AD Vs BD Modulation in Class-D

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Who would trade overall EMI performance for eliminating carrier residual only when there is no signal?

Eva!

BD modulation doesn't mean this! EMI performance can be identical or better then in case of AD (at fixed freq).

Quoted from Bruno:
"Output filter?
•Single-core
• no excitation current
• crossover distortion results"

Single core doesn't mean no excitation current. Differential (useful signal) and common mode inductance (excitation current) can be selected independently.
 
BD modulation

This is what I was talking about, the output is quantized to one of three levels vs one of two levels for AD. This lowers EMI because the quantization noise (PWM output minus analog signal you want) is 6 db lower. But there is that inherant nonlinearity. And there is also a paper out there that claims that BD performs worst in inductive loads? True or not I don't know but if so it could be an issue with full bandwidth amps in the high frequency area driving driving more inductive loads.
 
I think Kanwar is talking about what is described in Bruno's paper (which was mentioned some posts before) and not some three-state amp that is using an "open" state.

This has one advantage: The apparent switching frequency and also the actual sampling frequency is twice as high for a given carrier frequency when you use a carrier-based amp. Therefore you can use more loop gain.

Regards

Charles

Charles,
Do you think that reverse recovery in BD modulation is not that much drastic as in AD modulation?

Regards,
Kanwar
 
This lowers EMI because the quantization noise (PWM output minus analog signal you want) is 6 db lower.

I don't think it's good to talk about quantization here, since PWM is basically analog. We can say carrier freq. component, or switching freq., component. Carrier residual is 6 dB lower, but this has nothing to do with EMI. EMI is generated by switching of the MOSFETs, and determined by supply voltage, schematic of switching stage, FET type, switching speed, snubbering, attachment method of heat sink, etc... Modulation scheme is unimportant.

And there is also a paper out there that claims that BD performs worst in inductive loads? True or not I don't know...

Not. First of all, BD is a modulation. The load don't have any direct influence to the PWM modulator. If there were any indirect influence, then it would have been via output stage. Output topology can be different (BCA, or bridged), but if properly built, none of them care about the load current direction, so they don't care about inductive load. I tried both, especially with big (1 mH... 400 mH) almost pure inductive load, and there was no problem.
 
Modulation scheme is unimportant.

AD modulation has NO common mode ripple voltage and "common mode voltage is what radiated off cables".

BD modulation has ripple voltage in phase on both output wires, and at non zero modulation index the radiated frequency is doubled in compare to Fsw. So, having 500kHz Fsw, in BD modulation we will effectively radiate 1MHz stuff using long output cables.
 
AD modulation has NO common mode ripple voltage and "common mode voltage is what radiated off cables".

BD modulation has ripple voltage in phase on both output wires, and at non zero modulation index the radiated frequency is doubled in compare to Fsw. So, having 500kHz Fsw, in BD modulation we will effectively radiate 1MHz stuff using long output cables.

A separate common mode coil is used by Ecler to reduce the effect of common-mode ripple floating on the outputs.


Floating rails and grounded bridge is another way of solving this inherent problem but it has its own disadvantages also.
 
AD modulation has NO common mode ripple voltage and "common mode voltage is what radiated off cables".

I guess you mean: balanced (bridged) output has no common mode ripple. Yes. But 1 or 2 volts at 250...400 kHz is piece of cake. The switching transients are the real enemy! The frequency of common mode output ripple is the switching frequency (unless you use BCA topology). And - as Kanwar mentioned - it can be filtered out easily.

AD modulation in half bridge has the same common mode ripple, so I don't understand why would this be an issue.
 
Last edited:
Hi all,
my free opinion on "BD modulation" is that noise= *2,(double transient creates effective +noise),also not is true a greater efficiency than the modulation AD.
PWM out (pwm + DT) the integral of the area is almost equal to the modulation
AD.
EMI not changed but IMD yes.

Regards
 
I guess you mean: balanced (bridged) output has no common mode ripple. Yes.

Yes, I meant BTL configuration...

But 1 or 2 volts at 250...400 kHz is piece of cake. The switching transients are the real enemy!

Agree

The frequency of common mode output ripple is the switching frequency (unless you use BCA topology).

BTL configuration with BD modulation doubles the radiated frequency at non zero modulation indexes. Having Fsw=500kHz, we will get 1MHz. And this is something bigger, than piece of cake, I would say :)

And - as Kanwar mentioned - it can be filtered out easily.

Filtered means lowered by some factor. But it does not mean 'cancelled', reduced near to zero, as in AD modulation.

AD modulation in half bridge has the same common mode ripple, so I don't understand why would this be an issue.

