Pulse width Modulator..

[ozangazi]

i've desined a fixed frequency pulse width modulator for my class-d amplifier project.This is the first step of project and i will share all steps (output stage,speaker protection,feedback,PCB,ect...)of my desing

[IMG]http://img208.imageshack.us/img208/8120/adszbd9.gif
[/IMG]

Schematic Link:http://rapidshare.com/files/50794035/pwm_mod.zip.html

First stage is 400khz Triangle wave generator:

U1(ad826) is dual, Low cost, 50Mhz, Low power video opamp
U1A is working as a comparator and U1B is high frequency offset compensated inverting integrator.This configuration oscillates about 400khz.There is no frequency adjust but you can change it with variable resistances of R4 , R5 , R6.

R1,R2,R3 and C1 are used for integrators offset compansation

2-Triangle wave signal composer:

U2B(Ad826) is used for better performance of modulation.U can control your carrier waves amplitude and offset..

Offset control is necessary for classd application.Offset error of pwm pulses can lead to a high dc current at speaker output.

Amplitude control for triangle wave signal is directly effects amplifiers total gain or required input signal level..

U2A compares input audio signal and triangle wave signal and generates pulse width modulated signal for mosfet switching.

U3 is used for improving of pwm pulses rising ,falling times and creating inverted pwm pulsed for low side switching

Propagation Delay Time of U3 is about 10ns. This means there is 10ns difference at noninverted and inverted pwm output for fets dead time compensation..

i've uploaded a screenshot and simulation circuit for Labcenter's Proteus

waiting for advices about modulator.to be open to all views

[Whortless]

Hi ,

That is a known way to create a class-d amp.
Fisrtly I would suggest you to change the pin order of "offset1" trimpot. Mid pin should go to inverting input of the U2B.
On second stage I would suggest try LM360 or LT1016 to get inverted PWM signals.
The other point , I think 10ns is not quite good dead time to prevent your output mosefts overshooting . Please see and simulate the enclosed file by downloading LT-Spice SW-Cad. From : http://www.linear.com/designtools/so...witchercad.jsp

I think this will help you and you can understand the getting dead time with R,D,C. On my file A1 and A2 is shimt triggers. (many mosfet drivers has shimit trigered inputs ). U4 and U5 for simulation U can use other high speed opamps for your real desings.

Regards

[ozangazi]

You know this is a known and the basic way..

Yes it's false.. offset trimpot is escaped.i think i did it for PCB pinout fixing.thanks for this.

i will search your suggested opamps and i hope could get these components..

You are completely correct about dead time..AD826 is poorly for this circuit i know.And i place ttl gates for improoving ton toff values.But the the best way is using higher frequency opamp as far as i can see..

i will fix this falsities and share new modulator as soon as possible..

Thanks for all suggestion...Regards

[Whortless]

Hi,

I do not think AD826 is slow. But, you have to find a fast differantial comparator like LT1016 or similar for comapartor stage after triangle wave osc. OK, if you could not get these comps. try LM311 and add 4070 / 4077 or 4069(cmos) to get inverted PWM signals. I do not have an idea about your driver chip but you should use CMOS instead of TTL. If your driver chip is IR2XXX serie I would strongly suggest CMOS.
By the time, what is the desired output power? Which mosfets will be used on output stage? And did you check IRAUD Amps? IRAUD amps use selfosscillating topology but one can sync. them by using an external clock.

Regards

[ozangazi]

i have two matters about schematic

i looked up your design but the trianglewaves offset voltage is about +5 volts is this a good idea??. and input signal have to be float to +5 volts because of this...and it means one or more opamps..

And the other is about triwave frequency it's only 2k hertzs audio signals freq is much greater than 2khz.. Floundered.

[Whortless]

Dear Ozangazi,

I know the offset matters . But, this is only a simulation, real world is different than the simulation. You can do same circuit in proteus and can select different opamps(maxim, analogdevice, national etc.). The offset comes from LT1007. For simulation try a pulse source instead of Sq wave osc(1st opamp) and then integrate it with another fast opamp or use only R/C to get triwave.
You can understand what I mean for triwave osc by clicking this link : http://www.interq.or.jp/japan/se-inoue/e_ckt16.htm
R1 =1K, C1=1n and do not use fisrt opamp and R2,R3. Apply the sq wave to R1 and U will get a triangle wave. This is not real matter. In my desing one should get a good triwave instead of sinus on the output of the amplifier. Probably sinus modulates the triwave .

Regards

[ozangazi]

i restore modulator in accordance with your suggestion..

-i think now offset is much advisable.
-LM311 compares audio and triwave..
-and added a real dead time compensation with retriggable monostable multivibrator 74LS123.Currently deadtime value is about 100ns and audjstable by R12 ,R13..

i wait for leading..is ready now??

[ozangazi]

Sorry for attachment...

[Whortless]

Dear Ozangazi,

I do not have licensed proteus therefore could not open the attachment. Anyway pay attention to LM311 it has two output pins (1/7). If you carefully read the datasheet U can understand what I mean. It is a good idea to use 74LS123 but, TTL gates not my favorite logic (only 5v). What do you think about the feedback? It can be difficult to implement the FB to TTL levels, did you take this situation into consideration?

Regards

look attached triangle osc. Values for representation, u should find yours by sim. or breadboard tests.

[ozangazi]

Here is the schematic..

http://img259.imageshack.us/img259/9379/pwmmodkm8.png

i prefer TTL gates because of their speed.When i use CMOS logic square wave turns triangle wave.

i will work on feedback when then amplifiers base circuit is over.i think the best way for the feedback is differantial amplifier which compares input and filtered output as analog..and i'll modulate the error between input and output..