smps voltage feedback / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

smps voltage feedback

previous 1 2 3

Move Thread

LAN_403

KG7HF

Mon Feb 15 2010, 04:20PM

Registered Member #2683 Joined: Sun Feb 14 2010, 12:27AM
Location:
Posts: 14

I thought that might be the hidden term. Basically, Ray outlined a 10 step process for designing a half bridge (page 161), which I followed.

In a nutshell, the ripple voltage and current were give as design paramaters. In his design example, he set the ripple current to 10%, which I followed.

Step 5 was to calculate the output inductor. I did it three ways, using the equations presented in his book L = V(dt/di), using the method in mag-inc.com manual chapter 4 (really almost the same but based on Toff,Ton), and using the magnetics inductor design software. All methods came out to be about the same inductor size of 8 uH.

Steve Conner

Mon Feb 15 2010, 04:34PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

So, that means (please correct me if I'm wrong Richie! :-/) your supply will be in CCM down to about 10% of the rated load current, and that seems quite reasonable.

Next question.. What's the resonant frequency of your output filter? You're going to need to know this to design your compensator (or identify and steal somebody else's design with a similar resonant frequency to yours...)

GeordieBoy

Mon Feb 15 2010, 05:31PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Yes, if the ripple current is +/-10% of the maximum load current (20% total) then the supply should stay in continuous current mode down to about 10% of full rated load current. (* see below)

10 amps peak-to-peak at 200kHz is a moderate ripple current for the electrolytics at the output to support and the output ripple voltage will be equal to this figure multiplied by the combined ESR. This current ripple combined with the high operating frequency will also cause quite a lot of copper and core loss in the output choke.

* In practice the discontinuous current boundary will typically be at a slightly lower current. At full load current iron-powder looses permeability so you made up for it by adding more turns to get your required inductance of 8uH at 50A and get +/-10% current ripple. Once the load current falls, the permeability recovers and your 8uH inductor becomes a 16uH inductor! This reduces the depth of the current ripple and keeps the converter in continuous current mode for longer.

This swinging choke action of iron-powder is generally a good thing as it delays the onset of discontinuous mode operation. The continuous current / discontinuous current boundary is important because the dynamics of the power stage change when this boundary is crossed. The resonant peak of the output filter also wanders slightly with loading because of the permeability change.

When you ran the supply open loop giving 50V at 50A you probably noticed that the output voltage started to climb abruptly below a certain current as you decreased the load current towards zero. E.g. The output voltage might have started to rise rapidly for loads of less than 5 amps. This is because the converter is now operating in discontinuous mode. The output voltage is no longer just linearly proportional to the duty-ratio.

See here:

For choice of output filter inductor and capacitor parameters also read the info here about ripple current and output voltage overshoot specification:

-Richie,

KG7HF

Mon Feb 15 2010, 07:12PM

Registered Member #2683 Joined: Sun Feb 14 2010, 12:27AM
Location:
Posts: 14

I never calculated the resonant frequency of the output filter, but it looks to me like series LC circuit, so I'll take a stab at f=1/2pi * sqr(lc) = 5.6kHz. Finally get to use my radio knowledge?

I have to back track some on the 10%. When working through the design process, I was worried about finding low enough esr caps to handle 5 amps so I actually changed to 5% The 8 uH is based on 5% ripple current, rounded up. Sorry about that.

I tried to choose a core which held it's inductance over a broad range of currents. The reason is that since this can be used in an SSB, CW, or other variable load application, I wanted to make sure that the inductance stayed as close to the full load as possible.

I've attached some graphs of the inductor I chose. Note though I used 6 oz. copper foil 19mm wide (mixed units, gotta love that!) anyway, 6oz was what I had and it is about the right thickness to support the max skin depth at 400 kHz. I tried to design both the inductor and transformer for only 10C rise. You can see some details in the notes on transformer and inductor building details.

