Oh really? I don't know. I don't think I've read Electronics World since I was a teenager. They have a website. Phase lead comp is nothing special - it's in all the books. I imagine lots of designs use it...Naim for example.
traderbam said:
Hmm, maybe I've got the wrong name for the mag? The one that Baxandall, Self and others wrote articles for is what I'm thinking of. I think it was once called "Wireless World", then changed to "Electronics and Wireless World", then "Electronics World". I may be confused about one or more of these names though, as the mag is rarely seen here in the states.
Regarding the lead comp, in order to eliminate the overshoot of TPC, the zero of the lead comp had to be at a frequency somewhat lower than what one would ordinarily use. That ended up increasing the ULG frequency by 2x or so. Kind of risky I think.
Wireless World. I remember that. Maybe I didn't read EW as a teenager, but I did read Wireless World on occasion. The names change now and again over the years and when acquisitions occur.
I don't know what mikeks' exact scheme was, but I don't believe the ULG f needs to increase. You are daring me to do the maths, aren't you?
Hopefully, Glen will simulate it. 😀
I don't know what mikeks' exact scheme was, but I don't believe the ULG f needs to increase. You are daring me to do the maths, aren't you?
Hopefully, Glen will simulate it. 😀
traderbam said:
Nope 🙂. I tried Mike's technique in the simulator, but I wasn't able to do it without increasing the ULG freq by more than I wanted to.
andy_c said:
Hi Andy
No, you are not confused at all. His article hasn't been published in WW/EW/E+EW, because:
1. It was far too long.
2. He wanted too much money for it.
Regards,
Edmond.
andy_c said:
Well you don't get something for nothing. I've done a sim now. I guess the original zero moves down in f so some of the excess gain is sacrificed for stability. A few iterations has me with a FB zero at 122kHz and pole at 282kHz, ULG still at 1MHz. A little underdamped but no overshoot on the step. 650ns to 63%.
Personal communication.
Cheers,
Edmond.
edit: By the time he offered his article, WW/EW didn't pay contributors any longer anyhow.
Cheers,
Edmond.
edit: By the time he offered his article, WW/EW didn't pay contributors any longer anyhow.
Re: Re: Re: Re: Down with NDFL
Doesn't help the slew rate either. The most compromised part of Cherry's amp (I presume you have the ETI 60W amp module) I found was his VAS.
You can see in my schematic that I loaded the Rush current amplifier / intergrator with a current source and used an EF buffered VAS. That made a huge improvement to the VAS linearity by maximising the feedback via Cdom.
Cherry didn't do this. He just used a single transistor VAS and loaded the Rush intergrator with a 220 ohm resistor, which really made the VAS peformance mediocre. The resultant low feedback factor I'm guessing is why his VAS was stable with Cdom incorporating the output stage.
Cheers,
Glen
EDIT:
Not the schematic you linked to - the previous one.
Edmond Stuart said:
Doesn't help the slew rate either. The most compromised part of Cherry's amp (I presume you have the ETI 60W amp module) I found was his VAS.
You can see in my schematic that I loaded the Rush current amplifier / intergrator with a current source and used an EF buffered VAS. That made a huge improvement to the VAS linearity by maximising the feedback via Cdom.
Cherry didn't do this. He just used a single transistor VAS and loaded the Rush intergrator with a 220 ohm resistor, which really made the VAS peformance mediocre. The resultant low feedback factor I'm guessing is why his VAS was stable with Cdom incorporating the output stage.
Cheers,
Glen
EDIT:
Not the schematic you linked to - the previous one.
Edmond Stuart said:
I think the amount Mike wanted and other related info is personal to Mike.
Jan Didden
Re: Re: Re: Re: Re: Down with NDFL
Hi Glen,
First, if you are through with NDFL please ignore this post.
Although that 220 (actually 330) Ohm resistor does degrade the performance, this is what Cherry wrote about it:
"Replacing this resistor by a current source would be an expensive way of achieving half the sensitivities."
