Stopping the integrator during clipping....
..identical headaches in self oscillating designs and clocked as well...
The struggle with the integrator during clipping catches us all.
@ Bruno:
Figure 5 of your patent shows three embodyments of the integrator.
5A, zenering the cap: Obviously the most traditional solution, but not fortunate regarding distortion. Your excellent distortion values indicate that this is just a theoretic option, but not your real solution...
5B, shorting the cap: I guess - even if done not by a hard short but resistive, this solution might be not always act perfectly calm.
5C, disconnecting the integrator from the input:
That's temptating ! Do you need any kind of a parallel signal path?
Or do you fully rely on the information of the deviation detection circuit and simply reconnect the integrator again as soon as the deviation detection circuit 'thinks' that the output would now/soon go back below clipping? ....starting to love this idea....
BTW:
Instead of clamping the cap with zeners like in figure 5A, is there a reason
not to clamp the output of the integrator stage towards GND or against reference levels? In this configuration the low output impedance of the OP-Amp circuit theoretically keeps diode related distortions low until the voltage really comes close to the clamping voltage.
(Or if possible: Simply adjusting the OP amp supply, so that the integrator cannot run to far away).
..identical headaches in self oscillating designs and clocked as well...
The struggle with the integrator during clipping catches us all.
@ Bruno:
Figure 5 of your patent shows three embodyments of the integrator.
5A, zenering the cap: Obviously the most traditional solution, but not fortunate regarding distortion. Your excellent distortion values indicate that this is just a theoretic option, but not your real solution...
5B, shorting the cap: I guess - even if done not by a hard short but resistive, this solution might be not always act perfectly calm.
5C, disconnecting the integrator from the input:
That's temptating ! Do you need any kind of a parallel signal path?
Or do you fully rely on the information of the deviation detection circuit and simply reconnect the integrator again as soon as the deviation detection circuit 'thinks' that the output would now/soon go back below clipping? ....starting to love this idea....
BTW:
Instead of clamping the cap with zeners like in figure 5A, is there a reason
not to clamp the output of the integrator stage towards GND or against reference levels? In this configuration the low output impedance of the OP-Amp circuit theoretically keeps diode related distortions low until the voltage really comes close to the clamping voltage.
(Or if possible: Simply adjusting the OP amp supply, so that the integrator cannot run to far away).
You'll all understand I'm not going to disclose details so if I respond (very!) selectively to discussions about the guts of the amp as relates to the patent drawings that's entirely intentional.
CH: Clamping the output of an op amp amounts to shorting it. You could diode clamp the high z node of a folded cascode op amp but then they would make it distort quite heavily again.
CH: Clamping the output of an op amp amounts to shorting it. You could diode clamp the high z node of a folded cascode op amp but then they would make it distort quite heavily again.
What would be the difference between this and the diode clipper - apart from the fact that the diode clipper doesn't saturate "internally" (and would therefore behaves better) ?
IMO at least the diode-clipper and the reset with the switch are not actually novel since they have been used in noise-shapers before. Maybe the fact that these are used in a non delta-sigma style modulator is novel enoght - who knows ?
Regards
Charles
...to bad that we are touching confidential know how which you are making your living of...
But I definitely love the clear way of your answer - already noticed similar upright comments on commercial reality from you earlier in this thread.
Obviously yes. But most types are designed to handle not just clamping, but even a fully shorted output for unlimited time. The chip designer has spend some love to this feature (at least he copied the building block...
) - wouldn't it be a pity if nobody wants to use it?The zener clipper in the feedback is adding the worst amount of distortion, because in this configuration already few micro amps in the diode do massively influence the output. Means if you use two 4V7 zeners in the feedback, it will dominate the over all distortions already at 2V may be even earlier. If the zeners are placed from output to ground the distortions remain low until the diode current reaches the mA range and really starts to clamp.
The saturating limitation by the supply also is by far more 'sharp' than the zener feedback, and it is for free. In fact the clipping recovery of modern OP amps is fast enough for usual integrators in class D amps.
Well, patent laws and interpretation do not fit into my simple brain.
Without a lawyer and a powerful company I would not even start to file a single patent.
If you have time and money - feel free to fight Bruno's patent. It is filed, but not granted by now.
For me at least the idea around figure 5C is new and really nice!
Last edited:
Is this nonlinearity that bad ? Remeber that you introduce the integrator in the forward path and it is mainly there to increase loop gain not to introduce linearity in simple words.
Apart from that - what you dislike about the circuit might as well be optimised with some tricks.
Regards
Charles
Apart from that - what you dislike about the circuit might as well be optimised with some tricks.
Regards
Charles
...or is it already granted in NL & EP but not in US???
