[SOLVED] Oscillator frequency question

[d123]

Hi,

My SEPIC doesn't work, sometimes the voltage is too high, sometimes the voltage is too low, why?
(see attached image)

Sorry, just a little joke there. I am having a lot of problems getting the SEPIC to regulate the output voltage to 5V. I followed the advice of using a lower frequency and chose 100kHz, then dropped it to 90kHz to fit the 680uH inductors better.

The output is always between 2.2V or approx.4V whatever the load. The intended load should be 100mA, this very bad SEPIC of mine cannot even handle 4mA, it would seem.

I've been through anything I think that could be a problem more than once, and may hit the breadboard with a hammer, in case it helps:

It doesn't seem to be the input/coupling/output capacitor values.
I can't see the inductor current waveform, unfotunately.
Lowering the frequency to 50kHz or even 30kHz makes little difference/improvement.
The Vref is good at 2.48 - 2.49V, the divider quite accurately divides 2.440V into 1.221V.
FOsc is stable at 90.1kHz at measured points.
The error amplifier is on when Vout is below (Vref) 5V and off when above (tried it with no load at 30kHz and got approx 7V out).

Questions: Should the error amplifier and Vref be picking up the 90kHz, or could that be the DMM frequency counter misreading, as the oscilloscope shows no such thing?
Why could I be getting very random and varying frequency values on the breadboard? Is that because it's a breadboard?
If the SEPIC is never at 5V, and much less with a 56ohm load, is it probable that I have got the calculations wrong?

Thank you.

Attached is the snva168e where I got the calculations from for CCM, the measured results from last night, the schematic, and a document with a photo of the breadboard, the oscilloscope screen (empathise a little...) and how I'm measuring things (directly in the adjacent breadboard hole or directly on the pins of interest).

 

[KlausST]

Hi,

A simple test:
On your breadboard you have several rows of "GND".
Connect the scope GND to the lowest GND row of your breadboard.
Connect the scope input to one of the upper row of GND, near the inductances.

Adjust the scope timing to see about 1..3 full waves.
Adjust the voltage scaling and offset to see the peak to peak voltages.

I assume the voltages are well above some 10mV. (I indeed guess it's above 500mVpp, deoending on switching frequency and current)

********
In every switch mode IC datasheet you see PCB layout considerations. They talk about GND plane. Making the voltage drop across the GND plane lower than some mV. Maybe you think you can reduce voltage drop by using thick wires...it will fail. Because -about independent of wire thickness - every mm of length cause an inductivity. With a breadboard you easily get hundreds of nH.
Lets say 200nH at 100kHz means XL of at least 100mOhms. But this is only for the fundamental frequency. For clean switching you can calculate with the 50th overtone....making XL in the range of some ohms.
Every 100mA will case voltage drops of several hundreds of mV. It will decrease regulator stability.

Additionally those indutance may form an LC oscillator. Therefore you may expect heavy ringing. Making stability about impossible.

--> if you really want to experiment on a real circuit you need a copper foil, a copper sheet or similar as GND plane. Then solder all devices (connected to GND) directely to the copper..with short wires. And then solder all the other parts to these devices.
I admit, it doesn't look nice, more like a coloured spider web, but the results are way more reliable.
Just an example: https://www.w8ji.com/images/Amplifier/demons2.jpg

Klaus

 

[FvM]

Why are you using a hysteretic (on-off) controller instead of a variable pulse width PWM controller? It will never achieve continuous operation with stable frequency and pulse width.

To modify your controller design respectively, you would use a feedback amplifier with PI characteristic and a compare it's output with a sawtooth or triangle oscillator output. Or implement current mode control like LM3478 in the TI application note does.

 

[d123]

Hi, thanks for your input and advice.

There's a world of difference between a simple "static" circuit that'll work on a breadboard and an SMPS I see.

I assume the voltages are well above some 10mV. (I indeed guess it's above 500mVpp, depending on switching frequency and current)

340 - 400mV depending on which breadboard rail the scope input is in...

Maybe you think you can reduce voltage drop by using thick wires...it will fail.

Not sure if you mean "you" general or "you" meaning me, just for the record, I use 0.5mm for breadboard jumpers as it snaps less often than 0.25mm, but you've reminded me of some rather important aspects I forget to bear in mind.

Because -about independent of wire thickness - every mm of length cause an inductivity. With a breadboard you easily get hundreds of nH.
Lets say 200nH at 100kHz means XL of at least 100mOhms. But this is only for the fundamental frequency. For clean switching you can calculate with the 50th overtone....making XL in the range of some ohms.
Every 100mA will case voltage drops of several hundreds of mV. It will decrease regulator stability.

Additionally those indutance may form an LC oscillator. Therefore you may expect heavy ringing. Making stability about impossible.

Thanks, none of that even crossed my mind, and some points you make I didn't even know.

So, I need to do dead bug style prototyping for SMPS circuits... great, the whole point of breadboards is that I don't waste components on several incorrect versions, not to worry, it's worth giving a go as what you say makes a lot of sense, I like single-sided boards as I get to make an uninterrupted ground plane.