Now I wonder, were trannies in those days that expensive (I can't remember) or is it just an example of false economy?
One more remark: You have replaced the Rush amp by an integrator plus CCS. I'll take it that the distortion is lower now. But the time constant Ry*Cy has gone as well. So, I wonder if it is still a NDFL amp such as Cherry has meant. Did you get the same step response as from the original circuit?
Regards,
Edmond.
G.Kleinschmidt said:
Hi Glen,
First, if you are through with NDFL please ignore this post.
Although that 220 (actually 330) Ohm resistor does degrade the performance, this is what Cherry wrote about it:
"Replacing this resistor by a current source would be an expensive way of achieving half the sensitivities."
Now I wonder, were trannies in those days that expensive (I can't remember) or is it just an example of false economy?
One more remark: You have replaced the Rush amp by an integrator plus CCS. I'll take it that the distortion is lower now. But the time constant Ry*Cy has gone as well. So, I wonder if it is still a NDFL amp such as Cherry has meant. Did you get the same step response as from the original circuit?
Regards,
Edmond.
Re: Re: Re: Re: Re: Re: Down with NDFL
Hmmm... It's 220 ohms in the amp of the Jan 1983 issue of ETI. You must have a schematic for a different version.
In this version his VAS was not buffered by an EF, so I guess he is right about the CCS not giving a huge improvement - the VAS input impedance would still rather low and feedback via Cdom limited. With an EF buffer for the VAS it is another story though.
As a matter of fact the amps performance was a bit average overall. The sub-optimal VAS coupled with the extra distortions introduced by that Rush current amplifier intergator stage completely negated all the extra loop gain provided by the NDFL loop and then some.
For example, at only 6kHz at full power the 2nd harmonic measured 115ppm and the 3rd 100ppm - far worse than a basic "Blameless" with a double EF output stage using the same 2MHz fT power transistors.
Regarding my replacement for the Rush current amp, that was just an experiment to test the theory. The Ry Cy constant that makes the Rush current amplifier an intergrator is still there (provided by Cdom on the 1st VA stage). Step response and loop gain / phase are identical.
It is interesting to note however that a 2nd order NDFL does not have the overshoot issue of TPC.
Cheers,
Glen
Edmond Stuart said:
Hmmm... It's 220 ohms in the amp of the Jan 1983 issue of ETI. You must have a schematic for a different version.
In this version his VAS was not buffered by an EF, so I guess he is right about the CCS not giving a huge improvement - the VAS input impedance would still rather low and feedback via Cdom limited. With an EF buffer for the VAS it is another story though.
As a matter of fact the amps performance was a bit average overall. The sub-optimal VAS coupled with the extra distortions introduced by that Rush current amplifier intergator stage completely negated all the extra loop gain provided by the NDFL loop and then some.
For example, at only 6kHz at full power the 2nd harmonic measured 115ppm and the 3rd 100ppm - far worse than a basic "Blameless" with a double EF output stage using the same 2MHz fT power transistors.
Regarding my replacement for the Rush current amp, that was just an experiment to test the theory. The Ry Cy constant that makes the Rush current amplifier an intergrator is still there (provided by Cdom on the 1st VA stage). Step response and loop gain / phase are identical.
It is interesting to note however that a 2nd order NDFL does not have the overshoot issue of TPC.
Cheers,
Glen
220/330 Ohms
Indeed, I was referring to a different version, published in JAES, Vol. 30, No. 5, 1982 May. Anyhow, I don't think it really matters.
Cheers,
Edmond.
Indeed, I was referring to a different version, published in JAES, Vol. 30, No. 5, 1982 May. Anyhow, I don't think it really matters.
Cheers,
Edmond.
Re: Re: Re: Re: Re: Re: Re: Down with NDFL
That's because it doesn't introduce zeros into the CL transfer function.
Have you considered phase lead compensation for your TPC?
Brian
G.Kleinschmidt said:
That's because it doesn't introduce zeros into the CL transfer function.
Have you considered phase lead compensation for your TPC?
