Voltage Mode Pre and Post Filter Feedback

Status
Not open for further replies.
Jump to Latest
#4
Re: HD

classd4sure said:
Excellent! Certainly appreciate the examples, very useful. Can't wait for the rest. Writting style is :up: !
Click to expand...

Thanks! Loads of fun, I'm still learning myself.

koolkid731 said:
Excellent tutorial.

The idealized amp shows -47db distortion for 3rd harmonic and up.

I tried w/o success higher switching freq. to reduce its distortion.

I'll recommend this to newbees.
Click to expand...

It all 'depends'... I get -52dB for 1KHz 30V out. I did try reducing the deadtime and it made little difference. If you take out the clamp then it drops to -67dB. I might have to re-think that one.

Of course it's an idealised spice model but it does let you chase fundamental things.

Cheers

DNA
 
#5
I did take out the clamping zeners and got -67dB 2nd HD.

This is an example of drawbacks of sawtooth-based class-D: not enough loop gain at high frequency to reduce THD.

In a self-oscillating amp, the dominant harmonic is the third when modulation index is high, with other harmonics insignificant. I'm working on reducing that 3rd HD.
 
#7
classd4sure said:
Have you any plans for covering the topics of mixed feedback? It's a very non ideal solution but frequently implemented.

Seeing a current mode control design example could be interesting as well.
Click to expand...

Ahaaaa, errr not specifically mixed feedback, but you may notice that one of the sections on the front page mentions current mode control of a half bridge.

Strange to say I'm just dragging some things together to do that section. At the moment it's still a 'To Be Done'.

Cheers

DNA
 
#10
Very nice material.

Now that Charles is here, I would like to ask how the Nyquist theorem (that we have discussed several times here) matches into the explanations exposed in the Genomerics web. He speaks about slope matching, but how does that relate to the fact that having a crossover freq. higher than Fs/2 causes inestability?

Just that I wanted to hear Charles' point of view.

Best regards,
Pierre
 
#11
How about a inner current feedback loop that forces average inductor current to track a certain value, and an outer voltage feedback loop that estabilishes this average current value over time?

I have tried that in SMPS and it allows for lots of open loop gain while taking the feedback voltage after the output filter (it's stable because phase shift is just 90 degrees after the current loop is added). In such an arrangement, the voltage amplifier is stable just with a pole in the origin and a zero somewhere. I obtained sub-miliohm output impedance in a 15V 125A SMPS with that technique.

It's hard for me to believe that people just relies in inductor linearity and takes voltage feedback before the inductor in class-D audio amps. This is never done in SMPS, as the difference between the real output and the estimated output may be substantial.
 
#12
The combination of an inner current-controlled loop, and an outer voltage-control loop is exactly how the Mueta design (and others including my own design) works. Mueta have a patent on sensing the inductor ac current by measuring the filter capacitor current but there are other ways to do it.
 
#13
Hola, Eva.

(ahora en inglés). Although I agree with you in that before-filter feedback has drawbacks in Class-D designs (load dependency of frequency response, distortion due to filter components, increased output impedance, etc), there seem to be good-working (and sounding) amplifiers out there that use that principle, such as LCAudio ones. You know, audio is a completely different world 🙂

However, I also prefer to have post-filter control and we do that in all our coldamp designs, both amplifiers and switching power supplies.
 
#15
Ouroboros said:
The combination of an inner current-controlled loop, and an outer voltage-control loop is exactly how the Mueta design (and others including my own design) works. Mueta have a patent on sensing the inductor ac current by measuring the filter capacitor current but there are other ways to do it.
Click to expand...

I wonder how they could have managed to get that patent. These subjects have been public domain SMPS stuff for ages.

And that leads me to another question:
Is hysteretic current control (forcing constant current ripple without a clock) patented?
 
#16
phase_accurate said:
Very nice indeed. I always wanted to put tigether soething like that but haven't found the time so far.

Regards

Charles
Click to expand...


Hi Charles,

Wherever I sighted, I found your immaculate influence, but you are unaware of it.... 🙂

To everybody thats for the master: Charles Lehmann

K a n w a r
 
#17
Eva said:


I wonder how they could have managed to get that patent. These subjects have been public domain SMPS stuff for ages.
Click to expand...