I am not sure. Nomenclature of patent numbers is not my field of expertise...
And definitely I am not interested at all in making headache to Bruno.
He is contributing a lot to audio tech in general and to the DIY community.
I am glad that Bruno shows this topic to us and discloses one or the other idea at least.
I am not sure. Nomenclature of patent numbers is not my field of expertise...
And definitely I am not interested at all in making headache to Bruno.
He is contributing a lot to audio tech in general and to the DIY community.
I am glad that Bruno shows this topic to us and discloses one or the other idea at least.
...erhm, well... for me it is bad if my integrator shows ten times the k3 of my switching stage. Even if I get the nice benefit of higher loop gain at lower frequencies- I do not want to pay for that with a dominant degradation of the forward path.
So far I did not find tricks to bring down the zeners in the feedback to comparable distortion levels like clamping the OP amp output to GND or a ref.
Am I blind?
Instead of Zeners you may also use "normal" diodes in antiparallel configuration. Since you will need more than one antiparallel pair in series in order to achieve a higher clipping voltage you can shunt a node between two pairs to ground with a reasonably sized resistor. This should improve low-level linearity. I assume that the idea is not new - but maybe I just gave away a patentable idea (well - wouldn't be for the first time anyway).
Regards
Charles
Regards
Charles
...agreed, this should be better than the zeners.
Furtheron we might use BJTs (base connected to collector) instead of diodes.
But by the end of the day we start to throw around with ten components instead of two zeners. And I am not confident that its distortion level could rival the zeners from the OP amp output to GND.
Furtheron we might use BJTs (base connected to collector) instead of diodes.
But by the end of the day we start to throw around with ten components instead of two zeners. And I am not confident that its distortion level could rival the zeners from the OP amp output to GND.
If we were all in the pub I'd probably be a lot more forthright because I'd know who's listening. But anything I write here can be read by anyone. Most people who turn up at the bottom of the page ("Currently Active Users Viewing This Thread: 8 (6 members and 2 guests)") don't ever post anything so who knows who they are.
It's not too hard to get your head around the principles of patent law, but you do need a patent lawyer to explain things. The usefulness of patents is generally overstated. Protection isn't automatic - you still have to take copycats to court yourself. A patent is just evidence that you were there first. Take out a patent only if you are prepared to defend it. It does have some business value in that it shows someone's been checking, to the best of their ability, that the technology you're selling is truly your own. But the best protection is still staying ahead of the competition and making sure that if someone gets inspired by an idea you've had, they do so only when they see it in mass production.
About granting. It takes ages getting a patent granted, and it's done country by country. However, if you apply in Holland first for instance and in the US only a year later, the USPTO will not consider publications in the intervening time as "detrimental to novelty". What matters is the date of first filing in any country. That way you don't have to sink a huge amount of cash into the process in one go.
About what makes valid stuff for patenting. Every invention is based on something that went before. For instance, there are diodes in Ncore. Would it help to say "well DUH you're just using diodes to solve the problem. Diodes are old hat"? That's obviously ridiculous but the very same thing goes for more complicated standard techniques applied creatively. Integrators have been saturated before. That's not the point. What patent examiners look for is whether the sum of the parts is a new and unexpected way to solve a problem. These criteria are called novelty and inventive step respectively.
That's about all the patent banter I've got energy for.
It's not too hard to get your head around the principles of patent law, but you do need a patent lawyer to explain things. The usefulness of patents is generally overstated. Protection isn't automatic - you still have to take copycats to court yourself. A patent is just evidence that you were there first. Take out a patent only if you are prepared to defend it. It does have some business value in that it shows someone's been checking, to the best of their ability, that the technology you're selling is truly your own. But the best protection is still staying ahead of the competition and making sure that if someone gets inspired by an idea you've had, they do so only when they see it in mass production.
About granting. It takes ages getting a patent granted, and it's done country by country. However, if you apply in Holland first for instance and in the US only a year later, the USPTO will not consider publications in the intervening time as "detrimental to novelty". What matters is the date of first filing in any country. That way you don't have to sink a huge amount of cash into the process in one go.
About what makes valid stuff for patenting. Every invention is based on something that went before. For instance, there are diodes in Ncore. Would it help to say "well DUH you're just using diodes to solve the problem. Diodes are old hat"? That's obviously ridiculous but the very same thing goes for more complicated standard techniques applied creatively. Integrators have been saturated before. That's not the point. What patent examiners look for is whether the sum of the parts is a new and unexpected way to solve a problem. These criteria are called novelty and inventive step respectively.
That's about all the patent banter I've got energy for.
- Status
- Not open for further replies.