Simplest discrete op-amps

[JohnRoberts]

[quote author="Svart"]Yes it is totally objective, I don't mean to imply that I have tested this professionally. They just sound strange in the same circuits.[/quote]
Sorry if I'm not being clear. By objective I mean measured as in frequency response or distortion. I would refer to an uncontrolled listening test as subjective. Even controlled listening tests such as double blind ABX can only statistically confirm a difference, they give little insight into what that difference is.

JR

 

[Guest]

John;

if I measure how perfect horizontally are aligned piano strings, and how well it's surface reflects ultra-violet rays, what will be your conclusion?

Anatoliy

 

[JohnRoberts]

[quote author="Wavebourn"]John;

if I measure how perfect horizontally are aligned piano strings, and how well it's surface reflects ultra-violet rays, what will be your conclusion?

Anatoliy[/quote]
You have too much time on your hands? :wink:

JR

PS: I draw a wide distinction between musical instruments and even in-studio creation of original sounds which are exercises in doing "whatever it takes" to get a sound, and playback/reproduction paths which IMO should be arbitrarily accurate and let the chips fall where they may.

Imagine if every circuit block was designed by a different designer loaded with his subjectively determined enhancements. The cumulative effect of all these deviations from linear combined in a serial audio path could be interesting to behold.

If you choose to modify original source material within sundry audio blocks fine, if it sounds good to you do whatever floats your boat. Just don't confuse introducing euphonious enhancements with delivering sonic accuracy. My preference is to make individual blocks linear, and put any signal enhancers in a properly identified black box. It's easier to keep track of and identify what you're hearing.

 

[Guest]

John, "Panta Rei" was said a long-long time ago by an encient Greek philosopher. Nothing to invent!
However, it is much simplier to measure some static parameters, but they matter much less than dynamics, such as dependence of them on time, volume, etc... Speaking of audio we have to measure them against subjective perceived levels, otherwise measures are irrelevant like horizontallness of piano strings and coefficient of reflection.
It is my theory that works well...

 

[JohnRoberts]

[quote author="Wavebourn"]John, "Panta Rei" was said a long-long time ago by an encient Greek philosopher. Nothing to invent!
However, it is much simplier to measure some static parameters, but they matter much less than dynamics, such as dependence of them on time, volume, etc... Speaking of audio we have to measure them against subjective perceived levels, otherwise measures are irrelevant like horizontallness of piano strings and coefficient of reflection.
It is my theory that works well... [/quote]
from wiki

"Everything flows and nothing is left (unchanged)". or "Everything flows and nothing stands still".or "All things are in motion and nothing remains still."

======

Subjective listening is always the final test because it can reveal things we didn't measure, but I find it unreliable by itself as a design tool unless you're designing things like studio effects, which I did back in the '70s. But even that requires some discipline.

One story comes to mind that may help explain my position. I was in a studio doing listening tests on some minor circuitry variants in a delay line/flanger. After a couple of hours, the group needed a smoke break. Since I don't smoke I got a cup of coffee. Some 15 or 20 minutes later when we picked up where we left off, the results had flipped and the variant I favored now swapped with the former. I spent about an hour trying to figure out what changed before giving up. It was my ears that changed. Perhaps the caffeine, perhaps blood sugar, whatever.

My personal theory is that human hearing is like a 10 1/2 digit digital voltmeter that is only accurate to 6 digits. We have excess resolution which can be perceived in side by side listening but over time (even as short as a few minutes) isn't reliable.

Another anecdote relates to how our personal perception adapts to it's recent experience. My home brew DIY speakers were deficient in the top octave but I had adapted and music sounded OK to me. The next time I was in a studio with proper monitors they took my head off... Lesson learned, I tweaked (tweetered) my speakers.

There has been a great deal of study and knowledge documented about human perception. I took advantage of some of the time dependent perception characteristics while designing tape noise reductions and dynamics processors ('70s-'80s) to help conceal the gain manipulations.

