Boost Converter Noisy Output

[Angry Badger]

Hi,

Nice forum, my first post.

I’m attempting to build a boost converter using an LM2577.

**broken link removed**

Components: Cin1 = 220uF
Cin2 = 0.1uF
Cc = 3.3uF
Rc = 820r
L = 1000uH
D = 1N5181
R1 = 91K
R2 = 4.92K
Cout1 = 1000uF (low esr)
Cout2 = 0.1uF

Vin = 9V
Vout = 23.9V @ 95mA

No load current (from supply) = 9mA
Full load current (from supply) = 286mA

The voltage is almost spot on and holds up under the design load (100mA). However there are a lot of spikes on the output, I cannot get nice clean waveforms as shown in the National Semiconductor data sheet. I’ve fiddled with all the component values but they don’t make much difference. I originally built the circuit on a plugblock, only soldered everything together because I thought that might have been the problem.

**broken link removed**

You can see the spikes / ringing in the above screenshot. Below are the nice and clean National Semiconductor test waveforms. They are using a fixed 12V switcher.

**broken link removed**

I'd appreciate any advice about eliminating the spikes.

 

[Ubergeek63]

Well for starters you can replace the junk HV rectifier with a 40V schottky and tightening the power loops. A couple X7R ceramics to keep the HF loops as small as possible...

 

[Angry Badger]

Hi,

Thanks for the reply. I made a typo in the post - the diode is a 1N5818 Schotkky type.

Anyway, I've somehow managed to fry the regulator. I was fiddling about and there was a flash followed by smoke from the small tantulum used as Ccomp.

I'll have to order another and will also get some ceramic caps as you suggested. Where will these be placed, are they just decoupling caps across the supply? Any particular values?

 

[Speakerguy]

The ringing can be greatly reduced by the use of an RC snubber.

Measure the circuit with no load. Make note of the frequency of the ringing after the pulses.

Add a capacitor right at the output pin of the regulator, right before the inductor. Nothing else should be between this node and ground. Adjust the size of the capacitor until the frequency of the resonance is reduced by a factor of two.

Your parasitic capacitance is 1/3 the second value. Call this Cp.

Find the characteristic impedance of the line by solving for it (let's just call it Rs) :

Rs = sqrt(L/Cp)

Put an RC snubber right at the output pin with the resistor = to Rs and a capacitor 5-10x the value of Cp. This will greatly damp the initial pulse and the ringing will almost disappear. The downside is you need a few values of capacitance to have around to play with, you have to be able to measure it, and the parasitics in your ckt are probably huge on a breadboard, which means bigger C and more power dissipation in the Rc. But the output will look nicer.

 

[Roff]

You need short leads, tight layout, and a good ground plane. Use wide traces where the current (including transient current) is high.

 

[kchriste]

All good advice above. Also check your scope probe leads by connecting the probe tip to the same point as the probe ground clip and verify there are no spikes showing on the scope.

 

[RadioRon]

Roff's advice is, in my opinion, the most critical. The ringing is the result of current finding a wide array of parasitic inductances and capacitances in all those long skinny connecting wires. If you reduce these parasitic reactances first, then you will reduce the ringing and should you choose to add a snubber circuit, it won't have to work so hard. In this case a snubber is possibly a bandaid fix and may be unnecessary if you attack the root of the problem.

When the switch inside the IC (the one from pin 4 to GND, pin 3) switches on and off, the change in current is almost instantaneous and such a change has an extremely broad bandwidth. So if you want to get nice rectangular waveforms without ringing, you have to treat the critical current loops as high frequency (ie. radio) circuitry and connect them as if you were hooking up a 100 MHz amplifier, that is with extremely short, wide conductors and over a ground plane so that all connections have low characteristic impedances.

There are two critical current loops, the path of current when the switch is ON and the path of current when the switch is OFF. When the switch is ON the high frequency current follows a path through the input capacitor (the one from pin 5 to ground), the inductor, and then through the switch inside the IC. It comes out of the IC GND pin 3 and then finds its way back to the input capacitor to complete the loop. In this case the best layout puts the ground side of the input capacitor directly connected to the ground side of the IC. This is the shortest path. Also, make sure that the input capacitor hot side is directly connected to the input side of the inductor, and make sure the output side of the inductor is connected as close as possible to pin 4 as possible.

When the switch turns OFF, the high frequency current path is now from the input capacitor, through the inductor, through the rectifier and finally through the output capacitor to ground. From the output capacitor ground it then travels back to the input capacitor ground to complete the loop. To optimize this loop, make sure that the rectifier is connected directly to the inductor output with the shortest possible leads. The output capacitor then connects to the rectifier directly and with the shortest possible leads. The ground side of the output capacitor should connect directly to the ground lead of the IC with the shortest possible connection, which is in turn directly connected to the ground side of the input capacitor. The amount of metal that exists at the node connecting the inductor output to the IC pin 4 and to the rectifier is the most critical node in the whole circuit as far as ringing goes. You have to keep the parasitic capacitance of this node at a minimum, so make these connections short and the amount of total metal small.