Brian
NDFL
BTW Glen, did you know that by means of a few minor modifications an Alexander like amp can be transformed to a NDFL thingie?
If you are really interested I will dig in the files of an old PC of mine, though it'll take some time.
Cheers,
Edmond.
BTW Glen, did you know that by means of a few minor modifications an Alexander like amp can be transformed to a NDFL thingie?
If you are really interested I will dig in the files of an old PC of mine, though it'll take some time.
Cheers,
Edmond.
Re: NDFL
But if I did that wouldn't it just be a NDFL? BTW, I'm using TMC, not TPC 😉
I've never really looked at the Alexander amp, but when you find your files post them up, they should be interesting.
Cheers,
Glen
traderbam said:
But if I did that wouldn't it just be a NDFL? BTW, I'm using TMC, not TPC 😉
Edmond Stuart said:
I've never really looked at the Alexander amp, but when you find your files post them up, they should be interesting.
Cheers,
Glen
Re: Re: NDFL
I picked up the wrong end of the wrong stick by sticking my nose in! I admit I haven't read the thread in much detail and don't know why you were discussing TPC.
Anyhow, my understanding is that:
TPC - 2 pole compensation involves rolling off the fwd transfer function at 12dB/oct then pulling it back with a zero to 6dB/oct before the ULG f.
TMC - "transitional miller compensation" (c. Ed) is more complicated and involves taking the VAS/miller stage FB from the output at low f and then from some earlier point in the circuit at high f. Originally published by Cherry.
NDFL - is about wrapping NFB circuits within other NFB circuits, each outer circuit having a lower UGB than the inner one. Originally published by Cherry.
If that's right, then TPC with phase lead compensation is quite different from NDFL. NDFL is more bandwidth efficient in terms of poles/stages per increase in excess gain. TPC without compensation will always overshoot. In Ed's TPC example that Andy and I were analysing, the compensation to eliminate overshoot that I simulated is a network in the FB path consisting of 1.3k||1nF in series and 1k in shunt.
You seemed to be arguing against TMC and in favour of TPC. I haven't read back far enough to know why you were doing that - but arguing with Ed is good sport so who cares 😀 . Ed complained about the overshoot. Your rebuttal was to put an LPF on the amp input. Ed said this was pants because it ignored the problem. I then said a filter is needed in the FB path. Ed said true but it requires an unacceptable extension of the ULG f. I then simulated a solution that has the same ULG f, no overshoot and costs some 4dB of excess gain.
G.Kleinschmidt said:
I picked up the wrong end of the wrong stick by sticking my nose in! I admit I haven't read the thread in much detail and don't know why you were discussing TPC.
Anyhow, my understanding is that:
TPC - 2 pole compensation involves rolling off the fwd transfer function at 12dB/oct then pulling it back with a zero to 6dB/oct before the ULG f.
TMC - "transitional miller compensation" (c. Ed) is more complicated and involves taking the VAS/miller stage FB from the output at low f and then from some earlier point in the circuit at high f. Originally published by Cherry.
NDFL - is about wrapping NFB circuits within other NFB circuits, each outer circuit having a lower UGB than the inner one. Originally published by Cherry.
If that's right, then TPC with phase lead compensation is quite different from NDFL. NDFL is more bandwidth efficient in terms of poles/stages per increase in excess gain. TPC without compensation will always overshoot. In Ed's TPC example that Andy and I were analysing, the compensation to eliminate overshoot that I simulated is a network in the FB path consisting of 1.3k||1nF in series and 1k in shunt.
You seemed to be arguing against TMC and in favour of TPC. I haven't read back far enough to know why you were doing that - but arguing with Ed is good sport so who cares 😀 . Ed complained about the overshoot. Your rebuttal was to put an LPF on the amp input. Ed said this was pants because it ignored the problem. I then said a filter is needed in the FB path. Ed said true but it requires an unacceptable extension of the ULG f. I then simulated a solution that has the same ULG f, no overshoot and costs some 4dB of excess gain.