good question, lineup
as far as I understand this, assuming second stage is dominantly slow, than amp A has probably more linear input stage and therefore is less prone to tim and pim. On the other hand higher second stage gain may suffer from nonlinear loading of VAS by output stage and therefore higher crossover/GMdobling/classB distortion.
BTW, by audio spectrum borders I meant around 20kHz
best regards
as far as I understand this, assuming second stage is dominantly slow, than amp A has probably more linear input stage and therefore is less prone to tim and pim. On the other hand higher second stage gain may suffer from nonlinear loading of VAS by output stage and therefore higher crossover/GMdobling/classB distortion.
BTW, by audio spectrum borders I meant around 20kHz
best regards
lineup said:
I almost never use very large bandwidth.
Because I cant hear too high frequencies
and if i could, most 95% of music and sounds are within the audible spectra.
hehe - obvious, isnt it ....
But lets say my target values for amplifiers are:
10 - 100/200 kHz.
The actual value is of course dependent on
what kind of amplifier I will build.
A power amplifier seldome needs the bandwidth of a preamp, for example.
But I am more into the policy of making very good open loop performance ( a lot of emitter resistors )
within a more narrow bandwidth,
than being hunting for enormous figures in terms of high freq.
Very high freq may look good at your oscilliscope screen
with a pure resistive load,
but in real working environment such circuit can run into serious troubles.
One simple issue is radio waves transmitters causing unwanted effects!
Now the mentioned 100x, 40dB figure is not an absolute number for me.
It can be different depending what type of amp.
But what I try to avoid is MAXIMAL open loop gain.
---------------------
By the way, a question:
What is the lowest negative feedback factor possible?
If you want to use any negative feedback for correcting the signal.
In my mind it is 6dB.
Open loop gain is 2 x closed gain.
Is this correct?
At least I have drawn several such pre-amplifiers, on my papers.
You can't treat this all separated. Low bandwidth means early roll-of means phase shift means oscillations from high feedback. A statement like "I aim for 40dB feedback" is pretty meaningless unless you also specify your ol gain, ol freq response, cl gain etc.
All these factors are initimitely connected and I can only advise to read a good text on feedback amplifiers before actually yo put something together.
Jan Didden
traderbam said:
You phone up the genius who worked for Ferrari and ask him to design a better V12 engine for you than the one Ferrari has: Giotto Bizzarini !
Then you do not build a Ferrari, but a Lamborghini.
Some prefer Self and Ferrari boxes, others a Feruccio or a Pass box. Where's the truth ?
Miles Prower said:
Such as ? (given the nasty crossover behaviour of this device)
jacco vermeulen said:
Since you mentioned it, I have a couple of ideas in the works. Since the MOSFET has a wicked crossover characteristic, keep it out of Class AB operation, or at least don't push it too far into AB. One design idea involves the use of a Source Follower and Inductance loaded parafeed. Unlike VTs, this is a low impedance device, so that the required inductance doesn't amount to much: ~100mH will do nicely, and a coil for that won't be real difficult to build either. This being a Class A, obviously. Using the MOSFET as a SF also goes a long way to solving the problems that a large, and nonlinear, Cgs presents. Keep the DC resistance low enough, and you won't even need a coupling capacitor.
Of course, that'll mean some hefty iron, beefy heatsinks, and a power supply to match. This is DiY, and we're not constrained to designing for maximum cost reduction.
Another possibility is to use push-pull power MOSFETs in a transformer loaded topology. That way, both devices can be either N-Channel or P-Channel, thus avoiding the problems of complimentary pairs that aren't so "complimentary" in practice. Again, connect as SF's. Given a supposed 32R across the primary, you once again simplify the construction of the matching xfmr, as compared to a VT OPT, with the much larger impedances, primary inductance, far more turns required with the stray capaciatnce problems all that bring.
I'll do that someday.
Attachments
Which way is "right"
In a perfect world, a "perfect" (what this means, below) amplifier be it high global NFB, medium, bipolar, Mosfet, A, AB, single ended, vacuum tube or whatever topology, technology or combination, shoud sound the same.
Those who think there are - now or ever - unmeasurable performance parameters affecting sonics, need not keep reading.
But perfect worlds do not exist, neither perfect amplifiers. What we can do at most is settle on *provisional* target specifications subject to change as we learn more, particularly regarding how our ears aquire information and our brains process it.
If these specifications fail to be below the threshold of perception for the entire universe of listeners, then different design and implementation approaches measuring equal will obvioulsy sound different. Bad solid state high feedback designs will suffer from dense low level harmonic and intermodulation products among other gremlins. SE or for that matter designs without sufficient correction to inherent active device nonlinearity will exhibit significant harmonics and intermodulation though patterned in an acceptable form and so on.
In the real world average listeners will not perceive what an elite of gifted by nature and nurtured by experience do. And of course we must figure in human nature, leaning to perceived values like price or looks or brand or trusted reviewer opinion.
May be there will be a perfect amplifier, one that passes all tests setting performance safely below the most acute listener threshold, yet may be not accepted by everyone. For a perfect amplifier certainly will not add by itself components some find of benefit to the experience.
Rodolfo
In a perfect world, a "perfect" (what this means, below) amplifier be it high global NFB, medium, bipolar, Mosfet, A, AB, single ended, vacuum tube or whatever topology, technology or combination, shoud sound the same.
Those who think there are - now or ever - unmeasurable performance parameters affecting sonics, need not keep reading.
But perfect worlds do not exist, neither perfect amplifiers. What we can do at most is settle on *provisional* target specifications subject to change as we learn more, particularly regarding how our ears aquire information and our brains process it.
