help in Sigma Delta Modulation concept

[surZ90]

Hi all,
i want to make class D amp using Sigma Delta Modulation..
i have read many paper and i confuse about the concept of Sigma Delta Modulation..

first,
i want to ask, is SDM a self oscillation method, because SDM doesn't need to generate external carrier signal?
then, is SDM same with self-oscillating hysteresis-controlled?

second,
in conventional PWM modulation using external triangle generator and the frequency of output signal depends on the triangle generator, and pulse width of output signal represent amplitude input signal.
but how about in SDM modulation?
SDM will generate PDM, how SDM do that???and actually, what is PDM?i'm not clear about this..
and how about frequency in output signal PDM?is it bounded by using clock in quantizer??

third,
then in some paper i read, the quantizer using comparator with hysterisis. why using hysterisis? is there problem if just using simple comparator or it has any purpose?

last,
sry for my English..🙂

thank you for your help....

 

[gmarsh]

Sigma delta modulation is effectively "self oscillating", but implemented in discrete time as opposed to continuous time. The overall system is still clocked.

In a simple, first-order sigma delta encoder, there is a 'switching frequency' - this is at its maximum with no output voltage, and the frequency reduces with higher output voltage. Normally you have to clip the input so that you don't push the output of the encoder close to clipping, as this could reduce the switching frequency and potentially damage tweeters. If you're amplifying music, which is constantly changing in amplitude, the 'switching frequency' effectively moves around with it.

Hysteresis is used to reduce the output switching frequency, reducing FET losses. Generally the amount of hysteresis is changed based on the input signal level so that it doesn't cause the switching frequency to get even lower.

Other than that.. pick up this book and get cozy with it:

http://www.amazon.com/Understanding-Delta-Sigma-Converters-Richard-Schreier/dp/0471465852

 

[surZ90]

thanks gmarsh

In a simple, first-order sigma delta encoder, there is a 'switching frequency' - this is at its maximum with no output voltage

so the switching frequency switching in SDM (with no input signal) is frequency clock itself
and the frequency will change higher or lower depend on input signal?

and the frequency reduces with higher output voltage. Normally you have to clip the input so that you don't push the output of the encoder close to clipping, as this could reduce the switching frequency and potentially damage tweeters. If you're amplifying music, which is constantly changing in amplitude, the 'switching frequency' effectively moves around with it

what does it mean frequency is reduces?

with using hysterisis will it add more distortion, because it change time to switching?

i'm sorry about my question because i'm new in this topic...

 

[gmarsh]

so the switching frequency switching in SDM (with no input signal) is frequency clock itself
and the frequency will change higher or lower depend on input signal?what does it mean frequency is reduces?

Here's how a first order, 1-bit modulator without hysteresis behaves. + is a positive pulse, - is a negative pulse.

Halfway between rails:: +-+-+-+- (freq = fclock/2)
50% towards positive rail: +++-+++- (freq = fclock/4)
50% towards negative rail: +---+--- (freq = fclock/4)
99% towards positive rail: 99 +'s followed by a - (freq = fclock/100)
99% towards negative raill: 1+ followed by 99 -'s (freq = fclock/100)

Hopefully this is understandable. The closer your output signal is to the rails, the lower the switching frequency gets.

Using higher order modulators, hysteresis, and such throws this out the window for the most part.

with using hysterisis will it add more distortion, because it change time to switching?

i'm sorry about my question because i'm new in this topic...

Hysteresis has the same basic effect as lowering the switching frequency. Lower switching frequencies create higher distortion in theory, but trying to make a class D stage switch at a very high frequency can result in even more distortion - not to mention heat.

 

[surZ90]

Here's how a first order, 1-bit modulator without hysteresis behaves. + is a positive pulse, - is a negative pulse.

Halfway between rails:: +-+-+-+- (freq = fclock/2)
50% towards positive rail: +++-+++- (freq = fclock/4)
50% towards negative rail: +---+--- (freq = fclock/4)
99% towards positive rail: 99 +'s followed by a - (freq = fclock/100)
99% towards negative raill: 1+ followed by 99 -'s (freq = fclock/100)

Hopefully this is understandable. The closer your output signal is to the rails, the lower the switching frequency gets.

i understand this..thank you for this explanation!

Using higher order modulators, hysteresis, and such throws this out the window for the most part.

hm...i'm not really understand this...😕

Hysteresis has the same basic effect as lowering the switching frequency. Lower switching frequencies create higher distortion in theory, but trying to make a class D stage switch at a very high frequency can result in even more distortion - not to mention heat.

i see...thanks for explanation!