New Bee to Class D Capabilities

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi folks,

I am new to hardcore Class D audio power amplifier design . i plan to start with sigma delta modulation and error integrator design (about which i dont have any idea). :(
If somebody amonst our forum members could assist me on "How to go about "it would be great. Please spare a few minutes of your time and help.
 
http://www.diyaudio.com/forums/showthread.php?threadid=36852&perpage=10&pagenumber=1

there are a series of directions to beginners.

We usually use self-oscillating PWM scheme instead of delta-sigma,for it need a relatively low switching frequency.

to use delta-sigma,put a D flip-flop between the comparator and the gate driver & clock it with 1MHz ~ several MHz.

be a beginner,maybe you can try doing something with low power supply(5V logic,etc.) using some comparator and logic gates. thats easier than high power circuits.

a 555 timer connected into schmitant trigger(self-oscillating PWM ), or a 74HC74 D flip-flop (delta-sigma) plus some transistors(as voltage follower) is enough to drive a small speaker 8ohm/0.5 watts.

you can try to implement many kinds of feedback control techniques on these small devices.

remember to add the output filter--small things may not need them,but a big amplifier must.
 
Kenshin said:
http://www.diyaudio.com/forums/showthread.php?threadid=36852&perpage=10&pagenumber=1

there are a series of directions to beginners.

We usually use self-oscillating PWM scheme instead of delta-sigma,for it need a relatively low switching frequency.


Thanks Kenshin,

Why i wanted to start with delta sigma itself is ( please clarify if i am wrong) because, the dynamic range offered by sigma delta modulation is far more superior than that offerred by other topologies, better than 144 dB.

By the way are their any issues involved about PWM High frequency jitter when its driven by an external clock?

which one would be a better option to go about oscillator designed around a 555 timer or a 7404 based crystal oscillator?
 
phase_accurate said:
If you search the web for delta sigma class-d amplifiers you will find quite a lot of info, mostly from universities.

Hysteresis modulator type class-d has indeed some similarities with delta sigma modulation.


Thanks Charles (The Class D GURU :captain: in diyAudio forum) for the inputs. I'll go through it.

A question thats bugging me since quite long is how do we decide the PWM frequency??

Is their a thumb rule based of the bandwidth of the input signal ( oversampling thing) or their is a standard theroitical procedure for that??
 
Razor_Edge:

144 dB dynamic range? where did you saw the report?
I haven't heard anything about that.

basic (order 1) delta-sigma has a relatively low dynamic range -- perhaps 20dB dynamic range per 10 times of OSR.

to get a wide dynamic range,high order modulater is needed,but modulater with order>2 is complicated and difficult to make in amateur circuits.

about switching frequency:

the delta-sigma's sample rate is decided by oversample rate & SNR & the order of noise shaping.

in self-oscillating modulator or PWM with feedback, the switching frequency is decided partly by the efficiency of the output LPF, and mostly by the preferred feedback depth at audio frequencies(in the linear approximation model, it's the same thing as noise shaping in delta-sigma).

In an analog-to-analog delta-sigma converter,crystal oscillator is not needed--for you needn't make it synchrous with your computer.

you can try both the self-oscillating circuit and the delta-sigma.

the only difference is a schmitt trigger v.s. a clock-controled D flip flop.

both are simple and cheap digital devices.
(at the beginning,you don't need to care about the power stage)

Razor_Edge said:


Thanks Kenshin,

Why i wanted to start with delta sigma itself is ( please clarify if i am wrong) because, the dynamic range offered by sigma delta modulation is far more superior than that offerred by other topologies, better than 144 dB.

By the way are their any issues involved about PWM High frequency jitter when its driven by an external clock?

which one would be a better option to go about oscillator designed around a 555 timer or a 7404 based crystal oscillator?
 
The basic formula for calculating the expected SNR of a 1-bit delta-sigma modulator of loop order N is:

SNR [dB] = ld(0.5*fsample/finput)*(6 * N + 3)
= log(0.5*fsample/finput)/log(s)*(6 * N + 3)

I.e. per doubling of the sample rate (or also decreasing the input frequency) the SNR increases by 1.) 3dB due to naturally spreading the power of the error over double the frequency range and 2.) even more imporant it increases by 6dB per order of the noise-shaping loop.

