Philips UCD application note

[lumanauw]

Hi, Chris,

Is it possible to make this UCD amp a follower (G=1)? In the schematic, do not use R12/R11, output is directly attached to inverting input (like opamp buffer, to base of Q2).

What will happened? Will it blow out or not oscilating at all?

 

[classd4sure]

Hi,

My guess would be some sort of chaotic mess if you tried it like that.

The proportional gain of the output stage is set by R12 and R11, so if you want it to be a gain of one, just use R12=R11.

 

[RX5]

add to that:


higher value R12 = higher GAIN

lower value R11 = higher GAIN

 

[lumanauw]

Hi, Chris,

R11 and R12 are voltage divider. If we use R11=R12, then gain G=2.
Is it impossible to make it a follower (G=1)?

 

[djQUAN]

why would you want it to be a follower anyway?

 

[IVX]

why would you want it to be a follower anyway?

I guess, for more loop gain.

 

[classd4sure]

Hi, Chris,

R11 and R12 are voltage divider. If we use R11=R12, then gain G=2.
Is it impossible to make it a follower (G=1)?

Right you are! Sorry, In my little world it would have worked that way

I always have the differential version in mind, and I recommend you use that instead.

You'll have 2XR11 and 2XR12, R11=R12, A~1

 

[Pafi]

Hi, Chris,

Is it possible to make this UCD amp a follower (G=1)? In the schematic, do not use R12/R11, output is directly attached to inverting input (like opamp buffer, to base of Q2).

What will happened? Will it blow out or not oscilating at all?

Oscillation freq will be much lower depending on load impedance, and Q1-Q2 may will be blown up because of high feedback voltage.

 

[IVX]

Oscillation freq will be much lower depending on load impedance, and Q1-Q2 may will be blown up because of high feedback voltage.

I don't believe that such apocalypse will occur yet.

 

[classd4sure]

I don't believe that such apocalypse will occur yet.

I tried the single ended version with the feedback components removed it was pretty much a chaotic mess, I think Pafi's prediction is right on. First it really just slammed the output to the negative rail where it stuck, so I equalized the inverting/non inverting impedances and it oscillated very slowly, 50% duty cycle, output in full saturation (20V beyond the rails).

I haven't put much thought into it but there may be some chance at getting it to work if you configure it in non inverting mode, what do you think? Still, differential is vastly superior.

 

[fokker]

and Q1-Q2 may will be blown up because of high feedback voltage.

and that's something not many people appreciate in class d amp.

 

[IVX]

I tried the single ended version with the feedback components removed it was pretty much a chaotic mess, I think Pafi's prediction is right on. First it really just slammed the output to the negative rail where it stuck, so I equalized the inverting/non inverting impedances and it oscillated very slowly, 50% duty cycle, output in full saturation (20V beyond the rails).

I haven't put much thought into it but there may be some chance at getting it to work if you configure it in non inverting mode, what do you think? Still, differential is vastly superior.

I'm talking about Fsw. Really, differential is vastly superior, but it's other matter, if comparator ready to work well (e.g. comparator VDD>VDD power) , so no problem.

 

[Pafi]

IVX!

Fosc is the freq where loop phase shift reaches -180 degrees. This phase shift comes from 3 parts: filter, feedback, delay of comparator+powerstage. If we have feadback=1, then only 2 parts left. Delay of comp+PS introduces a small, negative component (phi=360*delay*freq). Worst case: an unloaded LC filter already has -180 phase shift above resonance freq, so with no load fosc=fresonance. I dont have to tell what does this mean. (Theoretically unlimited output voltage.)

With load it is less catastrophic, but still bad. You can simulate it with an ideal comparator, delay, and LC filter!

This is exactly why post-filter feedback is not trivial to design.

 

[IVX]

IVX!

Fosc is the freq where loop phase shift reaches -180 degrees. This phase shift comes from 3 parts: filter, feedback, delay of comparator+powerstage. If we have feadback=1, then only 2 parts left. Delay of comp+PS introduces a small, negative component (phi=360*delay*freq). Worst case: an unloaded LC filter already has -180 phase shift above resonance freq, so with no load fosc=fresonance. I dont have to tell what does this mean. (Theoretically unlimited output voltage.)

With load it is less catastrophic, but still bad. You can simulate it with an ideal comparator, delay, and LC filter!

This is exactly why post-filter feedback is not trivial to design.

Ok, let's make simulation, BTW, it's stable even completely without load.. AC analysis was done with stepping R1 from 8.2k to 200ohm.

 

[classd4sure]

Hi Ivan,

Your simulation doesn't seem to match the discussed conditions.

 

[IVX]

Hi Ivan,

Your simulation doesn't seem to match the discussed conditions.

I thought the moot point is G<=1 whatever detail.

 

[classd4sure]

nah, it was no feedback resistor in the given inverting mode, that's what Pafi and I were talking about.

Wicked sim's though

 

[Pafi]

IVX!

This was the question:

Is it possible to make this UCD amp a follower (G=1)? In the schematic, do not use R12/R11, output is directly attached to inverting input (like opamp buffer, to base of Q2).

BTW: is UcD really so fast? (I mean is td=220ns real? I haven't built it yet.)

 

[classd4sure]

If I recall correctly and I probably don't, 200ns for the comparator, 250ns output stage.

 

[IVX]

If I recall correctly and I probably don't, 200ns for the comparator, 250ns output stage.

So 200+250=450? Sure 200 or less even. AC show 330Khz vs 384Khz transient.