If these specifications fail to be below the threshold of perception for the entire universe of listeners, then different design and implementation approaches measuring equal will obvioulsy sound different. Bad solid state high feedback designs will suffer from dense low level harmonic and intermodulation products among other gremlins. SE or for that matter designs without sufficient correction to inherent active device nonlinearity will exhibit significant harmonics and intermodulation though patterned in an acceptable form and so on.
In the real world average listeners will not perceive what an elite of gifted by nature and nurtured by experience do. And of course we must figure in human nature, leaning to perceived values like price or looks or brand or trusted reviewer opinion.
May be there will be a perfect amplifier, one that passes all tests setting performance safely below the most acute listener threshold, yet may be not accepted by everyone. For a perfect amplifier certainly will not add by itself components some find of benefit to the experience.
Rodolfo
traderbam said:That's a little hard on Nelson
You want to design and build your own high performance sports car. Where do you go? Do you visit the guy down the road in the small shop which builds go-carts and bicycles and sells a book entitled: "high performance road vehicle design guide". No. Instead, you phone up the design department of Ferarri (or the design house you most respect) and ask them to share their expertise with you. They refuse but at least they do tell you exactly where to go!
Such is the nature of performance audio.
umm you ask ferrari to share the expertise, well thats one way
of doing it,the other you drop a rocket engine into a go-cart
and you have a super-high performance, but whats in an idea
design departments such as ferrari didnt start with an idea
they went to formula-one and asked for help, if you have the
funds you can play if you havent youre stuck with bicycles
john
DougSelf needs high OL gain (=high feedback) because he use class B (low biased) output stage. Output stage (in low biased mode) is the main part that has to be fixed by NFB. The differential and VAS usually already run in classA, not much to fix there. Mosfet output stage that runs in classB needs more effort to make right than bipolars.
The one that has to be fixed is the output stage, but usually the gain is made more and more in the front stage (differential+VAS). This has a danger. If the high gain is made in front stage and this huge gain hardly make it to the output stage (because of the bottlenecks in the process from the front stage to output stage), the sound will not be nice.
ClassB is appealing because of low heat radiation.
The better solution is to make local FB around output stage. Something like local output stage Error Correction or like the attachment below (this cct is just recently submitted in this forum). In the below attachment, BUZ21 output is made linear output (comparison of +input of OPA and Source of BUZ21) by FB with OPA627 (also BUZ171 with OPA131). High feedback occurs only around the output stage, means the source of BUZ21 will try to follow linearly +input of OPA627, OPA627 is doing this by burning it huge internal OL gain just for leveling voltage at +input/OPA627 with source of BUZ21 (linear follower).
The front end just have to give little effort only to drive +input of OPA627, need not to fix the BUZ21 non linearity.
In this mode, the front end not need to have huge gain, and huge gain to linearize BUZ21 is locally done by OPA627, not by front end.
If we parrarel 5 BUZ21's, then each BUZ21 will have its own OPA627 (5 OPA's for 5 BUZ's).
To distribute gain like when using local EC likely to make better sound for low biased output stage, rather than centralizing all the huge gain only in differential+VAS.
The one that has to be fixed is the output stage, but usually the gain is made more and more in the front stage (differential+VAS). This has a danger. If the high gain is made in front stage and this huge gain hardly make it to the output stage (because of the bottlenecks in the process from the front stage to output stage), the sound will not be nice.
ClassB is appealing because of low heat radiation.
The better solution is to make local FB around output stage. Something like local output stage Error Correction or like the attachment below (this cct is just recently submitted in this forum). In the below attachment, BUZ21 output is made linear output (comparison of +input of OPA and Source of BUZ21) by FB with OPA627 (also BUZ171 with OPA131). High feedback occurs only around the output stage, means the source of BUZ21 will try to follow linearly +input of OPA627, OPA627 is doing this by burning it huge internal OL gain just for leveling voltage at +input/OPA627 with source of BUZ21 (linear follower).
The front end just have to give little effort only to drive +input of OPA627, need not to fix the BUZ21 non linearity.
In this mode, the front end not need to have huge gain, and huge gain to linearize BUZ21 is locally done by OPA627, not by front end.
If we parrarel 5 BUZ21's, then each BUZ21 will have its own OPA627 (5 OPA's for 5 BUZ's).
To distribute gain like when using local EC likely to make better sound for low biased output stage, rather than centralizing all the huge gain only in differential+VAS.
Attachments
not exactly recent .
www.moehrenbude.de/Moehre/images2/download/No_Crossoveramp_W.Meusberger.pdf
www.moehrenbude.de/Moehre/images2/download/No_Crossoveramp_W.Meusberger.pdf
Jan, Jacco,
You guys are very good at languages - no question.
But if you really want to see speed language learning, examine David Lumanauw's written English over the last two years.
You both know Indonesia pretty well, you would know that only in the last generation has English been taught, and that the 1st language is emphatically Bahasa, with a polyglot of other widely used languages, particularly Javanese and Sundanese.
David, take a bow!!
Cheers,
Huigh
You guys are very good at languages - no question.
But if you really want to see speed language learning, examine David Lumanauw's written English over the last two years.
You both know Indonesia pretty well, you would know that only in the last generation has English been taught, and that the 1st language is emphatically Bahasa, with a polyglot of other widely used languages, particularly Javanese and Sundanese.
David, take a bow!!
Cheers,
Huigh
Let's face it guys, at the speed David is aquiring his audio knowledge he'll be our guru god in 10 years.
(some of us, the ones still in control of all the marbles that is.)
I wish i had his talent to grasp the picture the way he does so often.
Lumanauw, itu toean pintar. Satu hari, saya Janneman di Indonesia, David.
(some of us, the ones still in control of all the marbles that is.)
I wish i had his talent to grasp the picture the way he does so often.
Lumanauw, itu toean pintar. Satu hari, saya Janneman di Indonesia, David.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.