JR

PS: In a perhaps not surprising data point, customers preferred an older generation noise reduction wrapped around the analog delay chips that exhibited significant amounts of low order harmonic distortion. Since the customer is always right and it was after all an effect intended to alter the sound, adding distortion in this case was perfectly consistent with the products design goal (to make stuff sound good). IMO The design goal of a reference playback system is to reveal what's there, not change it. YMMV

 

[Kit]

Speaking of audio we have to measure them against subjective perceived levels

You said it yourself......"subjective" is the keyword here.

My ears might favour a different opamp than yours.

Data and specs is the only common language we got on this matter.

 

[Guest]

[quote author="JohnRoberts"]

There has been a great deal of study and knowledge documented about human perception. I took advantage of some of the time dependent perception characteristics while designing tape noise reductions and dynamics processors ('70s-'80s) to help conceal the gain manipulations.

[/quote]

Is it possible to find some of them?

PS: In a perhaps not surprising data point, customers preferred an older generation noise reduction wrapped around the analog delay chips that exhibited significant amounts of low order harmonic distortion. Since the customer is always right and it was after all an effect intended to alter the sound, adding distortion in this case was perfectly consistent with the products design goal (to make stuff sound good). IMO The design goal of a reference playback system is to reveal what's there, not change it. YMMV

I agree with you about "reveal ... not change". The question is, what "not change". It is impossible "not change" at all, something always will be changed. The question is, what. People tend to pay attentions to some "side effects" of "less changed" sound such as high level of lower order additional harmonics, while main reasons why it seems to be "not changed" slips through fingers...
Like all strings are aligned perfectly horisontally, but the piano still sounds out of tune... :roll:

[quote author="Kit"]

Speaking of audio we have to measure them against subjective perceived levels

You said it yourself......"subjective" is the keyword here.

My ears might favour a different opamp than yours.

Data and specs is the only common language we got on this matter.[/quote]

Sure. My point was, distortions' change means more than some absolute level of them. Everything flows, everything changes. Look at a snake in a cage if it does not move. You don't see it, I guarrantee, if it is still. But if it moves far from you, far from your direct sight, you will immediately turn your head, blood pressure will rise up a bit, adrenaline level will go up immediately. Similatry, switch on a light bulb in your bedroom in the middle of the night, to discover how bright the light is, despite it seems to be very dim during the day when the sun shines. Dynamics, this is what is more significant. Everything change, and we are trained to spot changes.

However, it is possible to measure sensitivity of human eyes to different light frequencies, also it is possible to measure gradations of light that seem to be different, depending on the background brightness. The same with sound... Human perceptions were trained a long time, generation to generation, no matter if to believe Darvin or Church clerics, anyway all our perceptions are best suited to live in the real world, where brightness, loudness, temperature, change in certain levels, and some levels are better recognizible, especially when they change.

Back in 30'th Fletcher and Munson experimented with loudness' perceptions and found that different objective measured levels of sound of different frequencies seems to be equal in loudness, and such loudness - frequency dependency curves depend on sound pressure, and such curves correllate well from listener to listener. I.e. subjective perception curves may be objectively measured!

You may experiment for yourself, and probably your perceptions are similar to perceptions of other people, no matter what kind of operational amplifiers do you prefer.

http://www.phys.unsw.edu.au/jw/hearing.html

 

[JohnRoberts]

[quote author="Wavebourn"]

There has been a great deal of study and knowledge documented about human perception. I took advantage of some of the time dependent perception characteristics while designing tape noise reductions and dynamics processors ('70s-'80s) to help conceal the gain manipulations.



Is it possible to find some of them?

[/quote]
For relationships that impact design of dynamics processors it varies from Haas's work in localization and fusion of sounds (also Madsen IIRC). There was work done at CBS in early days of their dynamics work for broadcasting and pressing records (no names come to mind). I don't have a specific list of references but would just suggest searching old AES papers, perhaps some IEEE consumer transactions with a specific area of interest. AES is probably more targeted to psycho acoustic phenomenon. One useful tidbit was time constants of our ears to adapt to large volume changes, that would provide some temporal masking after a loud sound goes away, etc.


JR

 

[Guest]

Hmmm... I googled Haas Madsen IIRC, no results...

AES requires $20 for each article that may have no value on the toipic...

