always absolute best possible.....
the OLG and OLB DO effect such things as output impedance and distortion above the open loop pole frequency. and it doesn't matter if it's a discrete design or monolithic, the limitations of the input stage's ability to respond to feedback are always going to have the same effect, so it's always best to start with the fewest limitations possible.
the OLG and OLB DO effect such things as output impedance and distortion above the open loop pole frequency. and it doesn't matter if it's a discrete design or monolithic, the limitations of the input stage's ability to respond to feedback are always going to have the same effect, so it's always best to start with the fewest limitations possible.
Real Men Don't Use Opamps
this statement is not true
it is very false!
i see real & very good audio men using both
a real man will use only the best solution
and will use op-amp vs. discrete whenever he finds this is the best solution to his application
fundamentalists are hard to deal with
be it audio or religion
discrete or op-amp fundamentalists are missing something:
the possible optimal solution to a specific application
Lineup - Sweden July 2008
this statement is not true
it is very false!
i see real & very good audio men using both
a real man will use only the best solution
and will use op-amp vs. discrete whenever he finds this is the best solution to his application
fundamentalists are hard to deal with
be it audio or religion
discrete or op-amp fundamentalists are missing something:
the possible optimal solution to a specific application
Lineup - Sweden July 2008
lineup said:
I think Scott means by 'gain at freq of interest' that this gain determines the available feedback. Suppose at your 12kHz the OL gain is 56db, and you want a closed loop gain of 26dB, you have 30dB feedback at 12kHz to straighten out the amp.
Jan Didden
lineup said:
To clarify, let's say an amplifier has 200MHz unity gain crossover and therefore 80dB of gain at 20kHz. Some people think that loading the VAS with a resistor so the DC gain is 80dB and flat to 20kHz has "huge" sonic benefits over the same amplifier with a DC gain of 140dB (by clever design) with the same AC roll off and 80dB (90 degrees of phase in the ideal case) of OLG at 20kHz.
I think this belief has no technical merit at all.
I once modified a Harris (remember them?) HA-911 op amp to have a 20,000 Hz bandwidth from an initial 200Hz bandwidth, AND the IM distortion dropped. How about that? 
John Curl.
This following is a project I have in mind always when talk is about
Open loop bandwidth.
It was a project in my Elektor Magazine.
(I have subscribed since 1985 of this great Netherlands electronics paper.)
This amplifier was designed to reduce IMD to very low level.
The article talked about that by increasing OPEN Loop BW
they could reduce this type of distortion, which is more hostile, bad for human ears.
( We can notice this type of distortion more easy than Harmonic distortion )
And so they did.
This amplifier had very low IMD and still had not very high THD either.
(you have to sacrifice a bit of feedback, with the result that Harmonics dist will go up a bit )
If anyone is interested, I could try to dig up that issue of Elektor
and scan the schematic 😉
What say PMA about IMD in amplifiers and how to cure it ....
I know he is very aware on this subject.
This following is a project I have in mind always when talk is about
Open loop bandwidth.
It was a project in my Elektor Magazine.
(I have subscribed since 1985 of this great Netherlands electronics paper.)
This amplifier was designed to reduce IMD to very low level.
The article talked about that by increasing OPEN Loop BW
they could reduce this type of distortion, which is more hostile, bad for human ears.
( We can notice this type of distortion more easy than Harmonic distortion )
And so they did.
This amplifier had very low IMD and still had not very high THD either.
(you have to sacrifice a bit of feedback, with the result that Harmonics dist will go up a bit )
If anyone is interested, I could try to dig up that issue of Elektor
and scan the schematic 😉
What say PMA about IMD in amplifiers and how to cure it ....
I know he is very aware on this subject.
I have posted a Sim schematic of a discrete amplifier
that had simulated Open Loop -3dB at 20 kHz
The distortion in open loop was very low, probably as I recall it below 0.2 %
But of course this was only a preamp .. not a power amplifier.
The topology was arranged to produce less 3rd than 2nd.
I use so called current feedback for this.
The feedback goes directly into the Emitters of the input complementary pair.
I have this circuit somewhere.
that had simulated Open Loop -3dB at 20 kHz
The distortion in open loop was very low, probably as I recall it below 0.2 %
But of course this was only a preamp .. not a power amplifier.
The topology was arranged to produce less 3rd than 2nd.
I use so called current feedback for this.
The feedback goes directly into the Emitters of the input complementary pair.
I have this circuit somewhere.
john curl said:I once modified a Harris (remember them?) HA-911 op amp to have a 20,000 Hz bandwidth from an initial 200Hz bandwidth, AND the IM distortion dropped. How about that?
Did you bother to analyse what you actually did? Extrapolating this to a rule without analysis, is not very thorough engineering.
I hear from others that even when this makes the amplifier measure worse it still sounds better. Who to believe?
I hear from others that even when this makes the amplifier measure worse it still sounds better. Who to believe?
I measured the difference, but I analyzed it by hand, like all we old geezers used to have to do. It is fairly straight forward, but the results were a surprise. After all, how could I reduce the negative feedback by 40 dB and get better IM? Perhaps I linearized the output stage.
