| Pabo |
I have tested putting a low pass filter in feedback loop of a self oscillating hysteresis stage with the result that one pole of the output filter is eliminated. I have though seen a fairly new patent claiming this technique but I would be surprised if this hasnīt been done about a thousand years ago in for example OPAMP circuits.
Has anyone seen such a circuit before which would make that claim invalid? |
|
|
| phase_accurate |
I guess you are talking about the patent of a fellow Scandinavian, aren't you ?
Building a lead filter the way as K.N. did is per se not new at all. It is not even new to have a lowpass in the feedback path of a class-d amp (I did this 15 years ago as well !).
But it was new to do this around a switching amp's modulator in order to generate a phase-lead ! It is 1.) the total combination of all aspects that counts and 2.) what the patent office regards as novel.
It actually depends on what is exactly claimed in the patent. If it was just a lowpass in the feedback path of a switching amp without any closer definition what it is used for, then it could easily be killed by the circuit from my thesis (which would then be prior art).
K.N.'s trick could even be considered as novel although someone else patented something like a DC/DC converter using such a modulator ages ago, since the application of it to audio would be a different one. It always depends on what the examiners let you get away with .......
Regards
Charles |
|
|
| Pabo |
Yes, unfortunately the claim is in combination with preceeding claims which in total gives a classd amplifier with global feedback.
Well, I can do without it. Using this technique would give me the possibility to apply about 6dB more of feedback but this isnīt the whole world is it.:) You still have to watch out for Nyqvist. |
|
|
| phase_accurate |
Yes there are more possibilities to generate a phase-lead. K.N.'s version is not the only possible solution but I must enviously admit that it is quite refined and elegant.
The 6 dB of additional NFB you are talking about could give you increased load independancy so it might be well worth it.
Regards
Charles |
|
|
| Pabo |
Another way of creating phase lead is of course to add a zero in front of the modulator but the disadvantage with this solution is that the switching ripple is increased which may cause high frequency instability for example during startup. This disadvantage is probably not present in K.N.s circuit, but I am not sure about it.
Transient simulations using K.N.s technique gives a very good behaviour at all frequencies and loads.
:( :( :( |
|
|
| classd4sure |
Hi,
IcePower patent? Or something newer? Care to share? |
|
|
| Pabo |
classd4sure
ICEpower, Yes, it is one of Karsten Nielsens earliest patents. The title is something like Multi Enhanced Cascade Control, I think. Use it for DIY but be careful if designing professionally.
I have been designing classd using a current loop in the output stage but there are some problems with this technique. First of all it is very difficult to sense the current in a linear manner. I have used 2010 MELF resistors creating a shunt of 100mohms. With this shunt I measured the return current from the output filter including both the capacitor and load current. It is a very good technique when it comes to phase shift and high frequency behaviour. A square wave looks allmost ideal.
My problems are: The negative output is not ground but a triangular wave giving more noise on the output. I have not been able to make the loop as linear at high frequencies as when using voltage integrator technique before the filter.
One would think that the output impedance at high frequencies would be a problem but actually it can be pretty low. I have managed to come down to 0,5 ohms at 20kHz using about 18dB of feedback.
One idea that I have is to measure the current flowing into the capacitor/load but this would require a measurement circuit tracking the output signal +/-50V (kind of difficult). The circuit has to be able to measure a triwave at 700kHz in order to make it future proof.
Does anyone know of such a circuit? |
|
|
| classd4sure |
Hi Pabo,
Thanks for sharing :)
You said "fairly new patent" which threw me, K.N.'s newest patent (2004) is the highly vague abomination straight off the mains with the modulator driving the voice coil directly with no output filter. :xeye:
SO, you're efforts are very interesting.
Have you read the mueta patent? Maybe you can get some ideas from that. You could possibly also do something similar to what their current datasheet shows. That's all I can think of. |
|
|
| Pabo |
classd4sure
Thanks for the tip:)
The Mueta patent is according to me a very complex way of going around the ICEpower patent and achieve a -1 slope in the output filter without using the output voltage. The amount of OPAMPs needed to create the current "set value" in the capacitor and to correct the output voltage is to high according to me. Otherwise itīs a good technique because it is a self oscillating current loop without the disadvantage of including the load current which causes higher output impedance. |
|
|
| Kenshin |
if you elimating the pole of the output filter for bigger bandwidth,
that's of no good in audio device.
At large input at high frequencies , to drive the filtered output to full amplitude,the output amplitude of the main amplifier will be greater than the rail--this is impossible. so TIM distortion may occur.
but I have once tried an idea like this in a buck SMPS to eliminate the output LC filter's overshoot completely.
| quote: | Originally posted by Pabo
I have tested putting a low pass filter in feedback loop of a self oscillating hysteresis stage with the result that one pole of the output filter is eliminated. I have though seen a fairly new patent claiming this technique but I would be surprised if this hasnīt been done about a thousand years ago in for example OPAMP circuits.