Thanks anyway! :grin:

 

[bcarso]

for example:

http://en.wikipedia.org/wiki/Haas_effect

Not to be confused with the de Haas-van Alphen effect:

(http://www.lanl.gov/mst/nhmfl/users/pages/deHaas.htm)

or a variety of avocado (yum).

 

[recnsci]

Few (uncorellated) thoughts on some issues arrised here and elswhere:

Subjective impressions:
Many people dismiss subjective impressions ("ears are lying, brain is
lying even more, so personal opinions are irrelevant"). And people
that prefer subjective impressions over measurments go to great
leanghts in fighting each other over which device sound good and
bad.
And here we have a problem, IMveryHO.

I will give you a story as example. Long ago I was very into analog
synths and their "guts" and saturated my inbox with mailing list very
similar to this forum. And everybody had opionion on various topologies
and incarnations of filters, oscilators and rest.
What I noticed is that my attention was focused on how voltage controlled
filters react to transients on voltage controll input, that is to fast envelope.
I am not refering to controll law ( linear or expo or something else). I am
reffering to the fact that different circuits react very diferently to sudden
changes of some state variables (like resistance of optoresistor).
So imagine little mind experiment. Take two filters that measure
same in THD, gain vs freq and other departments, but differ in a way
they respond to fast envelopes. And me and some guy that doesnt care
about reaction to envelope that much are listening tho those two filters.
He would say "They are same" and I would say "They are way off".
And, equaly imporant, I noticed one interesting pattern. As time went by
my capability to diferentiate various sounds got better. That is, at first
every filter and every EQ sounded same to me. But with more and more listenting I developed capabilitie to "focus" on some aspects of a sound
(like, fine details of what is 5KHz boost doing to a cymbal). This
focusing was actually concious brain activity. IMHO, ears dont get tired
at normal listening levels, but our brains have limited attention span.
I have a theory that what differs big money mixers (like Lord-Alge bros)
to us mortals is that their focusing capability remain intact during 12 hours
mixing session (and they know what to pay attention to).

Conclusion? Well, I think ears are as valid testing machines as any bench
procedure. And similar to bench testing, one pair of ears is focusing
on this aspect (because brain pays attention to that aspect) and other
pair focus on some other. One bench test will reveal CMRR other will
tell THD.

I'm not saying that everything people think they hear is real. There is
lot if imagination (especially in audiophile world) involved. People
decide do discern some info and make up nonexisting sounds.
I am however saying that ear itself is perfectly respectble tool. Its
up to us to train our brains to focus and to force our self into "unbiased"
mode.

And one final remark on this topic ( i promisse). I grow more and more
concerned on how unbiased ABX tests are (ok now I'm pulling flame
resistant suite). Especilay in psychology tests that revealed final limitations
of our hearing systems. I went numerous times to hearing tests. My
brain got lazy after 30 seconds of listening to static sinewaves. After
10 minutes I was barely able to hear. I asume that after 6 hours
of listening to that crap I would become psychotic. I guess you got my
point (and probably dont agree with it).

Now other stuff.

Testing
I think we should start new topic and resolve this issue once and
for all.
Test like THD number with barely loaded output (without notion of
exact topology) at 0 dBu sine on input are hardly important. No one
will dissagree that two circuits with same THD number could sound
raddicaly different. Also coments like "this IMD test will reveal a lot".
What da *** is that test revealing ? differences? apples? oranges?
holly grail?
We need:
1) comprahensive set of testing procedures
2) comprahensive set of theory papers that will discuss what are
we measuring and how
3) rigid set of rules for measurment like amp loading, topology etc
so that we are sure we are comparing apples to apples
4) disscusion of how to interpret results
5) (very important IMHO) growing database of conducted tests
so that we have benchmark to look at

So pretty please, could we open discussion on this matter ?

OpAmps and discrete vs IC:

Well, back to actuall toppic. When discussing merrits of those 8 legged
monsters, we should always keep in mind context, and that is using
them in actuall audio circuits. There are few important audio tasks
where ic opamps still fail short, stuff like driving dificult loads and
providing lot of gain. And if they do that stuff gracefully they tend to be
expensive. And, opamp=clean discrete=dirty(euphonic) is simply not
true as general statement. TL0xx-driving-1K = dirty , gordon audio
preamp = clean beyond capabilities of any current ic topology.
But in a next few
years we will probably have ic opamps that will do most of tasks in
true wire with a gain aproach at reasonable prices (maybe we have them
already).

