Audio LLC HB SMPS with Peak Load Capability

[ss44]

Hi All,

I'm designing half-bridge LLC converter for audio amplifier with initial parameters:

AC Input Voltage Range 95-265VAC (switched by voltage doubler+2 bulk caps, no PFC)
AC Input Voltage Nominal- 115/230VAC
AC Input Frequency 47-53Hz
DC Output Voltage +/-56V (regulated)
Output Power (Continuous): 200W
Output Power (Peak): 1kW
External Resonant Inductor
Main Transformer: ETD-39
Allowable switching frequency range: 68kHz-140kHz

I've already seen many examples, design tools, guides and read quite a lot... Most of them are for continuous output current and PFC in a front.
My question is - how to achieve that kind of peak power without having a problem with "hiccup" of the power supply (self-restarting when peak occurs - I already experienced that). I would like to create a power supply that is reasonably small and still have a possibility to create high peaks (also @110V mains) without blowing transistors/components or triggering OCP (if it's possible at all).
Is there any rule of thumb for that during calculations? Should I use the worst case for calculations (like minimum mains voltage and maximum peak current)?
I did example calculations for worst case and the inductances look really small (Lr:14uH, Lp:84uH, Cr: 180nF).
Please see attached file - I filled the design up to point 5.

 

[Silvio]

A few tips to help you out with your design

1 switching transistors should be strong enough to withstand the peak current with at least 30% more current.
It should be noted that the output capacitors be a bit more than necessary so that there will be enough storage for the peak. this will also give a good punch to the amp.

2 The ferrite core should handle the power at the peak without going into saturation. Please check data at the running frequency and see that an ETD 39 can give the peak wanted at the given frequency.

Ampere turns in the trafo should also be adequate and calculation of current should be calculated at 60% continuous load from the amp (1000w amp = 600w load on smps)

3 From what I have seen so far in commercial units input capacitance should be generous. You will be needing this especially due to using a voltage doubler when running at 120VAC. input capacitance should be around 1.5uF per watt across 320v.

To make it clear for you if you are using 200v capacitors and you have 1000uf on each side of the half bridge then the capacity of the capacitor across 320v is 500uf

4 Your short circuit protection should work at around 1200 watts to allow for 80% efficiency in the smps. This rating will allow the smps to peak the output up to 1000w without tripping.

5 Compromising size with efficiency will not always be a good choice. Either size or efficiency. Remember sub-woofers are rather hungry on the supply rails.

I did not see the circuit yet as it is not possible for me to open the zip file from here but will take a look as I go home.

Regards

Silvio

 

[ss44]

Thank you so much for your tips, it's a really good starting point!

Size/efficiency trade-off - obviously you're right. That's why I want to use ETD-39 - quite big for 200W (but I will check how far it can handle into 1kW peaks).

I didn't upload circuit yet, just a brief calculations in one of the spreadsheets. I haven't decided upon which controller I want to use but initial resonant calculation for all of them should be similar.

I couldn't find any designs with low Lp and Lr values and when I use my design assumptions to fill spreadsheets (any) they give me similar values.
I understand that I have to go for low inductance factor m (for higher boost gain and wide input voltage range) and because of that allow for low Lp value (which will generate conduction losses on the primary side because of high magnetizing current).
However - if I will go for let's say m=3.5, I will get:
Cres: 150nF
Lres: 18uH
Lpri: 65uH
Lsec: 34uH
The question is - is it technically and physically possible to make a transformer with that low values of Lpri Lsec? And if so, the most important - is it a way to go? Is there any "reasonable" minimum inductance for that power? I read that Cres for high current has to be quite big, but what about inductances?

Thank you for your help,
Best Regards,

Steve

 

[Silvio]

Hello Steve, I am not so keen on resonant smps and not much information is found on the net about them. I see quotes of people mentioning quasi resonant power supplies.

I do not know what type of amplifier you are going to use but from what I see their smps are seldom regulated. Through my experience with them the peak output voltage is rather stable if a good smps is built. Putting things to practice from my 1000w amplifier the output voltage varies very little with the amplifier playing at full volume at 1Khz tone and this varies from 80v peak to around 75v with full power before clipping. Link https://www.youtube.com/watch?v=sHH2GMITc7E&t=2s

Some advantages are that if the smps runs at full pulse width then noise at the output is minimal. Output inductors are not really necessary due to this. However good input and output capacitance will make it rather stable. Remember that a smps running at say 60Khz will fill back up the secondary caps around 1000 times faster than a normal iron trafo running at 60Hz during loading an amplifier, music does not impose a continuous load. Keep in mind that you are not going to run the amplifier at 200 watts output especially if its at home. Your ear drums will not feel so good LOL.

