[SOLVED] Oscillator frequency question

Hi,

Am learning a little about SEPICs, the attached circuit is a 2MHz oscillator for the PWM. Question is why does the frequency change from 2MHz at CD4049 output to 6MHz at MOSFET output, then goes back to 2MHz at triangle wave generator output?

The same thing happened when the P-Channel MOSFET was connected to an NMOS at a previous stage of experimenting, done to understand a few things about buffering the oscillator/crystal (reason for three inverters in series).

Thanks for any input.

You no significant load (102k) on the drain and circuit inductance appears to resonate on the 3rd harmonic.

Try a resistive load from drain to ground knowing that Coss=45pF to give rapid decay to 0 when turned off.
e.g. EST. RC=45ns then R=1k.

When R=100k C=5pF= 0.5us approx then the 2&6MHz pulse integrates and results in 0.5us sloped triangle wave at 2MHz

Consider a bipolar half bridge driver. Actually your Pch FET offers no benefit as 74HCxx logic is lower average impedance than a single switch With no load.

Hi,

From your description I assuume you don't have a solid GND plane, maybe no PCB at all...
Is it built on a breadboard?

Klaus

Suggest to monitor the waveforms with an oscilloscope rather than "frequency" with a frequency counter.

Hi Sunnyskyguy,
Thanks, a harmonic, makes sense, I'll try a resistor from drain to ground, and as you say the PMOS serves no purpose, I'll try with an op amp, which the 2MHz is supposed to go into.

Hi Klaus, yes, this is just on a breadboard, I had no success in finding out how to simulate the oscillator on the TI Tina free version. I don't mind so much, as I get to see the real thing and take some measurements.

Hi FvM, I did check the waveforms with a simple oscilloscope (just a black and white LCD small screen, not too satisfactory but better than nothing), the PMOS outputs a dirty/messy square wave, the 4049 and the RC part output a triangle wave that is not neat, slightly rounded at the peaks.

I understood the crystal on the 4049 doesn't like capacitive loads, can I connect the third stage of the 4049 directly to an op amp?

Hi,

Breadboard....

If you post a picture of your breadboard wiring, then maybe we can give recommendations for better signals.

Klaus

A further problem in the original schematic is the ZVN2106 is an N-channel MOSFET so connected as it is, the substrate diode is always conducting. It would make more sense to omit the MOSFET and 1K resistor completely and use the output of the CD4049 directly to the filter.

Brian.

Hi,
I tidied up the breadboard this evening to swap the "antenna loop" jumpers for short flat cables, and trimmed leads etc., which I hate doing by now as the components used to make a tidy breadboard circuit end up useless for a lot of other things, and it rarely seems to make any difference to circuit operation when you're just testing out a rough functional circuit, but anyway... it made no difference to how the oscillator works, so I've omitted the photo.

Also, I found that by removing the PMOS, the oscillator stops oscillating, don't know why but for the time being I'll stick with something that works so I'll leave it in. Also discovered that the double low pass filter is unnecessary, one stage is enough, copied it from a webpage that admitted the writer was not too advanced theory-wise.

The ZVN2106 is the P-Channel, so that bit is right, the ZVN4206 is the N-Channel device.

I'd like to understand why two of you say the MOSFET is unnecessary, but without it the oscillator stops oscillating, I started this circuit about two months ago then had to put it to one side, and had forgotten that the PMOS is to boost the 4049 2.18V output voltage to ~4.5V, it seems to serve no other purpose to me, but in my ignorance I'd have said that it could be acting as a buffer between the 4049 and the filter, which is only ~50pF, yet the fact that the PMOS has 100pF input capacitance throws me.

20pF is about 5k at 2MHz so 47k attenuates too much so reduce to 4.7K
Then remove FET

Hi,

Breadboard:
It's not only a question of long cables.
In your case I assume it's just a question of choosing the correct GND reference points.
I assume it's rather wired coupling than air coupling, because you work with relatively low frequencies.
6MHz needs a wire length of more than 10m to be a good lamda/4 antenna.

Klaus

I'm confused. According to Zetex and Diodes Inc, both the ZVN2106 and ZVN4206 are N-channel devices.

In the schematic, only the first section (on the left) is the oscillator, the following two inverters are acting as buffers so the filter should be well isolated anyway. Either there is little or no supply decoupling, the wiring is really bad or the component values are not what they say. The 1K resistor serves no purpose at all except to increase the current it draws, the gate capacitance of the MOSFET (if used) should discharge quickly enough through the inverter output stage.

Brian.

betwixt said:

I'm confused. According to Zetex and Diodes Inc, both the ZVN2106 and ZVN4206 are N-channel devices.

Click to expand...

...Apologies Brian, too late I saw my spelling mistake, really sorry about that, it seemed completely incongruous that you'd give incorrect info... ZVP2106 was what I meant...

betwixt said:

In the schematic, only the first section (on the left) is the oscillator, the following two inverters are acting as buffers so the filter should be well isolated anyway. Either there is little or no supply decoupling, the wiring is really bad or the component values are not what they say. The 1K resistor serves no purpose at all except to increase the current it draws, the gate capacitance of the MOSFET (if used) should discharge quickly enough through the inverter output stage.

Click to expand...

Okay, I'll try taking the gate resistor off tonight/as soon as I get a moment. Normally only use 11uF supply decoupling for all breadboard circuits and seem to be okay, will see if adding more helps.

- - - Updated - - -

SunnySkyguy said:

20pF is about 5k at 2MHz so 47k attenuates too much so reduce to 4.7K
Then remove FET

Click to expand...

I see, thank you, will also give this a go when tonight or time permits.

Hi,

Normally only use 11uF supply decoupling for all breadboard circuits

Click to expand...

