Fast Response Low Pass Filter

[v1.5]

Hi all.
I'm thinking of some kind of voltage converter circuit from pwm. this circuit should definitely react fast and the output ripple should be as little as possible. pwm input frequency is 10khz and around. the output is in the range of 0 to 5 v.
I tried various methods ( usually low pass fltr. )via simulation programs, but I did not get what I wanted exactly. I saw that some integrations of the Analog Devices in LTC series perform this function, but these integrates are not on sale in my current location. For this reason I can not provide. The ltc family uses dac in its content, so it can react quickly. I want to perform this process without using microcontroller and dac. dutycycle value should react to rapid changes. therefore, typical low pass filters are non-functional. There may be some kind of circuit structure that you may know. Or you can give me an idea on this subject. That's why I wanted to ask you.
Thank you

 

[Nigel Goodwin]

Why do you need a fast response?, and how fast does it need to be? - low pass filters are usually perfect for this kind of application, so it sounds like perhaps you're doing something 'wrong'?.

 

[v1.5]

Why do you need a fast response?, and how fast does it need to be? - low pass filters are usually perfect for this kind of application, so it sounds like perhaps you're doing something 'wrong'?.

thank you so much.
The speed of the converter should be able to react with every pulse. For example; Let's say the first pwm pulse contains 50% dutycycle. at that moment the output should be 2.5v. If the dutycycle for the next pulse is 25%, the output should be 1.25v. In other words, we can say 1/10000 that is 0.0001 second as speed.

 

[Nigel Goodwin]

thank you so much.
The speed of the converter should be able to react with every pulse. For example; Let's say the first pwm pulse contains 50% dutycycle. at that moment the output should be 2.5v. If the dutycycle for the next pulse is 25%, the output should be 1.25v. In other words, we can say 1/10000 that is 0.0001 second as speed.

Again why? - this isn't something that would normally be desirable, and probably isn't possible? - so why not tell us exactly what you're trying to do, rather than how you're trying to do it.

 

[v1.5]

Again why? - this isn't something that would normally be desirable, and probably isn't possible? - so why not tell us exactly what you're trying to do, rather than how you're trying to do it.

sorry i forgot to mention this
I want to use this circuit in some kind of machine tool. therefore it needs to be as sensitive as possible. I concentrated my thoughts on some kind of circuit model and this circuit model requires the circuit structure I mentioned above. I think that the faster he can respond, the more sensitive I can take. this is all thought, not a realized project yet!

 

[v1.5]

and probably isn't possible?

I want to know why you think that way ?
Fast response time seems impossible without using DAC. Because as you can understand i don't want any delay in my thoughts. This makes it impossible for me to use low pass. Do you think a control algorithm made with dac and microprocessor will work?
Thank you.

 

[Beau Schwabe]

Isn't a low pass filter by nature slow? ... essentially blocking the high frequency (faster) signals.

You could look at the "delta" between the output of the low pass filter and your "raw" signal and based on the amount of error between the two, proportionally update the low pass filter output node. At this point though, it's not exactly a low pass filter anymore.

The circuit I am thinking of in addition to the low pass filter is a push-pull drive commonly used to drive a capacitive load of a MOSFET gate.

Drive a MOSFET via BJT?

My intention is to drive a MOSFET (say, a IRF840 or some logic-level) from a low-current (<10mA) SoC output pin. Using appropriate resistors from gate to pin and ground, this works... but the g...

electronics.stackexchange.com

 

[Reloadron]

thank you so much.
The speed of the converter should be able to react with every pulse. For example; Let's say the first pwm pulse contains 50% dutycycle. at that moment the output should be 2.5v. If the dutycycle for the next pulse is 25%, the output should be 1.25v. In other words, we can say 1/10000 that is 0.0001 second as speed.

I doubt you will achieve that. The reason being is the RC time constant involved. You mention a low pass frequency of about 10 KHz.

The above image illustrates a 100 KHz 5.0 Volt PWM signal with a 50% duty cycle. The 100 KHZ has a period of 10 uSec. Note the time it takes for the capacitor to charge to 2.5 volts. You will not get an instant change. Also keep in mind a LP filter is not going to supply very much current depending on your application. In the example the load is 10 Meg Ohm. Think about the RC time constants and how many it takes to charge the capacitor.

Ron

 

[crutschow]

The fastest way would be to measure the PWM period with a micro, which would give a cycle-to-cycle measurement, but since you don't want to do that, then I don't know anything faster than a high-order (e.g. 3-pole) low-pass filter.

