3710-T drivers from Texas with me...
Originally I planned them for some monster Fets in the amp section.
But the amp section is coming less heavy now (full bridge with 4 IRFB52N15D, - nice! ).
Nevertheless, your design makes me curious. Quite curious!

BTW: High di/dt in the inductor? di / dt of the inductor is causing conducted noise into the mains, right. But usually this mainly causes
noise in the 30KHz.....500 kHz. I can filter this without VooDoo dancing.
Much worse: We usually figure out that the very,very high di/dt in the Fet and in the boost diode and du/dt sloping of drain cause trouble in the 10MHz....1 GHz.... Filtering here is 100% Voodoo.
And usually fast and hard switching devices cause even more trouble in the MHz area. So I try to avoid hard switching if
possible.... Unfortunately radio and TV are quite sensitive here....
If your design is fortunate in this regard, it would become even more interesting. Your magnetic snubber looks really promising!

Bye
Markus
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.