Regarding magnetizing current flux density etc. a compromise has to be found. Seeing your recommendations there is quite a current swing between 200w and 1000w. Do not forget that there is stored energy in the output capacitors and this will give you quite a punch when power is needed intermittently as in audio amplifiers.

Why don't you start with a small basic resonant SMPS till you get some experience. This will teach you a lot and then you will learn how it really behaves. Not many people will give away their hard earned experience on a silver plate. I found help here but to fully learn is to try yourself just like I did.

Lastly be careful when playing with offline smps they can kill you. I suggest you read smps information on this site so that you be aware of any hazard that may occur. There are also blog posts with different subjects that can help you out during this build.

Regards Silvio

 

[ss44]

Hello Silvio,
Many thanks for advice again.

Sorry, I wasn't precise about the power - I am talking about 200W of SMPS output (it transfers into 2 channels of class-AB power amplifier including 50-60% efficiency of power amp itself).

In a past I have already done similar project with smaller power and successfully built it and tested it. Please see attached files (any comments welcome). It's built around LCS705 and work pretty well up to around 200-300W power (out of SMPS). However - when it goes to the more demanding pulse peak currents (aka "difficult to drive" speakers) - it's restarting itself creating a sort of "hiccup" behavior. Whole design (as you can see in the spreadsheet) is based on 200W, not more. Now I am planning to change the initial resonant design to deliver about 1kW and (most likely) change the integrated device to discrete MOSFET + controller topology. I also didn't wind the transformer for 200W version by myself so I am not sure if it is possible to make that kind of low inductance transformer on ETD-34.. Have you got any good links about how to wind the transformer to achieve certain parameters?

Regards,
Steve

 

[Silvio]

Hi Steve,
If the smps is not giving enough power on peaks it means that its not strong enough for your amplifier
I am just wondering how are you loading it and I hope you are not driving the amp into clipping.

All you need is a couple of 100w resistors (Quite cheap on aliexpress) 8 or 4 ohms a scope and and a signal generator to see what is happening. If the smps is going into protection mode then, either the setting of the peak on the smps is not set right or it is giving in and cannot cope on the maximum load. Keep in mind that the fets within the chip are only capable for their designed power

About your plan to make the smps to deliver 1KW and you are driving the same amplifier I think it will be a bit exaggerated.

Here are some links to help you out

http://www.diysmps.com/forums/entry.php?145-Making-litz-wire-at-home

http://www.diysmps.com/forums/entry.php?143-Transformer-winding-practices-for-smps

http://www.diysmps.com/forums/showt...s-700w-(IR2153)&highlight=audio+smps+700watts
The full pdf file is on post #13

https://www.youtube.com/watch?v=_K3ixhcTYFg&t=1s
This is the transformer for the 700w smps

Regards Silvio

 

[ss44]

Hi Silvio,

I tested the SMPS for continuous power up to 200W (smps into dummy load) and it performed ok. Then I tried to use the amplifier as a load and connected dummy load (2, 4, 8 ohm) into the amplifier. For 8 ohm, the amplifier is capable to produce maximum power (1kHz Sine, 1% THD max) without any problems (around 2x80W/8Ohm), into 4 ohm it is still ok for a short time (just don't want to do it for a long time as the temperature rises and sine wave is not a music representation anyway). After that, I tested it with the speakers - quite demanding floorstanding ones (because of the passive crossover, the impedance can even drop to 3.5 ohms). Eventually I found that with high transients the output likes to drop out and the controller is restarting itself with about 200ms "hiccup" mode. It is not a thermal protection (because in LCS OTP is latching), it is not over-current protection (I set it to very high value, like short circuit protection). Now I think that resonant circuit is just for too small power and I need to decrease Lres, Lpri and increase Cres. Basically, do another SMPS, more powerful in terms of peaks.

Thank you for the links, wow, that's actually all your work, I admire your contribution to that forum

Well, I guess I need to wind the transformer by myself and see if I can fit that low inductances into ED-39. Thakns for help, I will keep this thread updated.