Has anyone seen such a circuit before which would make that claim invalid? |
|
|
|
| Pabo |
Kenshin
The total system bandwidth is allways kept at about 80-100kHz. The reason for creating a -1 slope in the output filter is that the phase margin allows for more feedback. Actually, TIM distorsion is less likely to occur if the output filter is faster. |
|
|
| phase_accurate |
Kenshin is basically right.
But keep in mind that not every conventional amp's power bandwitdth is as high as it's small signal bandwidth. If you don't overdo it then there are not much problems to be expected. Bruno does this as well witch his UcD. The UcD 180 has an output filter with a cutoff-frequency of 30 kHz approx and the closed-loop bandwidth is 50 kHz.
This comes at the risk of of increased susceptibility to TIM but has the big advantage of getting a more or less load independant frequency response, because the NFB is now dominant.
If you want the best of both worlds then use a lowpass in front of your amp and make the whole combination behave like a bessel LPF.
If you add an LPF to the feedback just for getting rid of some RF hash then it would not touch touch the ICE patent as long as this filter's cutoff frequency is higher than the carrier frequency. It all depends on what you want to do.
Regards
Charles |
|
|
| Kenshin |
NFB from total output is a good idea,if no TIM--it's good for ballasting the resonance and reducing nonlinear distortion of the output LPF.
how could a -1 slope in the output filter decline the switching ripple sufficiently?
at 300KHz switching frequency,the amplitude of the open loop gain should be 1. Hence at 30KHz frequency,it will be 10. So it's hard to introduce an deep NFB. Another problem is that a hyerstis
modulator running at high ripple level will produce more nonlinear distortion from the RC integrator. |
|
|
| phase_accurate |
The trick is to get a sifficient phase-marging AND still have an output filter of at least 2nd order AT ONCE !
That's what Pabo wants to achieveand that's also feasible with clever circuit topologies.
Regards
Charles |
|
|
| Kenshin |
The output filter is of order 2--this is OK.
but the feedback circuit- if add some positive phase shift in the NFB loop to get bigger phase margin,the ripple in feedback will also rise. |
|
|
| Kenshin |
ICE patent contains two switches operating at different phase.
would the design of creating positive phase shift in the feedback loop touch the ICE patent? |
|
|
| phase_accurate |
| quote: | | but the feedback circuit- if add some positive phase shift in the NFB loop to get bigger phase margin,the ripple in feedback will also rise. |
Yes, it will be of the same height as if there was only a 1st order transfer function in total. This is the same as if feedback is used from the output stage directly and fed into a 1st order integrator.
Because the feedback thingie is not restricted to polyphase PWM in B&O's patent it does cover almost any PWM topology where the phase-lead is generated by a lowpass around an inner feedback loop.
It does however not cover any other topology that generates a phase-lead.
Regards
Charles |
|
|
| classd4sure |
| quote: | Originally posted by phase_accurate
Yes, it will be of the same height as if there was only a 1st order transfer function in total. This is the same as if feedback is used from the output stage directly and fed into a 1st order integrator.
Because the feedback thingie is not restricted to polyphase PWM in B&O's patent it does cover almost any PWM topology where the phase-lead is generated by a lowpass around an inner feedback loop.
It does however not cover any other topology that generates a phase-lead.
Regards
Charles |
Would the UCD patent cover the other method?
Seems to be very little one can do without violating someones patent... kind of silly isn't it, as soon as you compare a sine to a triangle you've lost, yet, few of them actually invented anything.. at least there's free reign on negative feedback huh :rolleyes:
Pabo I wasn't really sugesting you duplicate mueta, but they have a few techniques for measuring the current, if I remember right, their patent covers them all. I certainly agree with you, too many op amps! |
|
|
| phase_accurate |
| quote: | | as soon as you compare a sine to a triangle you've lost, |
I don't think so since the oldest patents of class da amps are already more than 25 years old and therefore obsolete !
And compensating for a pole as such is basic control theory, and might be successfully fought against unless the method is really elegant and new (like the aformentioned detail of ICE or the UcD patent).
Regards
Charles |
|
|
| Pabo |
| quote: | ICE patent contains two switches operating at different phase.
would the design of creating positive phase shift in the feedback loop touch the ICE patent? |
Kenshin
No, applying a zero in the feedback network is not included in this patent. |
|
|
| Kenshin |
| And,how about UcD's patent? connecting a NFB from output and make the system self-oscillate, will this touch their patents? |
|
|
| Pabo |
Kenshin
I am not sure what you are referring to specifically but the ucd-patent and the ICEpower-patent are different in the way that ICEpower uses a separate 2nd order filter before the output filter to create the phase modulation while ucd uses the existing 2nd order filter itself.
None of these patents claim a zero in the feedback loop. |
|
|
| classd4sure |
Hi Pabo,
It is because of what Charles said I believe:
"...Because the feedback thingie is not restricted to polyphase PWM in B&O's patent it does cover almost any PWM topology where the phase-lead is generated by a lowpass around an inner feedback loop.