This is of course theoretical since I don't belive that an SD amp with a sampling rate of 2 Msamples/s and a loop order of 4 would achieve an SNR of more than 150 dB (@ 20 kHz) like the formula indicates. Otherwise Sharp wouldn't use a 7th order loop for their amps while only achieving an SNR of 100 dB approx.

But I am still convinced that it is possible to build decent SD amps.

Regards

Charles
 
Kenshin said:
Razor_Edge:

144 dB dynamic range? where did you saw the report?
I haven't heard anything about that.



:confused:

Hi Kenshin,

I think I read that somewhere on one of the Sigmatel's sites. i should blame my ignorance about the subject i guess :confused: (might have had misinterpreted something).

I am still confused as to how to start learning all this stuff. Self switching or SD. (newbie).

I copied a schematic from johan's 100 watt Class D amplifier about a year and a half back. I faced some difficulties getting the output (on a general purpose prototype board ) but the results were amazing. I think that kind of topology would be the most simplest and result oriented Designing. Can you throw some light on that part??
 
Razor_Edge said:


:confused:

Hi Kenshin,

I think I read that somewhere on one of the Sigmatel's sites. i should blame my ignorance about the subject i guess :confused: (might have had misinterpreted something).

I am still confused as to how to start learning all this stuff. Self switching or SD. (newbie).

I copied a schematic from johan's 100 watt Class D amplifier about a year and a half back. I faced some difficulties getting the output (on a general purpose prototype board ) but the results were amazing. I think that kind of topology would be the most simplest and result oriented Designing. Can you throw some light on that part??

I think that's how the ZapPulse works too doesn't it? I'm not sure.

Start learning this stuff with persistance, that's all I can say.

Self oscillating has alot of advantages and no disadvantages, so I agree with you, it is simpler, higher performing, it can naturally compensate for alot of errors clocked types suffer from.. there's no downside that I know of.
 
classd4sure:

downside of self-osc:

the switching frequency varyes in a wide range. during output near the rail, the frequency will drop to very low -- this cause bad perfomance and even cause resonance of output filter.

feedback from the total output instead of the switching stage could produce a stabler frequency and output ripple lefel if designed properly. But if designed wrong, the oscillation may happen on the resonance frequency of the output filter. If the load is open-circuited, god know how much voltage could the output be.

perhaps add two clamp diods between output and +/- rail may be safer.
 
Hi,

Those are things to look out for, nice to have them mentioned, but I'm not convinced it's a problem. Every single topology has areas that need careful attention.. "If designed right" applies to everything.

Sonically speaking, it is far superior.


That's why phase oscillation is nice, it is easy to take post filter feedback.

The Fs can vary but it isn't that big a variation, nothing to be concerned with. Clipping can be problematic, but you can easily clamp the input to avoid that.

Fs variation is alot worse for hysteresis modulators, but once again..they can, and should be stabilized just by modulating the hysteresis band.

Why clamp the output, if the load is removed you still have post filter feedback.

When I speak of self oscillating you can be assured I have post filter feedback firmly in mind, and not pre filter or even mixed. We have to learn from what works and improve on that instead of always trying to reinvent the wheel.

Anyway, properly designed, there's still no downside, it has every advantage over other methods.

I did mention the Zappulse though, it has pre filter feedback, load dependant frequency response... it isn't a design I'd base my own on.
 
Last September there was a thread all about the workings of various self oscillating class d schemes:
Non clocked (e.g. free running, self oscillating) class d

One little known design fact about the phase shift type of d-amp is that the phase shift network must provide 180 degrees of shift at amplifier quiescence and only 90 degrees of shift at when the amplifier output is railed. The usual feedback phase shift scheme is to augment the a dominant pole (90 degrees shift - perhaps from the output inductor) with two identical high frequency real poles at the intended oscillation frequency. These are in the form of RC "noise" filters (45 degrees shift each) placed in the feedback path (one is often right at the comparator input). When the amplifier is nearly railed these RC section must only supply 10 or 20 degrees of additional shift over the constant 90 degrees from the dominant pole, thus, the operating frequency must drop under the nominal RC frequency by a factor of about three.

The obvious way to limit this drop in operating frequency is to increase the rate of phase change of the feedback phase shift network. This can be done by increasing the Q of the phase shift elements (if gain peaking can be tolerated) or by adding in additional low Q, low pass elements (i.e. more RC sections).
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.