When I say "shortest possible" connection, I mean that you should push each component down onto your perf board so that the leads are short. Resistors should be laying down, not standing up. Part connections underneath should be 0.2 inch wide pieces of copper tape (wider is better), not wires, especially if the connections are longer than about an eighth of an inch. The exception to this is the node at pin 4 which should have all three components directly conneced with no unnecessary "traces" or wire length. Its hard to make a ground plane on perf board, so direct connections of ground terminals should also be copper tape, quarter inch wide or more.



Consider rebuilding the circuit on some hand cut copper plated board (double sided). The easy way is to cut islands of copper on one side, one island for each junction and then surface mount the components to these islands of copper. You can cut such a board with a sharp knife in short order, if you plan it carefully to minimize the number of cuts. Ground connections are made by drilling holes only for ground connections and using the other side of your board as a ground plane. This sort of thing is easier and smaller if you use surface mount components, but leaded parts with the leads cut very short are OK too.

This construction approach reduces interconnect inductance, lowers characteristic impedances of connections, lowers node parasitic capacitance, and ultimately reduces the ringing.

It should also be said that R1 and R2 must also be placed close to the IC so that the feedback voltage at pin 2 is a faithful copy of the output voltage.

 

[blueroomelectronics]

I noticed you have a OWON scope, I've been looking for reviews/opinions on them. How do you like it?

 

[eng1]

Components:
Cin1 = 220uF
Cin2 = 0.1uF
Cc = 3.3uF
Rc = 820r
L = 1000uH
D = 1N5181
R1 = 91K
R2 = 4.92K
Cout1 = 1000uF (low esr)
Cout2 = 0.1uF

Vin = 9V
Vout = 23.9V @ 95mA

Have you tried running a Webench simulation on the National website? For the desired output voltage and nominal current, setting Vin min = 6.5 V for example, they suggest a 114 uH inductor (you should choose a power inductor with suitable current ratings, the peak current is about 0.76 A). They suggest using two 22 uF capacitors at the output. You should be using the adjustable version of the LM2577.
As already explained, the layout is a critical part of the design.

 

[Ubergeek63]

silly me... I made the first comment assuming he had selected components based on the easy switcher nomographs

 

[Angry Badger]

Hi,

Just a quicky to say thanks for all your replies. I'm carefully studying all the points you made. I'll let you know how I progress in due course. By the way I selected my components through the example calculations in the data sheet.

Blueroomelectronics:I'm so far pleased with the Owon scope although I have to point out that I'm not a seasoned scope user (this is my first - purchased off ebay a couple of weeks ago). It only took me an hour or so to master most of the controls/functions and that was without recourse to the manual. The build quality seems ok although I expect it's not in the same league as a Tek' or Fluke etc but then one gets what one pays for. I think it's a good purchase for the amateur, like myself.
I can't really find fault with it at the moment. It would be nice to be able to change the colours of the waveforms to suit oneself, they are red and yellow by default. The software is ok but you cannot paste directly to a wordprocessor as far as I can tell.
There is a battery supply facility but a battery is an optional extra. The max input is 300V pk so I cannot measure our mains voltage (340v pk). The supplied usb lead is rather short at about 65cm so the scope needs to be pretty close to your pc to transfer data, unless you buy a longer lead.
If you wish to know anything else please ask.

Regards,

Jeff

 

[geko]

I noticed you have a OWON scope, I've been looking for reviews/opinions on them. How do you like it?

Hi Bill,

I bought a Owon PDS6042 a month or two back.

I was in two minds whether to try and pick up a Tek TDS210 off E-Bay as these seemed to be going for a similar price to the Owon scope. In the end I went with the Owon because it has warranty, includes scope probes, and interfaces to the PC using USB.

It is styled like a Tek TDS200 series scope and the manual is almost a copy of the same. This is reflected in the fact that where the TDS210 has a 'Hardcopy' button, the Owon has a button labled 'Not used'. You can of cause use the PC software to get a hardcopy but you get the idea


Some things Owon hasn't got that the Tek has are:

No trigger hold off
No trigger view
No bandwidth limiter
No fine adjustment on the vertical Volts/Div
Vertical sensitivity is 5mV to 5V (2mv to 5v on Tek)

There are probably some others I've missed but they're the main ones I've noticed although non are show stoppers.

The display use a STN colour LCD not TFT so anyone who can recall the early laptop STN displays will know what I mean when I say 'limited viewing angle' Display contrast is adjusted with a thumbwheel pot below the display. It's not a great display if you compare with a good TFT but it's adequate for the intended purpose.

The control legends printed on the plastic case look a bit cheap but the controls themselves are okay.

It can interface to a PC through a USB port (no RS-232, it has blanks where 9/25 way D connectors would go) The software is fairly basic but it does let you capture data and has a real-time mode where the PC functions as the display.

I like its compact size and light weight. The cursors, measurements, and not having to multiply by ten when using a 10x probe makes life so much easier. Being able to capture one-off events like IR remote signals and pre-trigger capture are great facilities to have too.

Given that I've managed with my old Hameg 20Mhz dual trace scope that I bought with one of my first pay packets 25 years ago I'm quite pleased with the Owon. It's more than adequate for my hobby use and comes at a price I can justify.

I've put a few photo's of the scope in use here **broken link removed**

Pete