It does however not cover any other topology that generates a phase-lead.
Regards
Charles
"
How many other ways can you create a phase lead? UCD uses lead compensation... you're right though I don't think it is one of their claims.
Ooooooooooohhhhhhh look.... tripath owns that patent :)
http://www.acutechnology.com/ClassD...S05909153__.pdf
You know .... IMHO, there's far too much concern here over who's got what patented, use what works best! |
|
|
| phase_accurate |
I was aware of that patent but it 1.) mentions the old-fashioned method with the parallel feedback capacicor as prior art (i.e. the patent claims don't cover this) and 2.) the proposed topology is a little complicated IMO. AFAIK they don't use it for their own audio amps. I somehow remember that they wanted to use this principle for RF applications rather than audio but have to think where I got this from.
I just saw that my statement about the feedback around the inner loop was a little unprecise. No one could say anything if you use lowpass-filtered NFB around maybe an OP-AMP for making a phase-lead, even if it is the part of an inner loop of a switching amp. What is claimed is the lowpass-filtered NFB around a modulator stage.
Regards
Charles |
|
|
| Kenshin |
I found the UcD patent WO03090343 at
http://l2.espacenet.com/espacenet/a...DB=EPD&DRDB=EP1
and the feedback do include a zero.
http://l2.espacenet.com/espacenet/b...++03090343A2+I+
| quote: | Originally posted by Pabo
Kenshin
I am not sure what you are referring to specifically but the ucd-patent and the ICEpower-patent are different in the way that ICEpower uses a separate 2nd order filter before the output filter to create the phase modulation while ucd uses the existing 2nd order filter itself.
None of these patents claim a zero in the feedback loop. |
|
|
|
| Kenshin |
Rf//Cf...UcD has it,Tripath has it...and SODA has it....
is it still a valid patent?
if somebody use it commercially,what would happen?
nothing? or be charged of BOTH UcD & Tripath?
|
|
|
| phase_accurate |
A diifferentiating feedback branch as such is a well-known feedback control topology. That part alone would not be enforceable IMO.
Regards
Charles |
|
|
| Kenshin |
what about use 180 deg of phase lag to oscillate?| quote: | Originally posted by phase_accurate
A diifferentiating feedback branch as such is a well-known feedback control topology. That part alone would not be enforceable IMO.
Regards
Charles |
|
|
|
| phase_accurate |
| quote: | | what about use 180 deg of phase lag to oscillate? |
Then you would definitely have to think about a topology that isn't covered by either UcD or ICE. It may be possible but I don't have any usable idea in this direction at the moment.
Regards
Charles |
|
|
| Kenshin |
| quote: | Originally posted by phase_accurate
Then you would definitely have to think about a topology that isn't covered by either UcD or ICE. It may be possible but I don't have any usable idea in this direction at the moment.
Regards
Charles |
But any feedback from the filtered output implies a possibility of phase shift exceeding 180 deg and oscillate--at a frequency of 500KHz, parastical paramters such as delay of the opamp could cause unexpected oscillation easily.
A namely(topologicly) "hysteresis" or "carrier-driven" D-amp (even a namely class AB amplifier)may well be a "180 deg of phase-shift" one in fact. Boundaries between tham are clear theoreticly, ambiguous in schematic (can you tell me how SODA oscillate at a glance on its schemetic?) and almost impossible to recognize in real-world systems.
This means that "180 deg of phase-shift" modulator is rather a combine of parameters than a topology.
So it's probably that enforcing the "180 deg of phase-shift" patent really means patent the feedback from output filter and enforce it. |
|
|
| Kenshin |
Now,audio amplifier industry is now at its "Bakumatsu" time.
The story,is not yet ended....
When achieveing for deep feedback & fast response,it's far easier to let the system oscillate than remain it stable. High-frequency unstable system may even have better perfomance at audio fequencies. Designing linear amplifiers is an art of tradeoff between performance and high frequency stibility.
The boom of self-oscillating D-AMP today is something like the boom of chaos(random oscillations in nonlinear systems) in 1970s~~1980s.
It's first considered something highly advanced, then turns out to be something happening everywhere but we always disliked and tried to prevent.
But at last,there will be some hard core remining. Though building a chaotic system costs only a few lines of code or a few RMBs components, it's hard to understand a chaotic system clearly even today. Optimizing a D-AMP including its nonlinear nature may be far difficult than merely build it.
The mysterious nonlinearity are now causing a dramatic revolution again -- just as it has done in math & physics.
...The new age. |
|
|
| Pabo |
Kenshin
I enjoy reading your posts:)
The ucd topology also oscillates at 180 degress lag, the zero is only there to more accurately set the switching frequency. Relying completely on parasitics for creating 180 degrees phase lag would give bad precision and large temperature dependancy. You could actually also put in a lag in order to more accurately set the frequency but it would lower the frequency which is more impractical. |
|
|
|