I would say, AD or BD modulation is applicable for BTL configurations only, not for half bridges. :confused:
 
BTL configuration with BD modulation doubles the radiated frequency at non zero modulation indexes.

Radiation can come from different parts, but previously you said common mode output voltage, I think we would better to stay with this now. Common mode freq does not double. Please think it over! What doubles is the output ripple on the speaker, wich is in differential mode!

Filtered means lowered by some factor. But it does not mean 'cancelled', reduced near to zero, as in AD modulation.

Yes, but who cares, when a much more agressive source of EMI still exists?

I would say, AD or BD modulation is applicable for BTL configurations only, not for half bridges.

So you havn't heard about BCA? But it doesn't matter. My point was: if common mode ripple were a serious problem at BD modulation, then why it isn't at single ended AD? (Or you really think AD is only for bridge?)
 
Last edited:
So you havn't heard about BCA? But it doesn't matter. My point was: if common mode ripple were a serious problem at BD modulation, then why it isn't at single ended AD? (Or you really think AD is only for bridge?)

Bingo...!!!;)

2 AD half bridges with clock in phase and audio signal out of phase results in BD modulated bridge[BCA does the same in half bridge while using 2 inductors and both switches switching at same time at idle.], there fore i don't think that the common mode component in BD is that much troublesome, when the same also exists in Half-Bridge AD.
 
Last edited:
So you havn't heard about BCA? But it doesn't matter

Sure I've heard about it. ;) And if you once look at the K-series schematics, then you will find at least 2 notch filters tuned for different frequency in its output filter section: Fsw and 2*Fsw. And please do not say, that BCA doubles the radiated frequency, and usual BTL configuration with BD modulation doesn't! Each output wire in BTL configuration with BD modulation radiates its own ripple voltages, which are always in phase and in sum they are giving doubled frequency at non zero modulation index.

Yes, but who cares, when a much more agressive source of EMI still exists?

Ok, you may don't care about the output ripple voltage till you have more agressive source of EMI ;)
 
And please do not say, that BCA doubles the radiated frequency, and usual BTL configuration with BD modulation doesn't!

I've already asked you to not mix radiation in general and common mode ripple. Common mode ripple freq is Fsw in 3-level modulated bridge, but 2*Fsw in BCA. If you don't believe me, then do your homework: run a simulation, or use your mind to analize it!

which are always in phase and in sum they are giving doubled frequency

You can't be serious! sin(x)+sin(x) wont be sin(2x)! Why would sin(x) be cancelled? In the difference there is no sin(x), that's why only the even components remain in the spectrum of the differential signal. (Even components are out of phase when there is modulation, otherwise zero.)
 
Last edited:
I don't think it's good to talk about quantization here, since PWM is basically analog. We can say carrier freq. component, or switching freq., component...

Interesting, I guess it's a different point of view. Since my loop is comparing the switched output to the analog input to create the error, and we can do noise shaping, I just tend to think of BD as a quantizing choice regardless of sampled time or not. I do see your point of view as well though. I think it may be useful to look at it in different ways.
Your second point on distortion I agree with 100%.

Cheers
 
I've already asked you to not mix radiation in general and common mode ripple. Common mode ripple freq is Fsw in 3-level modulated bridge, but 2*Fsw in BCA. If you don't believe me, then do your homework: run a simulation, or use your mind to analize it!

Ok, I will do simulation and I will post result here ;)

You can't be serious! sin(x)+sin(x) wont be sin(2x)! Why would sin(x) be cancelled? In the difference there is no sin(x), that's why only the even components remain in the spectrum of the differential signal. (Even components are out of phase when there is modulation, otherwise zero.)

do you really think, that the residual is of a sinusoidal form? :confused: Please, take a look on the ripple at high modulation index. It is some quasi-sinusoidal periodical stuff, and when it is summed from both output wires (radiated in common mode) in BD modulation, then its period is halved (freq doubled)...

In AD modulation it does get cancelled in common mode, because this quasi-sinusoidal stuff is of the same form on both wires, but of opposite sign.

stuff(x) - stuff(x) = 0

Sure, in real world it is not fully cancelled, because of difference in output filter parameters between half-bridges, wire lengths, etc, but the main amount is usually effectively removed when using AD modulation.
 
The conversation is interesting but perhaps I get lost in translation.
For me they are two separate issues. One is the noise (pink-noise at output), this is much greater than in the BD modulation. second point is the noise (EMI radiation). this depends on the shape of the square wave, the angle tips contain GHz and these are double in the case of BD. eg, an SMPS at 120V generate minor40% of EMI compared to an A 60V. therefore not a question of square wave amplitude only.
I agree with 81Bas, quantization error is double, in fact bd modulator has a lot more than IMD (3.5 kHz-7, 5KHz on IMD test).:)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.