And finaly:
DO we need transparent ?

Transparency (for whatever that means) seems to be goal of audio for
quite some decades (whole ****** century). Well, I have very personal
opinion that differs. It's not that I like "euphonic" here and there, its
that I need it most of the time. Two very important elements in audio
chain are badly lacking in performance after all these years, mic at front
end and speaker at back end. In every aspect, freq response, thd, transient response, you name it. Even more, whole concept of recording
is doomed from begining. Realism would mean that soundfield in recording
environment is reproduced in listening environment. Harsh physical
reality of transducers, rooms and formats tells us "Nope honey, no can do"
We are always creating arteficiall image and even if by direct divine
intervention "perfect" reproduction chain apeared tommorow, 90% of
music genres today rely on "arteficiall" interventions in recording.

While I agree that "colour" should be added when its needed, and that
we should have transparent block in chain on other occasions, problem
as I see it is not in making that transparent segment. ****, I think
5532 is transparent for me (and most of others) in most ocassions.
Problem is creating boxes that give more on output than you provide
on input, cus I need those. And if box has 0.1% thd it doesnt mean its
good sounding. It will probably sound like ****.
So, issue is box that is making "good" 0.1% thd on some sources
(and if some box is doing good on every source in every context
I'm willing to donate kidney for schematic, layout and BOM).

If you are still reading this that means you have too much time on your
hands. Go to studio and record, or go solder something.

cheerz
urosh

 

[JohnRoberts]

[quote author="recnsci"]Few (uncorellated) thoughts on some issues arrised here and elswhere:

Subjective impressions:
Many people dismiss subjective impressions ("ears are lying, brain is
lying even more, so personal opinions are irrelevant"). And people
that prefer subjective impressions over measurments go to great
leanghts in fighting each other over which device sound good and
bad.
And here we have a problem, IMveryHO.

I will give you a story as example. Long ago I was very into analog
synths and their "guts" and saturated my inbox with mailing list very
similar to this forum. And everybody had opionion on various topologies
and incarnations of filters, oscilators and rest.
What I noticed is that my attention was focused on how voltage controlled
filters react to transients on voltage controll input, that is to fast envelope.
I am not refering to controll law ( linear or expo or something else). I am
reffering to the fact that different circuits react very diferently to sudden
changes of some state variables (like resistance of optoresistor).
So imagine little mind experiment. Take two filters that measure
same in THD, gain vs freq and other departments, but differ in a way
they respond to fast envelopes. And me and some guy that doesnt care
about reaction to envelope that much are listening tho those two filters.
He would say "They are same" and I would say "They are way off".
And, equaly imporant, I noticed one interesting pattern. As time went by
my capability to diferentiate various sounds got better. That is, at first
every filter and every EQ sounded same to me. But with more and more listenting I developed capabilitie to "focus" on some aspects of a sound
(like, fine details of what is 5KHz boost doing to a cymbal). This
focusing was actually concious brain activity. IMHO, ears dont get tired
at normal listening levels, but our brains have limited attention span.
I have a theory that what differs big money mixers (like Lord-Alge bros)
to us mortals is that their focusing capability remain intact during 12 hours
mixing session (and they know what to pay attention to).

Conclusion? Well, I think ears are as valid testing machines as any bench
procedure. And similar to bench testing, one pair of ears is focusing
on this aspect (because brain pays attention to that aspect) and other
pair focus on some other. One bench test will reveal CMRR other will
tell THD.

I'm not saying that everything people think they hear is real. There is
lot if imagination (especially in audiophile world) involved. People
decide do discern some info and make up nonexisting sounds.
I am however saying that ear itself is perfectly respectble tool. Its
up to us to train our brains to focus and to force our self into "unbiased"
mode.

And one final remark on this topic ( i promisse). I grow more and more
concerned on how unbiased ABX tests are (ok now I'm pulling flame
resistant suite). Especilay in psychology tests that revealed final limitations
of our hearing systems. I went numerous times to hearing tests. My
brain got lazy after 30 seconds of listening to static sinewaves. After
10 minutes I was barely able to hear. I asume that after 6 hours
of listening to that crap I would become psychotic. I guess you got my
point (and probably dont agree with it).