Me I had once made an op amp line preamp after I measured and listened to about 20 types. It was old time, the OPA627 was just new in BB brochures. I made a special order for 627 and 637. I think it cost me $150 in 1992 Dollars. Fortunately it proved the best sounding chip back then. Very nice preamp it ended up to be. I used it for long. Then I made one with just 2 JFETS per channel. Much to my dismay for my former effort the JFET pre just sounded so much better. What the heck, I told my self, make one with a single triode and a triode buffer. So much better it sounded again. And now I made one with just an output duty beam tetrode wired as triode, good current (25mA) with no buffer. It opened up much more again. I just hate it, that whatever I do logically with technical excellence in mind, underachieves musically versus the simplest, most dated stuff.
I understand your experience, Salas. I had a similar experience, when I replaced a discrete, 100V/us trans-amp with a pair of jfets as a phono input amplifier. This sounded better, even with A-B testing. This ultimately developed into the JC-80 phono input stage, and then into the Vendetta Research input stage, which is the best sounding input amplifier that I have ever designed. Why? I don't really have a complete understanding of why, but the Vendetta input stage will beat the socks off any op amp that I can design or buy, even today, so far as I know, UNLESS the op amp is operated open loop (hint, hint) and then a good design can give me a run for my money.
john curl said:I once modified a Harris (remember them?) HA-911 op amp to have a 20,000 Hz bandwidth from an initial 200Hz bandwidth, AND the IM distortion dropped. How about that?
I found the datasheet online - vertical dielectric isolated op amp in 1975??
http://www.datasheetarchive.com/preview/1526866.html
the datasheet schematic shows a quasi-comp output stage with a lower "pnp" cfp follower - I guess the pnp must still not have equaled the npn in that process
the frightening thing for audio quality is the 500 Ohm typ output Z - woefully underbiased Class AB is known to not be the best idea with cfp
negative feedback shouldn't asked to "fix" the expected dynamic deadzone expected when trying to commutate the weakly biased quai-comp output
I would expect to see major improvement from Class A biasing the output
I wouldn't expect significant distortion improvement from just resistively loading the BW pin - but it is possible that "tuning" the loop gain and phase could cause some specific cancellation of a few distortion products
JCX, the HA-911 is what I found in 1970 as the answer to a good quality audio amp. Independently, Dick Burwen found it as well, and used it in all his designs, including the LMP-2 preamp, that he designed for Mark Levinson. In his case, he screened the op amps for xover distortion and used the rejects for signal processing rather than audio through-path. His selected IC's were pretty darn good, but were ultimately be replaced by discrete op amps and trans amps that I designed.
In any case, it would be first supposed that feedback can fix just about anything, and less feedback would make MORE IM rather then LESS IM.
I think that you are on the right track. It is the beta non-linearity of the output stage and its extremely high drive impedance that made a lower drive impedance (1 meg ohm) better than its nominal drive impedance, even with a reduction in open loop gain at low frequencies.
It is important to point this out, just because 'hard and fast' rules sometimes fail, and you just can't assume everything in advance.
The same thing happens with listening to differences. You just can't always predict what is audible and what is not. We are still scratching our heads about hi feedback op amps, and why they are less than perfect for audio circuits
In any case, it would be first supposed that feedback can fix just about anything, and less feedback would make MORE IM rather then LESS IM.
I think that you are on the right track. It is the beta non-linearity of the output stage and its extremely high drive impedance that made a lower drive impedance (1 meg ohm) better than its nominal drive impedance, even with a reduction in open loop gain at low frequencies.
It is important to point this out, just because 'hard and fast' rules sometimes fail, and you just can't assume everything in advance.
The same thing happens with listening to differences. You just can't always predict what is audible and what is not. We are still scratching our heads about hi feedback op amps, and why they are less than perfect for audio circuits
john curl said:
The flatpack HA909 in my Alembic guitar has a noble heritage, and still sounds pretty good to this day, albeit as a producer, not reproducer.
john curl said:
John Curl
Very good post, John.
And very right you are again:
We are still scratching our heads about hi feedback op amps
and why they are less than perfect for audio circuits.
Ever since Matti Otala tried some other ways than Highest Possible gain & feedback,
there has been a debate of what is the optimum gain for a global negative feedback correct amplifier.
Noone has still come up with the perfect answer.
So keep on scratching our heads .. is what we will do 😀
Even Scott Wurcer has scratched his head many times.
He is almost bald headed by now
I guess
I've given up scratching and started looking for an invite to hear the same things everyone else seems to be hearing. Still no offers. I'll never bother to put together a system with enough resolution myself.
BTW I'm thinking of registering "Three leads good eight leads bad!" and offering T-shirts at the Burning Amp.
jcx - Your guess is along my first thoughts. I've seen fortuitous cancellation effects many times. You tweak something a little here or there and it comes and goes. A single datapoint is meaningless.
BTW I'm thinking of registering "Three leads good eight leads bad!" and offering T-shirts at the Burning Amp.
jcx - Your guess is along my first thoughts. I've seen fortuitous cancellation effects many times. You tweak something a little here or there and it comes and goes. A single datapoint is meaningless.
- Status
- Not open for further replies.