We need actual requirements.
You can't design for "as little as possible" ripple, with "as fast as possible" response.

 

[Nigel Goodwin]

I want to know why you think that way ?
Fast response time seems impossible without using DAC. Because as you can understand i don't want any delay in my thoughts. This makes it impossible for me to use low pass. Do you think a control algorithm made with dac and microprocessor will work?
Thank you.

Well now you've explained what you're trying to do - all becomes clear. You're trying to control a slow mechanical process, so you have no speed issues at all - use a low-pass filter, job done. Your original question makes little sense, because there's no context, and your original idea is mistaken - you've no need (or use) for such speed as you envision.

 

[ronsimpson]

10khz PWM. Zero delay. 0-5V.

i don't want any delay

We need to define "any delay". Using your example of 10khz, there will be 100uS of delay just for the PWM time. That seem OK to you I think.
If you move the frequency up to 100khz then the filter will be much better. Now we have 10 cycles of time to reach a output. If the filter had 100uS of "delay" the output would be pretty stable.

We also need to know what you are driving? I have driven motors with out any filter. The mechanical momentum forms a delay-filter function.

Here I made two filters. The bottom one is a simple RC filter with a op-amp buffer after it. (high ripple voltage) The top one is a second order filter with a small amount of over shoot. (using 10khz foing from 0% to 50%)

If we moved the PWM frequency to 100khz the p-p ripple voltage will be 1/10 as much. Or we could reduce the delay through the filter. Or we could a combination of less ripple and less delay.

Having done this many times, the "over shoot" can help with some of the mechanical delays in the system. I can make it over shoot any amount you want.

 

[ronsimpson]

Speed up the frequency by 10:1. Did not change the filter.

 

[Beau Schwabe]

You can get similar results to post #11 and #12 with what I was describing in post #7 ... by adjusting the "Attack" resistor you can affect the filter response time.

 

[gophert]

I want to know why you think that way ?
Fast response time seems impossible without using DAC. Because as you can understand i don't want any delay in my thoughts. This makes it impossible for me to use low pass. Do you think a control algorithm made with dac and microprocessor will work?
Thank you.

you are confusing instantaneous (single pulse) PWM power with motor speed In your machine. if you want instantaneous voltage control to a 10kHzto control a motor spinning at 10kRPM, do you really expect to change the rpm 60 times during each revolution of that motor? Motors have inertia and take some finite time to slow down or speed up.

Also, your desire to convert a PWM power control to a voltage control just disables the slow speed capabilities of your controller - a hallmark of PWM vs voltage control. Specifically, a 5v motor won't turn with any smooth motion at 0.5v but it will run perfectly smoothly at 10% duty cycle on a 5v supply.

 

[gophert]

One more thing...

one of my rules to live by in electronics, home/auto repair, parenting, finances, ...

if you are having trouble implementIng your solution and it seems like nobody else is running into the same problems, you're likely (LIKELY)...
- using a sub-optimal solution, or
- you are not understanding the problem, or
- you are not understanding the solution.

 

[ronsimpson]

by adjusting the "Attack"

The attack resistor is the same as picking a different "10k" resistor. Example: 10K||10K_arrack=5k

 

[Beau Schwabe]

" The attack resistor is the same as picking a different "10k" resistor " - Not exactly ... the transistors provide an active circuit that provides low impedance when OUT1 does not match the 0V to 5V input and a high impedance when they agree with one another.

 

[crutschow]

Speed up the frequency by 10:1. Did not change the filter.

Why does your simulation have a 1KHz PWM frequency instead of 10kHz?

A 3-pole filter can improve the ripple to rise-time ratio.
For example, below is the sim for a 125Hz 3-pole filter with a 10KHz PWM signal:

 

[rjenkinsgb]

Back to the basics - what is the source of the PWM and the duty cycle control input to that?

Real-world machine factory machine tools & robotics systems in general either use a DAC to directly produce the command voltage for an analog-setpoint servo drive, or the "servo drive" is digital with the setpoint transferred as data.

Either way, the only PWM is in the output stages that drive the motor.

 

[v1.5]

One more thing...

one of my rules to live by in electronics, home/auto repair, parenting, finances, ...

if you are having trouble implementIng your solution and it seems like nobody else is running into the same problems, you're likely (LIKELY)...
- using a sub-optimal solution, or
- you are not understanding the problem, or
- you are not understanding the solution.

the most accurate interpretation I've ever seen. really thank you so much. I will keep this word in my mind.