There's a number of standard techniques used in one-quadrant buck regulators that are now being used in four-quadrant switching amplifiers. Although the techniques are prior art for use in power supplies, they are being patented for equivalent use in amplifiers. Whether the patents would stand up in law is another matter.

I have being doing some searching on patents for hysterectic current mode amplifier control and as far as I can see, everything apart from a few implementation techniques (such as Mueta's use of a differentiator to measure the capacitor current) are all prior art in general use.
 
#18
Pierre said:
Very nice material.

Now that Charles is here, I would like to ask how the Nyquist theorem (that we have discussed several times here) matches into the explanations exposed in the Genomerics web. He speaks about slope matching, but how does that relate to the fact that having a crossover freq. higher than Fs/2 causes inestability?

Just that I wanted to hear Charles' point of view.

Best regards,
Pierre
Click to expand...



phase_accurate said:


I have reread it and that choice of crossover frequency is in fact soemthing that i don't like. I'd personally go for < fswitching/2.

Regards

Charles
Click to expand...

It's something I've wondered about but am not clever enough to produce a decent answer for. I've seen elswhere that Charles mentions the fact that, with a triangle wave modulator, there are in fact two sampling periods per switching cycle.

Even so, and you can see this elsewhere on the site, with something like a buck converter and current mode control with a leading edge modulator, slope matching results in an indicated loop crossover frequency of Fs/2piD. If D less that 1/pi or about 0.3 then that exceeds half the switching frequency. This is with a sawtooth modulator so there is only one sample per switching cycle.

I know it's not a good position to argue from but I have regularly run simulations under these circumstances and the loop has been stable up to the slope matching limit. It has apparently been stable above it but then it is possible to 'knock' it over the edge into subharmonic oscillation by injecting appropriate noise. Perhaps there is something missing in the analysis?

I'll be dangerous here. I wonder if there is some confusion that arises from considering Nyquist Stability Criteria and Nyquist Sampling Theory. Feel free to shoot me for this one because I am hand waving.

Perhaps it is incorrect to call an SMPS or Class D amplifier of this type a sampled system because the error amplifier is in continuous contact with the signal it is trying to correct......

mumble mumble

:headshot:

OUCH!

DNA
 
#19
Eva said:
How about a inner current feedback loop that forces average inductor current to track a certain value, and an outer voltage feedback loop that estabilishes this average current value over time?

I have tried that in SMPS and it allows for lots of open loop gain while taking the feedback voltage after the output filter (it's stable because phase shift is just 90 degrees after the current loop is added). In such an arrangement, the voltage amplifier is stable just with a pole in the origin and a zero somewhere. I obtained sub-miliohm output impedance in a 15V 125A SMPS with that technique.

It's hard for me to believe that people just relies in inductor linearity and takes voltage feedback before the inductor in class-D audio amps. This is never done in SMPS, as the difference between the real output and the estimated output may be substantial.
Click to expand...

Cool Idea! That's what exposure to different technical fields gets you. Fair enough, it's all been done before (maybe) but that's not the point.

That's the next section on the site but at the moment it's causing some stress. My previous models and analysis have not been totally precise, I tried something different and it caused an Uh-Oh moment.... Then I've got myself into trouble with a particular argument that I shouldn't be allowed because I've done something different elswhere. 🙁

DNA
 
#20
It's something I've wondered about but am not clever enough to produce a decent answer for. I've seen elswhere that Charles mentions the fact that, with a triangle wave modulator, there are in fact two sampling periods per switching cycle.
Click to expand...

One has to be more specific: There are two sampling periods only at IDLE (or very faint input signals). The higher the output voltage the more they move together. Therefore one is on the safe side to regard it as one sample per switching period only.

Bruno once suggested choosing a crossover frequency of less than 30% (or was it 40 % ?) of the switching frequency .

Though I am currently unable to offer a correct and indepth mathematical analysis I assume that the crossover frequency has to be chosen more conservatively for the switching amp than the PSU. Keep in mind that the "input" for the PSU is static (i.e. the reference voltage) while it is a completely arbitrary signal for the amplifier.

Regards

Charles
 
Status
Not open for further replies.
Top Bottom