Sawtooth ramp generator 500kHz

[tvgste]

Hi there

In order to learn more about power electronics, I am working on my own simple 500 kHz buck converter. I am done with the basic design, components, input filter. gate drive and feedback control with a type-iii compensator. Works like a charm. However,  I am missing a 500 kHz ramp generation and I want to have this included as well instead of hooking up my function generator.

How can I  create a 500 kHz sawtooth ramp with basic components? It doesn't need to be perfectly linear, it should just be monotonic and smooth from 0V - 4V (with e.g. a 5V supply). What would be that state-of-the-art approach to tackle this problem? What about using a triangular-wave?

I am looking to any reply or fruitful discussion 

Cheers,
Dave

[moffy]

A current source feeding a capacitor with a comparator, sensing the capacitor voltage,  controlling a switch across the capacitor e.g. 74LV07, is one possible way.

[Konkedout]

This buck converter...will it be DC to DC?

The poster above mentioned what I was thinking...but I would use a small logic level MOSFET (2N7002?) to discharge the timing capacitor.

The comparator should be moderately fast and with active pullup, such as TLV3201.  You also need a bit of AC coupling from comparator output to the input to make sure that the capacitor gets slammed down very close to ground with each cycle.

Relax and start with a relaxation oscillator similar to this  https://www.ti.com/lit/ab/snoa998a/snoa998a.pdf?ts=1732381495131

But use the 2N7002 to discharge the FET.  Then you will need a resistor in series with the comparator - input and maybe a 200 nSec RC coupling the - input to the comparator output.  If you draw it and I remember  to return here, I can recommend tweaks from that point.

[jonpaul]

Suggest to read and learn before you hop in and build.

Do you have any books and papers on SMPS/driver design?

I would first choose IC controller, eg old PWM type, SG3525 etc.

Internal ramp gen, comparators transistor drivers.

After that you can DIY a discrete design,

Just my thoughts,

Jon

[PCB.Wiz]

How can I  create a 500 kHz sawtooth ramp with basic components? It doesn't need to be perfectly linear, it should just be monotonic and smooth from 0V - 4V (with e.g. a 5V supply).

It does not need to be 4v, as the compare is done using a offset amplifier.
If you make the peak < 50% of the supply, you are close enough to linear on a normal RC charge, for SMPS use.
If you really want a current ramp, you can make one using a LED and transistor and resistor, but it will have more variation in frequency than a simpler RC.
Or you can use a matched pair PNP and make a current mirror.

[Someone]

If you make the peak < 50% of the supply, you are close enough to linear on a normal RC charge, for SMPS use.
If you really want a current ramp, you can make one using a LED and transistor and resistor, but it will have more variation n frequency than a simpler RC.

Have you tried to make saw or triangle waveforms for modulators at these sorts of x00kHz rates? Tricks used in silicon to get around linearity and switching issues fall apart when trying to do it discrete. nano henries of interconnect and picofarads of coupling start being problems.

[PCB.Wiz]

If you make the peak < 50% o the supply, you are close enough to linear on a normal RC charge, for SMPS use.
If you really want a current ramp, you can make one using a LED and transistor and resistor, but it will have more variation n frequency than a simpler RC.

Have you tried to make saw or triangle waveforms for modulators at these sorts of x00kHz rates? Tricks used in silicon to get around linearity and switching issues fall apart when trying to do it discrete. nano henries of interconnect and picofarads of coupling start being problems.

That's why keeping to the simple end of thigs is a good idea.

A useful part for a 500kHz sawtooth generator, would be a single gate open drain schmitt part. (eg Nexperia 74LVC1G07)

You use two RC's around this, a charging RC on the output pin,  that sets the ramp time, and a flyback RC to the input pin, that ensures the open drain discharges all the way to GND, whilst the IP is still within the hysteresis band.

A single RC will also give a sawtooth, but limited to an amplitude slightly larger than the hysteresis band, due to device delays.

[Someone]

If you make the peak < 50% o the supply, you are close enough to linear on a normal RC charge, for SMPS use.
If you really want a current ramp, you can make one using a LED and transistor and resistor, but it will have more variation n frequency than a simpler RC.

Have you tried to make saw or triangle waveforms for modulators at these sorts of x00kHz rates? Tricks used in silicon to get around linearity and switching issues fall apart when trying to do it discrete. nano henries of interconnect and picofarads of coupling start being problems.

