SMPS Hissing & Whining Noises

[julianbenson]

Hi Guys,

This is my first post on this forum, so go easy please

I'm aiming to to build an offline SMPS for larger audio power amplifiers, but have decided to start with a simpler 12V SMPS as it seems a bit safer without the mains voltages and there is loads of documentation etc.

I'm a keen SMPS constructor, but until recently haven't really had too much success. I had really noisy transformers, latched / blown SG3525 outputs, fried mosfets etc etc.

Anyway, I have now designed a transformer around a large ferrite inductor core ITL501 obtained from Wilco. After studying various sources on the design of such a transformer I created an Excel spreadsheet that allowed me to adjust parameters easily and determine the turns ratios etc as shown below:

Ap
Height 1.590 cm
Outer Diameter 4.910 cm
Inner Diameter 3.380 cm
Ac 1.216 cm2
Wa 8.973 cm2
Ap 10.914 cm4
B 800.000 Gauss
k 0.006
f 70000.000 Hz
Po 965.531 Watts

Windings
Po 965.531 Watts
Vp 24.000 Vac
Vs 78.000 Vac
B 800.000 Gauss
f 70000.000 Hz
Ac 1.216 cm2
Np 8.809 turns (5+5) Skin effect: 4 conductors
Ns 28.628 turns (15+15) Skin effect: 2 conductors

The 12V SMPS runs now with a simply POT input unregulated output setup. I have pretty much followed the SG3525 datasheet exactly, apart from adjusting Ct & Rt for desired frequency.

If a adjust the POT i.e. pulse width to about 30%, the thing runs fine with load apart from a low output voltage/current. If I then adjust the pulse width to say closer to 45%, it starts to hiss and whine even with and without a load connected.

My question is what causes these transformer to hiss and whine so much even without any load?

Thanks for any help guys!

BTW: Eva, I love your work, you are a legend ;-)

[TechGuy]

[/QUOTE]My question is what causes these transformer to hiss and whine so much even without any load?[/QUOTE]

Saturation is the leading cause. It looks like your doing a Push-pull. set up. Uneven winding can cause flux walking, one half of the push pull cycle magnetizes the core more than the other half. windings in a push pull circuit must be dead even. You might be able to correct the problem by adding dead time.

A dead giveaway of core saturation is if your input current is high with no load.

BTW: At low switching frequencies, the switching can produce audiable noise because the winding will vibrate in response to the magnetic field generated. Even at higher switching frequencies audio noise can be produced either from harmonics or because of dynamic loads that change in the audiable range. Mass produced transformers have vacuum impregnanted windings with varnish to prevent vibration. The vibration can rub off the insulation over time.

FWIW: a transform with 5+5 turns seems low for a transformer switching at 70 Khz at the power level you stated (900W)

I have now designed a transformer around a large ferrite inductor core ITL501 obtained from Wilco

This is probably inappropriate for a SMPS transformer. Inductor cores consist of materials that make poor transformer cores and are designed for filtering.

[julianbenson]

Hi TechGuy,

Thanks for your helpful response, I owe you a beer or two no doubt.

I think it is core saturation due to core imbalance as increasing the dead time as suggested significantly reduces whining noises and also improves the no load current consumption to about 80mA from about 900mA but also reduces output voltage as expected Some other reading indicates that I may need to use the bifilar winding technique as I’m only currently covering about half the core with each primary winding. So at the moment each primary winding either side of the CT only sees half the core circumference and not all the way around. So I'll rewind the transformer and expect a great improvement hopefully as the primary windings will be cover the same core area and be better balanced.

But first I’m now also wondering if I’m using the correct equation to calculate the Np for a push pull toroidal transformer and whether I am using the correct parameters. The equation I am using is:

Np = (Vp * 10^8) / (4 * f * B * Ac)

Where:

Np = Number of primary turns
Vp = Input voltage
f = Oscillator frequency of SG2525
B = Maximum permissible +/- Gauss value for material / core used -> I usually use 2000
Ac = Ae = Cross sectional core area from core datasheet or calculated from Ac = ((OD - ID) / 2) * H

Questions:

• Does the above equation look correct?