In your opinion, would it make any difference starting by trying to make a buck or a boost converter? i.e. Are they in any way "easier" than a SEPIC, or do they equally have their own issues for people with basic skills/knowledge, so I may as well just try to do the SEPIC, which seems a useful one to know how to make?

Thanks a lot, Klaus

Regards,
Daniel

- - - Updated - - -

Why are you using a hysteretic (on-off) controller instead of a variable pulse width PWM controller?

Because I've had to piece together what I understand from different app notes and similar documents, beyond "error amplifier" and "comp" amplifier/comparator I have little idea how to fit these things together or what they're supposed to do, and evidently still haven't understood a lot of it. I'm not being sarcastic. There seem to be many overview descriptions that make a lot of assumptions about what the reader understands, but few detailed ones I've found for beginners about the control circuitry explained in simple terms regarding one or two points.

It will never achieve continuous operation with stable frequency and pulse width.

Thanks, that would explain why the circuit never has a 5V output, much less at 100mA.

To modify your controller design respectively, you would use a feedback amplifier with PI characteristic and a compare it's output with a sawtooth or triangle oscillator output. Or implement current mode control like LM3478 in the TI application note does.

No, not current mode control, thank you, I'm avoiding that, that involves right-hand-plane zeroes and literature says they are harder to control than voltage-mode, if I'm not getting the RHPZ thing mixed up, but I remember that voltage-mode is the beginners "easy" option.

Im sorry - Does "PI" refer to "PID", I have only briefly read about it. I'd wondered about the need for a sawtooth into the comparator as in the end the MOSFET sees a square wave. When I tried to put a low pass filter on the 555 output the circuit stopped working as far as I remember, so I tried a totempole then a low pass filter to get the triangle and the same thing happened, so tried to omit the triangle wave.

That's another doubt I've been having and have found no explicit description for: Can you say what voltage the sawtooth needs to be at in comparison to the EA output, if that's how it works?

I have an XR2209, which can output a triangle wave, I can use that for the triangle wave, can't I? (having said that, I need to re-read the datasheet in case it's only good up to 5V max.).

If possible to compare, and not pointless, and you're familiar with all three, having a 4049UB, XR2209 and an LMC555 that can operate up to 3MHz, which would be a better choice to generate the triangle wave? Or is it much of a muchness/comparing apples and apples?

Thanks for the reply and clarifications, FvM, has helped with floundering in the dark with this circuit.

Regards,
Daniel

 

[FvM]

There are reasons why flyback and SEPIC controllers use mostly current mode control, but yes, voltage mode should basically work. Other than with current mode, you don't get simple built-in overcurrent protection.

The sawtooth generator + comparator is a classical pulse width modulator topology which can be found in many voltage mode controllers like 3524. It sets a defined gain output voltage variation / error amplifier signal. The error amplifier is usually designed as PI or PID. The "type II" and "type III" error amplifier discussed in switched mode design text books are modified PI respectively PID circuits.

 

[KlausST]

Hi,

There's a world of difference between a simple "static" circuit that'll work on a breadboard and an SMPS I see.

Very true.
******

"You" in general, so "you" are included, too. ;-)
******

SEPIC, BUCK, BOOST.... all the same

Klaus

 

[d123]

Hi,

I have been trying to get to grips with type III compensation calculations ("cumbersome and iterative" as one app note stated, how true...), and see I need to understand transfer function, how to draw a bode plot and other concepts before the compensation calculations will mean anything valuable/usable...

-------------------------------------------------------------------------------

Main question: I've been trying to put together a suitable sawtooth circuit, and am having several difficulties as far as I can see from this picture:


As this is about as detailed a description as I've found across several/multiple app notes and the like, I understand that the EA Voh needs to be let's say about two thirds of the sawtooth amplitude. Is that assumption correct, please?

I've tried several sawtooth circuits, some disappointing (probably my fault), some okay, but I'm finding all seem to be capacitor based, which means the output voltage will never be over two thirds of the supply voltage. That's a problem, I think, as the EA I am using is a rail-to-rail device, and for VCC 4.76V has Voh ~4.5V and Vol ~100mV, but all the sawtooth generator's I've tried output about 2.8V max, and perhaps worse never reach 0V for Vol, they range from 100mV to 400mV and some seem to only reach mid-supply for Vol and max Voh 3.95V.

Some examples:




The 555 + NPN is surprisingly "good", but has the 0.3 low to 2.8 high range.
Haven't tried the Schmitt Trigger one, I'm guessing it also only reaches 2/3rds of VCC?
The OpAmp one pictured here, couldn't get it to work, don't know why.

a) If I have to use the 2.8V max version, can I use a divider on the EA output to scale it to the right amplitude? Or not?

b) Are there sawtooth circuits that are not capacitor based?

c) Is there some way of amplifying the sawtooth to get it to swing from rail to rail, without degrading the waveform? ...Is that idea basically adding a bad solution on top of an already bad basic design/circuit?

I've attached the documents I got the circuits from, in case anyone wants to peruse them.

Thank you.