Class d loop gain determination using simulation

Status
Not open for further replies.
Jump to Latest
K

Kootman

Member
Joined 2014
#1
Hello everybody,

I'm new here so bare with me....
I'm working on a class d amplifier which uses pre filter feedback for oscillation. Now i want to incorporate the lc filter with a second feedback loop which gives a flat frequency response. (not the ucd method).

I have some practical applications where this has worked (at least the no load part...). But that doesnt say anything about stability etc..

How do i determine the stability of a simulated amplifier (or a real one) ? I have used a single opamp with lc filter and feedback and tried the middlebrook approach to get the loop gain response. But i am missing the gain in the response and i'm not even sure if i am looking at the right picture....

Maybe someone can give me some hints on finding the loop gain of a class d amp. Which steps to take and if looking at the loop gain from the closed loop system is the right way to go...

Thanks...
 
#2
In the averaged model approach the modelled forward gain of the circuit section from input of the comparator to output of the halfbridge is given by the ratio:
(Pos rail - Neg rail) / Vpp triangle

The triangle does not need to be the triangle of a clocked PWM generator.
The method is also applicable to the carrier signal which you can find at the comparator input of a self oscillating design.
 
#3
Hello Chocoholic,

Thanks for the reply, maybe i should clarify something...
I have a model with comparator, delay and switching stage but when i do an ac simulation the gain is not correct....
Ideally this model i want to use to work on...
But because of this problem i tried to simplify things and used an opamp for the gain which drives a LC stage.
When i want to look at the loop gain of this model the gain isnt correct either.

So my main problem is how to get a model which gives a reasonable real response and look at the loop gain to determine the stability.
I want to compare this with real live data...

Thanks...
 
#4
AC analysis makes sense for linear systems, but not for flipping non linear comparators. For this reason people started to use linear averaged models since some decades.

For linear modeling with the averaged model method, you do not need an OP amp, you need an idealized gain stage with a gain which equals (Pos rail - Neg rail) / Vpp triangle.
First try to get the fundamental idea of averaged modeling. The WEB is full with papers, which hide the basic idea by an over complex describtion, but sooner or later you will get it.
Then have a look to page 7:
http://www.hypex.nl/docs/papers/AES124BP.pdf
 
#7
Hello ChocoHolic,

Thanks for the tips ! I'm relative new to this subject and still have lots to learn..
I will look into the concept of average modelling and see if i can get my head around it.

For now thank you and if i have any questions i will post again.
 
Status
Not open for further replies.
Top Bottom