That's why keeping to the simple end of thigs is a good idea.

A useful part for a 500kHz sawtooth generator, would be a single gate open drain schmitt part. (eg Nexperia 74LVC1G07)

You use two RC's around this, a charging RC on the output pin,  that sets the ramp time, and a flyback RC to the input pin, that ensures the open drain discharges all the way to GND, whilst the IP is still within the hysteresis band.

A single RC will also give a sawtooth, but limited to an amplitude slightly larger than the hysteresis band, due to device delays.

Sounds like something that would work with spherical cows in a vacuum (simulation) but then fail in the OPs stated application. The linearity of the modulator in a switching regulator has to be rather tight to avoid instability. Triangles are hard enough at that frequency, let alone a sawtooth. Do you have some examples of tested circuits that produce 500kHz sawtooth waveforms? and what errors they exhibit compared to "ideal" ?

[PCB.Wiz]

Sounds like something that would work with spherical cows in a vacuum (simulation) but then fail in the OPs stated application. The linearity of the modulator in a switching regulator has to be rather tight to avoid instability. Triangles are hard enough at that frequency, let alone a sawtooth. Do you have some examples of tested circuits that produce 500kHz sawtooth waveforms? and what errors they exhibit compared to "ideal" ?

Since you are the self appointed expert, we'll wait to see your pearls of wisdom here. 

[Someone]

Sounds like something that would work with spherical cows in a vacuum (simulation) but then fail in the OPs stated application. The linearity of the modulator in a switching regulator has to be rather tight to avoid instability. Triangles are hard enough at that frequency, let alone a sawtooth. Do you have some examples of tested circuits that produce 500kHz sawtooth waveforms? and what errors they exhibit compared to "ideal" ?

Since you are the self appointed expert, we'll wait to see your pearls of wisdom here. 

Having built analog modulators in that frequency range and knowing the control theory behind switching regulators, my answer was simple: dont

Getting something useable for the purpose of the OP is not simple or small or cheap if it has to be built from scratch. So much easier to use an existing chip with the function integrated even if the other parts are ignored/disabled.

You started off on claims about how it could be done, but provide no evidence of those methods being successful. Not my job to prove every possible method is impractical, you just have to show one.

[mtwieg]

Years ago I developed high-speed (1.2MHz) PWM modulators with discrete components, see attached schematic. The current mirror is made using ALD1102 (could use other FETs or BJTs, or a resistor pulling up to a much higher voltage, but very low capacitance is important). The comparator U101 was TLV3502, it creates the ramp waveform and compares that with the error amplifier output (VERROR) to create the PWM output.

Getting the reset part of the waveform fast and clean was the trickiest part. I found that resetting the ramp using a very low capacitance schottky (D101, SMSA3923) worked better than an open-drain FET or BJT. RC values were adjusted manually to provide a specific reset pulse duration (was between 10-50ns, I forget exactly). The ramp doesn't reset all the way down to 0V, but this isn't a problem. The resistor network R110, R112, and R113 shifts/scales the error amplifier output (VERROR) to match the amplitude and offset of the ramp waveform.

[PCB.Wiz]

Getting something useable for the purpose of the OP is not simple or small or cheap if it has to be built from scratch. So much easier to use an existing chip with the function integrated even if the other parts are ignored/disabled.

Of course, but the OP has already stated this is an education and experience exercise, they do not want to put this into production.

[PCB.Wiz]

In order to learn more about power electronics, I am working on my own simple 500 kHz buck converter. I am done with the basic design, components, input filter. gate drive and feedback control with a type-iii compensator. Works like a charm. However,  I am missing a 500 kHz ramp generation and I want to have this included as well instead of hooking up my function generator.

How can I  create a 500 kHz sawtooth ramp with basic components? It doesn't need to be perfectly linear, it should just be monotonic and smooth from 0V - 4V (with e.g. a 5V supply). What would be that state-of-the-art approach to tackle this problem?

Here are some Spice examples using a single-gate package plus Dual RC, that also show the difference between a simple RC (which is  monotonic and smooth ) and a current source (nominally linear)

What about using a triangular-wave?

A fast '555 variant can manage 500kHz 'RC triangle', that swings between ~33% and 67% of supply, useful for experiments.