• Should I be using Vp = 12V or 24V?

• How does Np relate to the physical windings, should the transformer be wound using Np-Ct-Np or 0.5*Np-Ct-0.5*Np?

• SG3525 frequency = 70kHz then transformer f = 70kHz, I don’t have to divide the SG3525 frequency by 2 for some reason?


Thanks again for all your help, it's much appreciated

[julianbenson]

Also just a side issue not my biggest issue right now, but using individual diodes, if I connect the two diodes in RED the current consumption increases and rail voltages reduce too with or without a load.

If I leave them unconnected the rails voltages seem fine and less current is used. It almost seems like the two RED diodes are shorting something out. I've seem some SMPS that only have one diode per rail i.e. no full bridge.

Any ideas what is going on here?

[dtproff]

OK... I'll jump in here.

The full article on this supply can be found at
http://www.scribd.com/doc/4071900/Sw...-for-Car-Audio

You will find all the equations there also.

Tony

[julianbenson]

Hi dtproff,

I am very familiar with this article as it was one of the first ones I ever studied with great enthusiasm. Unfortunately there are no equations relating to the design of the transformer. There is some maths relating to RMS audio power and the turns ratio calculation, but no transformer primary turns equations.

The article although fantastic and very inspirational is limited to using a particular core with a rule of thumb primary turns count. It makes no mention as to how to adapt the design for other cores and parameters. What if I want to change the core or the operating frequency?

Thanks.

[dtproff]

The basic equation is Np=(Vin*ton)/(Bm*Ae)

Typically you will use the lowest expected voltage on the bulk cap for Vin. Set Ton to 48% of you switching frequency. Ae is defined in the data sheet.

For Iron powder cores you can run them up to about 1T (10000 Gauss) but be sure to check the data sheet. If you use a regular ferrite core you will need to make sure the leakage between primary and secondary won't saturate the core when you hit full load since their is no gap in a toroid.

The tuns ratio is then set by Ns=Np*(Vo+Vfd)/(Vin-Vdfet-VdxfmrPri-Vrsense).

Vo=Output Voltage
Vfd=Forward voltage drop on the rectifier diode)

I usually drop the following from the equation:

Vdfet=Voltage drop across the FET
Vrsense=Voltage drop from the current sense resistor
Vdropxfmr = voltage drop through the primary winding

I hope that helps.


Tony

[dtproff]

By the way, I've seen McLyman use that equation you had above for forwards and push pulls and was never happy with the derivation. Pressman uses E/N=Ae(Bm/ton)*10^-8

Billings also uses N/V=ton/(Bm*Ae)

Tony

[Conrad Hoffman]

"For Iron powder cores you can run them up to about 1T (10000 Gauss) but be sure to check the data sheet. If you use a regular ferrite core you will need to make sure the leakage between primary and secondary won't saturate the core when you hit full load since their is no gap in a toroid."
---------------

There have been a few companies offering gapped ferrite toroids, but they typically discover there's no money to be made because manufacturing costs are high, and drop the offering.

I've had some great results with gapped high perm ferrite toroids in some applications. There will be fringing at the gap, but the overall efficiency and resistance to saturation overshadows all else. For inductors, I could obtain Qs far beyond what I could do with even the best iron powder, and still handle significant currents without saturation.

You can gap a toroid using a thin brass shim disk on a motor, and some carborundum power in water. Slow, messy, but easily doable. Or, just get a core you can gap, like a pot core. The necessary gap will usually be very small, like a few thou.

[dtproff]

I'm in agreement with you for the most part. The problem I've had with the gapped torroid (of which I include the metglass cores) is the bad radiated EMI patern I get. If I'm using a ferrite I tend to use the PQ or RM cores.

Excellant comments on the use of gapped torroids. There are almost always more than one acceptable way of solving an issue an it is good to see lively discussions on the issue.

Tony