11uF.. quite an unusual value. But it seems to be an electrolytic one. If so, they often are not suitable for the high current transients on power supplies. I recommend to add fast ceramic capacitors.
Maybe 100nF ones.

Klaus

You using the CD4049ub

That is an unbuffered device, excellent for class "A" applications (like the crystal oscillator itself) but with low drive capabilities.
You still have three unused inverters in the hex package. Parallel them with the very last inverter. That way you have 4X the drive capabilities. Then get rid of the external mosfet.

Also, I'm a little curious. What does this circuit have to do with a SEPIC converter, anyways?

Hi,

You still have three unused inverters in the hex package.

Click to expand...

The schematic says "1/3", so I tought there are three inverters ... all used... --> but now: confirm that no inputs (especially of the unused inverters) are left floating!

Klaus

Hi,

I haven't had time today to make use of the good advice, I hope tomorrow evening will provide free time to do so, I miss (making a mess of) my circuits a lot lately.

Normally use 10uF electrolytic, 1uF polyester, and 100nF ceramic decoupling on breadboards, with an occasional 100nF polyester as well. I have very, very few 10uF X7R ceramics, and am trying hard not to squander them as they are needed for the SEPIC capacitors. If I understood well, Panasonic 10uF polyester are not suitable for SMPS, maybe I got that wrong. No floating inputs on the CD4049ub!

"You still have three unused inverters in the hex package. Parallel them with the very last inverter. That way you have 4X the drive capabilities. Then get rid of the external mosfet."
- That's great to know, thank you. No-one approves of the MOSFET it seems...

"What has it got to do with a SEPIC?" - I wanted to attempt to make the PWM circuitry, as I've had great difficulty researching and piecing together the information for the control circuit and gate drive (error amplifier and PWM op amp before Totempole? gate drive) as the SEPIC part itself seems to be - in principle and based on the calculations at least - simple-ish/simple enough, it's sitting on the breadboard unpowered waiting for the control circuit to be attached.
I'm wondering if I shouldn't have done the oscillator circuit using two op amps for the triangle wave instead, but the 2 MHz means I can use two 50uH, 1A inductors I have and the switch MOSFET won't go over its max. PD, and I had a couple of crystals spare.

I want to learn as far as my ability allows to make a basic SMPS, and SEPIC seems a good choice for a DC-DC converter as it covers bucking and boosting input voltages, notion was power supply from a 6V battery (4x1.5V) to go down to 4.8 but output at constant 5V, or 3x2V solar panels in series. Motivation is learning about SMPS.

I have a terrible confession to make, and a wiser person would not admit it: The 6MHz seems to have stemmed from an N-Channel MOS and LED which were also connected to the PMOS output, I didn't put that in the schematic as it seemed irrelevant/I assumed the PMOS was the cause, and it wouldn't have been part of the actual circuit anyway. When I removed the NMOS the PMOS output went down to 4MHz. Sorry about that dreadful omission.

Hi,

When I removed the NMOS the PMOS output went down to 4MHz. Sorry about that dreadful omission.

Click to expand...

Why 4MHz? Why not 2MHz?
Please confirm.

Klaus

2MHz PWM is that what you want?

KlausST said:

Why 4MHz? Why not 2MHz? Please confirm.

Click to expand...

Hi, I do not know why it does that, all I can see is that each additional MOSFET adds 2MHz more to the DMM frequency counter. Maybe it's something to do with harmonics as Sunny said. The crystal/inverter frequency is 2MHz, the PMOS added 2 more, the NMOS in series with the PMOS and in parallel with the filter (the filter in series after/from the PMOS) added another 2. Presumably the sooner the PMOS is removed, the better, so I'll try what schmitt trigger suggests about paralleling the remaining 3 inverters on the IC.

- - - Updated - - -

SunnySkyguy said:

2MHz PWM is that what you want?

Click to expand...

Hi, yes. It seems a somewhat high a frequency comparing to the frequencies I see in datasheets and in SMPS/SEPIC app notes I read, but also understand that such frequencies of PWM are used in DC-DC converters. The components I have would fit with that frequency, otherwise I would have used 500kHz to 1MHz. I have to fit the circuit to the components to hand rather than enjoy selecting new stuff to fit a lower frequency, e.g. first calculations were for 1MHz CCM, and only a power MOSFET like IRF520 in my little store of components would cope with the PD, and had least bad RDSon of the rest, but VGSth of 4V at 250uA not going to work, and needed inductors of ~60uH so I went for 2MHz to fit the ZVP2106 (1 Ohm, RDS on, apparently...), and the 50uH inductors. Pretty grim McGyver/A-Team circuit design approach...

I'm becoming aware that higher frequency = lower amplitude, so the triangle wave is very small Pk2pk from what I'm seeing, and looks like it will (probably) need amplifying.

OK, I understand now...you want to design and build each of the building blocks in a SMPS controller to learn how they work.....this is good, congratulations!

I then will provide you with the best advice I can, which is: Buy, borrow or beg for an oscilloscope.
You want to understand a circuit, then you must understand what is going on. You need a scope to do so.

For instance, your paragraph "...all I can see is that each additional MOSFET adds 2MHz more to the DMM frequency counter..." does not make sense. Most likely there is ringing, but without actual waveforms, all are guesses.

You think you have problems right now with a simple oscillator stage? Wait until you actually start switching some real power. You WILL really require a scope then.

BTW, if this is your very first SMPS you design and build, I would suggest to start with a far lower frequency, let's say 100 Khz or perhaps even lower. For a number of reasons, the possibilities of success will grow incredibly.

Also..........You don't require crystal-accuracy for SMPS work. A simple RC oscillator is enough.
Nevertheless if you still want to design a high stability oscillator, then a very common 32.768 Khz tuning fork crystal would fit the bill perfectly.