Regards,
Steve

 

[iop95]

Hi,

At high transient I think controller voltage go below UVLO and enter in restarting mode; add more caps in rectifier and at pin 1.
In restarting mode output voltage is uncontrolled, feedback loop si broken, power not transfered from input to output. Need to correct this to solve problem.
What mean "resonant circuit is just for too small power"? Lp/Lr impose maximum gain at lowest frequency and Cr value will influence Q of tank. I think no need to modify them.
Low Lp or Lm value may obtain with low AL cores or by adding air gap to a high AL core.

 

[Silvio]

From what I can see the transformer or the switching transistors cannot sustain the 1KW transient. I think you are asking too much from 200w continuous to 1000w peak. You have to sustain more capacitance in the input and also the output to enhance performance.

In half bridge topology we do not make gaps in the center leg of the transformer. There is no energy to be stored between pulses. The flux uses 4 quadrants in the core.

Silvio

 

[iop95]

In LLC converters gap is not to store energy.
Is needed to reduce primary inductance (Lp) to designed value, often below 100microH for frequency above 100k and high gain (small Lp/Lr ratio).
Depending of FB/HB, voltage, minim frequency, core area, and flux density, number of primary turns with ungapped core may lead to high value for Lp. So, to reduce it, must use a gap.

 

[Silvio]

In a way yes if it could be the cause of reducing inductance. Well there has to be a balance with number of turns to provide adequate coupling and the total inductance required for the operating frequency. In that case the gap has to be rather small. This will usually happen if you have an oversized core and voltages are small. This can be overcomed if one uses copper sheet instead. This will provide more coupling across the ETD or E core on the centre leg. however it may not be practical for ring cores.

Regards, Silvio

 

[iop95]

Depending of inductance needed, gap may be high (up to few mm).
Very good coupling is not desired in LLC; some time even a high leakeage inductance is needed when no external resonant inductor is used.
Number of turns is related to input voltage at minimum frequency, but primary inductance Lp is a resonant tank factor, linked to resonant inductance Lr; ratio Lp/Lr impose gain below resonant frequency fr, at low input voltage and high loads.

 

[Silvio]

Depending of inductance needed, gap may be high (up to few mm).
Very good coupling is not desired in LLC; some time even a high leakeage inductance is needed when no external resonant inductor is used.
Number of turns is related to input voltage at minimum frequency, but primary inductance Lp is a resonant tank factor, linked to resonant inductance Lr; ratio Lp/Lr impose gain below resonant frequency fr, at low input voltage and high loads.

Yes indeed if you opt to have the leakage inductance instead of an external inductor before the trafo then loose coupling will be necessary to get your resonant inductor combined with the transformer. For this reason they make the windings side by side instead of on top of each other. Regarding the gap in the center leg well it is seldom more than one millimeter.

Regarding inductance Lp it is calculated at maximum input voltage with minimum switching frequency.

Through my experience with a 50Hz inverter making it resonant it was found that with no load the resonant wave form was very nice sine wave but as soon as you load the inverter the wave form changes as resonance is lost and the frequency has to be adjusted again according to the load at that instance to maintain resonance. With the chips available today they do this and the smps will work at resonance at all times.

 

[chas11]

I checked out your pdf file and have a couple of questions, does the finished power supply agree with the schematic 100% except for small component changes?,did you buy or wind the transformer?what basis did you use for selecting values for resonant components (Lr,Lmag and Cr),was the feedback implemented as shown in pdf file?do you take waveforms of FB as you power up supply using a Variac ,plus lower Mosfet gate wave form?How did you arrive at the Cr value?Spreadsheets are nice but I think a good simple guide is AND8255 & AND8257 then use LTspice to simulate your design gains(min,max,nom curves),set your operating frequency at the resonant frequency or slightly below, also simulate your feedback circuit using spice I think you will find that it does not work as expected(simple fix)I think it is a waste of time to use 2153 when the new 8 pin controller IC's (UCC25600, IRS27952) 16 pin favorites NCP1395,96,97, L6599AD offer so many advantages and simply overall design in the end I think you will find Lmag ~ 150-155uH, Cr 82-100nF, Lr 26 - 34uH and ratio of 5 will do the trick with Fr about 90 to 95mHz,ETD49 or EER4220 core set gapped and primary and secondary windings side by side with a 3mm divider between them which you can make with a plastic credit card or plastic cable tie.