You are correct, actually the rising voltage is exactly the problem I am trying to solve. When I only have my load resistor, the voltage rises to about 60V, but under load it falls to just about 48 volts, which I am very happy with. As suggested I need to see what the "snap back" voltage is. Just my initial observation is that the rise is slow and I think maybe that the rise can be slowed even more by increasing the filter cap size, (wouldn't it take longer to charge?)

Having the discussions is really helpful to the learning process. Like a software designe review! Thanks for the valuable input.

Paul
]filter_inductor.pdf[/file]
]filter_inductor_acperm_graph.pdf[/file]
]filter_inductor_dcperm_graph.pdf[/file]
]filter_inductor_inductance_graph.pdf[/file]

GeordieBoy

Mon Feb 15 2010, 08:54PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

> I have to back track some on the 10%. When working through the design process, I was worried about finding low enough esr caps to handle 5 amps so I actually changed to 5% The 8 uH is based on 5% ripple current, rounded up. Sorry about that.

That's fine. I thought that +/- 10% of ripple current was a little high for a 50A supply anyway. +/-5% ripple current lowers the discontinuous current boundary to 5% of full load too which is good.

The 88% permeability of the inductor at full load current is fine. This is actually quite conservative and you could have let it saturate down to about 70% at full load current if you wanted to skimp on iron. Minimising the inductance swing does make the control stuff a little easier though so that is good.

I still think that 400kHz ripple frequency at the output choke is about 5 times too high though! If you get away with it, it will be because the duty ratio might be low for your application?

I'm guessing that the peak power is 2.5kW but the quiescent carrier power level is only something like 700 watts?

The worst case for the "snap back" voltage overshoot that you mentioned is removal of full rated load from the output. All of the energy stored in the output inductor (0.5 x I x I x L) is transferred to the output capacitor causing it's voltage to rise. If the output capacitor is too small this voltage can massively overshoot the designed target output voltage. What's more, there is nothing that the control loop can do to prevent this situation. Only an over-voltage clamp, (or correct sizing of the output filter capacitor) can limit the overshoot voltage during "load shed".

A similar voltage overshoot situation can occur when the supply is current limiting into a short-circuit and then the s/c is removed from the output. All of the energy stored in the magnetic field of the choke is transferred into the output capacitor. If the output filter is "all choke and no capacitor" the output voltage can overshoot dramatically.

I hope this helps,

-Richie,

KG7HF

Mon Feb 15 2010, 10:16PM

Registered Member #2683 Joined: Sun Feb 14 2010, 12:27AM
Location:
Posts: 14

"...have let it saturate down to about 70% at full load current if you wanted to skimp on iron...."

Yes, I thought that staying above 75% would be OK although I never did see any design criteria which really stated that. One of the deciding factors about the size was to make it a bit more flexible, besides, winding large wire or many small wires on a smaller core might be harder.

"I'm guessing that the peak power is 2.5kW but the quiescent carrier power level is only something like 700 watts?"

It really depends on the mode. SSB and CW might be 50% duty cycle, RTTY and digital modes such as wsjt are more along the lines of 90% duty cycle, AM and FM are 100% duty cycle. But of course not in a continuous duty, there are transmit and receive periods. WSJT is sometimes 30 seconds xmt, 30 seconds rcv.

I may have overbuilt somewhat, but if we assume 65% efficiency of the RF amp, producing 1500 watts will require 2300 watts of power from the supply I figure I need some overhead for loss in the power supply itself too. I think by keeping the permeability a little higher and the power transformer flux density a little lower then normal as well as having sized things for smaller temp rise I should have some decent margins. although it will be heavier, and more expensive, it's still less weight and cost than a comperable linear supply.

BTW, another smps designer I know also agrees with you about the 200 kHz switch rate, although he did say he had done some higher power converters at a higher switch rate as well but ususally doesn't.

Oh well, it's a learning experience. I do have a really big heat sink on the FETs (I think it is big) aluminum 6"x4" with 2" fins every 0.5". I could also mount the FETs to a 1/2 copper heat spreader then mount that to the al, but I haven't monitored it long enough to determine the need yet. (running for a minute or so at full load doesn't quite count.)

previous 1 2 3

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob