Re: Error correction?
If this arrangement works as advertised, then it would appear to possess the advantage of simplicity with respect to Hawksford's complementary error amplifier, which, it would seem, requires a low source impedance at its output to operate correctly.
Discuss.
mikeks said:
If this arrangement works as advertised, then it would appear to possess the advantage of simplicity with respect to Hawksford's complementary error amplifier, which, it would seem, requires a low source impedance at its output to operate correctly.
Discuss.
Hi, Mikeks,
In patent #4,785,257 in fig.2, there's something strange to me. The idle current of the VAS stage is not well defined? Will this oscilate?
An interesting idea about this patent is that the 2nd differential pair is used for unity gain of the output stage. If we move this 2nd differential to the front end, it will be complementary differential input 😀
In patent #4,785,257 in fig.2, there's something strange to me. The idle current of the VAS stage is not well defined? Will this oscilate?
An interesting idea about this patent is that the 2nd differential pair is used for unity gain of the output stage. If we move this 2nd differential to the front end, it will be complementary differential input 😀
Error Correction Patents
The Iwamatsu patent is just a clumbsy implementation of the Hawksford technique applied to bipolar output stages. Hawksford published his error correction technique in the JAES in January 1981. I think Hawksford may have presented his idea at a convention (with a preprint) a year or so earlier. The original Iwamatsu patent application was filed in Japan in April, 1981. It is inconceivable that he was unaware of Hawksford's work, yet he did not cite it as prior art.
The Hamamatsu patent application was first filed in Japan in June 1986. It is again applicable to bipolar output stages. It is also a variation of the Hawksford scheme where the error correction feedback is injected into the VAS stage. This is a cute idea, but suffers more from stability issues than the Hawksford architecture, so less effective correction out to high frequencies is achievable. He does not cite Hawksford as prior art in the public domain. His Claim 1 reads directly on Hawksford's technique, although in the specification the implementation details are different.
I am not an attorney, but am very familiar with patent law. A basic tenet of patent law is the duty of disclosure imposed on the inventor. It is very unlikely that this second patent would have been issued in its current form had the Hawksford prior art been disclosed to the examiner, as required.
Then, of course, there is the Halcro patent 5,892,398 filed in 1994 which appears to be remarkably similar in EC circuit details to that in my MOSFET power amplifier (even down to the black art of frequency compensation), the only difference being the use of output-based bootstrapping for some of the circuits. The bootstrapping is a legitimately distinguishing feature (it greatly narrows the patent), but the failure to cite the prior art that Hawksford and I thrust into the public domain 10+ years earlier makes the patent very vulnerable to un-enforceability due to the failure in duty of diclosure. Curiously, a second Halcro patent filed in 2000, disclosing a non-Hawksford scheme for distortion reduction, did finally cite the relevant prior art by Hawksford and me. I do not know if any of the Halcro amplifiers use this second scheme, which is basically a bunch of tricks to use an astounding amount of local and global negative feedback using high-order loops.
As many others here have stated, the patent business is extremely imperfect, and depends very heavily on the inventors' exercise of due diligence and honesty in the duty of disclosure. The examiners are overwhelmed by work, and cannot be expected to be up to speed on developments in a field that are not already in the form of patents or pending applications. As a result, many patents get issued that should not be issued, either because they read directly on prior art, or because their distinguishing feature from the prior art is inadequately novel. That's just the way it works.
Bob
mikeks said:
The Iwamatsu patent is just a clumbsy implementation of the Hawksford technique applied to bipolar output stages. Hawksford published his error correction technique in the JAES in January 1981. I think Hawksford may have presented his idea at a convention (with a preprint) a year or so earlier. The original Iwamatsu patent application was filed in Japan in April, 1981. It is inconceivable that he was unaware of Hawksford's work, yet he did not cite it as prior art.
The Hamamatsu patent application was first filed in Japan in June 1986. It is again applicable to bipolar output stages. It is also a variation of the Hawksford scheme where the error correction feedback is injected into the VAS stage. This is a cute idea, but suffers more from stability issues than the Hawksford architecture, so less effective correction out to high frequencies is achievable. He does not cite Hawksford as prior art in the public domain. His Claim 1 reads directly on Hawksford's technique, although in the specification the implementation details are different.
I am not an attorney, but am very familiar with patent law. A basic tenet of patent law is the duty of disclosure imposed on the inventor. It is very unlikely that this second patent would have been issued in its current form had the Hawksford prior art been disclosed to the examiner, as required.
Then, of course, there is the Halcro patent 5,892,398 filed in 1994 which appears to be remarkably similar in EC circuit details to that in my MOSFET power amplifier (even down to the black art of frequency compensation), the only difference being the use of output-based bootstrapping for some of the circuits. The bootstrapping is a legitimately distinguishing feature (it greatly narrows the patent), but the failure to cite the prior art that Hawksford and I thrust into the public domain 10+ years earlier makes the patent very vulnerable to un-enforceability due to the failure in duty of diclosure. Curiously, a second Halcro patent filed in 2000, disclosing a non-Hawksford scheme for distortion reduction, did finally cite the relevant prior art by Hawksford and me. I do not know if any of the Halcro amplifiers use this second scheme, which is basically a bunch of tricks to use an astounding amount of local and global negative feedback using high-order loops.
As many others here have stated, the patent business is extremely imperfect, and depends very heavily on the inventors' exercise of due diligence and honesty in the duty of disclosure. The examiners are overwhelmed by work, and cannot be expected to be up to speed on developments in a field that are not already in the form of patents or pending applications. As a result, many patents get issued that should not be issued, either because they read directly on prior art, or because their distinguishing feature from the prior art is inadequately novel. That's just the way it works.
Bob
The problem is not the invalidity, but the cost to anyone being sued by patent holders to get to the point where invalidity can be considered by a judge. That's fine if you've got a few hundred thousand bucks to spare...
Re: Error Correction Patents
It occurred to me that the error loop in Yokoyama's patent may require some capacitive compensation in the vicinity of transistor 47 (fig. 2).
However, this may not necessarily be the case if the exceptionally heavy degeneration of the error amplifier is taken into account.
It occurred to me that the error loop in Yokoyama's patent may require some capacitive compensation in the vicinity of transistor 47 (fig. 2).
However, this may not necessarily be the case if the exceptionally heavy degeneration of the error amplifier is taken into account.
Hi, Mikeks,
That is indeed the main issue about so called "Error Correction".
Some of the EC scheme (not all offcourse) make the whole amp cct becomes very unstable when EC is attached. The problem disappear if we put compensation cap around the EC mechanism. Until the compensation cap so big, it disables the EC work for rather high frequencies. At this point, what is the usage of that EC? We need EC mainly to shoot the problem of higher spectrum. When compensation needed (to make the whole amp stable), but disables the EC to work to higher frequencies, maybe its better to skip that EC 😀 Just let the ordinary global feedback loop (via input differential) works as usual, but the output stage must be a better one, like classA or heavy biased AB. The main problem mostly will be at the output stage non-linearity.
Yes, another advantage of Yokoyama over Hawksford may be that degeneration rather than heavy reactive compensation is all that's required for error loop stability.
This, however, is just speculation; an existing design may have to ''hacked'' for thorough examination.
The later is far easier to do with this approach than with Hawksford.
This, however, is just speculation; an existing design may have to ''hacked'' for thorough examination.
The later is far easier to do with this approach than with Hawksford.
Hi, Bob Cordell,
I've been wondering about this for a long time, but couldn't find the answer.
In your EC (Hawksford), looking at the EC central there are 2 transistors, that also works as VBE multiplier.
The input signal (from VAS, then to the predriver) is entering the emitors of these transistors. The output condition is entering the base of these transistors. The base is more sensitive than the emitor (emitors need big current, base need only small current to stimulate the transistor). If there is anomalies (like speaker impedance non-linearity, or HF intrusion, or loudspeaker's back EMF), it will be entering the EC system from the base (the sensitive part).
Is there any alternate design of your EC (Hawksford) where the input (from VAS-predriver) is entering a base, and the output signal is entering emitors? The key operation is still the same, that is the Vb-e, but changing the input/output position. I tried to draw it, but until now no success....
I've been wondering about this for a long time, but couldn't find the answer.
In your EC (Hawksford), looking at the EC central there are 2 transistors, that also works as VBE multiplier.
The input signal (from VAS, then to the predriver) is entering the emitors of these transistors. The output condition is entering the base of these transistors. The base is more sensitive than the emitor (emitors need big current, base need only small current to stimulate the transistor). If there is anomalies (like speaker impedance non-linearity, or HF intrusion, or loudspeaker's back EMF), it will be entering the EC system from the base (the sensitive part).
Is there any alternate design of your EC (Hawksford) where the input (from VAS-predriver) is entering a base, and the output signal is entering emitors? The key operation is still the same, that is the Vb-e, but changing the input/output position. I tried to draw it, but until now no success....
lumanauw said:
I find this unattractive: voltage inefficiency.
Another advantage of Yokoyama perhaps?
traderbam said:
They are for biasing, to emulate the real circuit that implements this. For AC, they don't have an effect. One thing to look at in the first sim is "how does the current in V2 and V3 vary as the error correction current changes?".
Here is the input stage separated out. Have a look at V(x) and V👍 vs the swept V1. You should be able to identify the real circuit that this is a semi-idealized model of. The "table" statement gives a piecwise-linear approximation of it.
After a pint of ale it look like an idealised differential amp to me with 270 ohm emitter degen (ignoring Vbe).
Yes, that's it. The idea was to duplicate the bias conditions of an ideal complementary diff amp, and give it zero distortion as long as its dynamic range isn't exceeded. You get a linear I-V characteristic until it hits the stops determined by its tail current.
It shows that error correction can be achieved with complementary diff amps. This eliminates the level-shifting via the zener diode in the Halcro, and the cross-coupled resistor arrangement of Hawksford (which the Halcro also eliminates). Because its input impedance can be made large, it has neglibible loading on associated circuitry. This allows for error-correcting both the driver and output stage. Further, it presents a constant-current load to the follower circuit that precedes it.
Complexity is a big disadvantage though.
It shows that error correction can be achieved with complementary diff amps. This eliminates the level-shifting via the zener diode in the Halcro, and the cross-coupled resistor arrangement of Hawksford (which the Halcro also eliminates). Because its input impedance can be made large, it has neglibible loading on associated circuitry. This allows for error-correcting both the driver and output stage. Further, it presents a constant-current load to the follower circuit that precedes it.
Complexity is a big disadvantage though.
Andy,
I've looked at the whole circuit a little. I see what it does. What were you intending it to do?
I've been acquainting myself with Bob Cordell's MOSFET with error correction paper. I'm not sure I understand fig. 11. In the text it says wrt e(x) that "This error signal is then added to the input of the power stage by summer S2 to provide that distorted input which is required for an undistorted output. Note that this is an error-cancellation technique like feedforward as opposed to an error-reduction technique like negative feedback".
But the diagram doesn't add up to me. If the input to the output stage is x + e(x) then the output must surely be x + e(x) - e{x + e(x)}, rather than x. And doesn't this assume that the e(x) function is linear itself? So I don't see how the output error can ever be cancelled unless e{x + e(x)}=0, which resident mathmaticians may well offer some solutions for, other than e(x)=0.
It may be the ale, but the diagram looks unstable to me, as it were
. The x and e(x) tail chase round the loop reminds me of something...
I've looked at the whole circuit a little. I see what it does. What were you intending it to do?
I've been acquainting myself with Bob Cordell's MOSFET with error correction paper. I'm not sure I understand fig. 11. In the text it says wrt e(x) that "This error signal is then added to the input of the power stage by summer S2 to provide that distorted input which is required for an undistorted output. Note that this is an error-cancellation technique like feedforward as opposed to an error-reduction technique like negative feedback".
But the diagram doesn't add up to me. If the input to the output stage is x + e(x) then the output must surely be x + e(x) - e{x + e(x)}, rather than x. And doesn't this assume that the e(x) function is linear itself? So I don't see how the output error can ever be cancelled unless e{x + e(x)}=0, which resident mathmaticians may well offer some solutions for, other than e(x)=0.
It may be the ale, but the diagram looks unstable to me, as it were
. The x and e(x) tail chase round the loop reminds me of something...
Agreed. Good idea.
In what way...number of parts or the challenge of getting them all to behave like their ideal counterparts in the model?
traderbam said:
Both 🙂.
With regard to Bob's paper, figure 11 holds the keys to the error correction kingdom. The labeling is a bit odd. It uses "x" to refer to two different signals. That may be the source of confusion. Let's change the names of everything.
Let:
out = output of output stage
in = input of output stage
actual_in = input of system
Output stage:
out = in + error
Therefore
out - in = error (diff amp does this)
Let:
in = actual_in - error (where "error" is obtained from the subtraction above)
then
out = actual_in - error + error = actual_in
I have a problem with that last equation. As I see it the two "error"s are not equal unless "error" is independent of "in". Since the "error" is a function of "in" the "actual_in - error" will produce a different "error" at "out".