Now other stuff.

Testing
I think we should start new topic and resolve this issue once and
for all.
Test like THD number with barely loaded output (without notion of
exact topology) at 0 dBu sine on input are hardly important. No one
will dissagree that two circuits with same THD number could sound
raddicaly different. Also coments like "this IMD test will reveal a lot".
What da *** is that test revealing ? differences? apples? oranges?
holly grail?
We need:
1) comprahensive set of testing procedures
2) comprahensive set of theory papers that will discuss what are
we measuring and how
3) rigid set of rules for measurment like amp loading, topology etc
so that we are sure we are comparing apples to apples
4) disscusion of how to interpret results
5) (very important IMHO) growing database of conducted tests
so that we have benchmark to look at

So pretty please, could we open discussion on this matter ?

OpAmps and discrete vs IC:

Well, back to actuall toppic. When discussing merrits of those 8 legged
monsters, we should always keep in mind context, and that is using
them in actuall audio circuits. There are few important audio tasks
where ic opamps still fail short, stuff like driving dificult loads and
providing lot of gain. And if they do that stuff gracefully they tend to be
expensive. And, opamp=clean discrete=dirty(euphonic) is simply not
true as general statement. TL0xx-driving-1K = dirty , gordon audio
preamp = clean beyond capabilities of any current ic topology.
But in a next few
years we will probably have ic opamps that will do most of tasks in
true wire with a gain aproach at reasonable prices (maybe we have them
already).

And finaly:
DO we need transparent ?

Transparency (for whatever that means) seems to be goal of audio for
quite some decades (whole ****** century). Well, I have very personal
opinion that differs. It's not that I like "euphonic" here and there, its
that I need it most of the time. Two very important elements in audio
chain are badly lacking in performance after all these years, mic at front
end and speaker at back end. In every aspect, freq response, thd, transient response, you name it. Even more, whole concept of recording
is doomed from begining. Realism would mean that soundfield in recording
environment is reproduced in listening environment. Harsh physical
reality of transducers, rooms and formats tells us "Nope honey, no can do"
We are always creating arteficiall image and even if by direct divine
intervention "perfect" reproduction chain apeared tommorow, 90% of
music genres today rely on "arteficiall" interventions in recording.

While I agree that "colour" should be added when its needed, and that
we should have transparent block in chain on other occasions, problem
as I see it is not in making that transparent segment. ****, I think
5532 is transparent for me (and most of others) in most ocassions.
Problem is creating boxes that give more on output than you provide
on input, cus I need those. And if box has 0.1% thd it doesnt mean its
good sounding. It will probably sound like ****.
So, issue is box that is making "good" 0.1% thd on some sources
(and if some box is doing good on every source in every context
I'm willing to donate kidney for schematic, layout and BOM).

If you are still reading this that means you have too much time on your
hands. Go to studio and record, or go solder something.

cheerz
urosh[/quote]

The golden ear vs. meter reader debate is probably older than some readers here. The argument is constantly corrupted by straw men (misleading characterizations) about each other's position. The GE will argue all MRs are over reliant on tests that don't reveal all flaws, while MRs argue GEs are constantly reinventing the wheel and too invested in things only they can hear that seem to disappear under scrutiny. There are surely examples of both with more than a little truth, but IMO that is not the norm for both groups.

I have a slightly different take on the GE/MR issue. If you can reliably hear some audible phenomenon that is real and not a perceptual distortion, that phenomenon will have a physical basis that can be measured. Once that is measured, it can be managed in designs. Further as perceptual distortions are understood, they too can be integrated into designs and/or practices (like loudness eq decades ago) .

I won't waste bandwidth with a full list of straw men, but the "same THD as" is a classic. Low order harmonic distortion will have a dramatically different perceived sound quality than high order crossover distortion. Trying to equate the two is disingenuous.

I believe we all have the same goal (good sound), and can probably learn from each other. If someone identifies a truly new phenomenon they'll get their name in the technical journals, but please don't ignore the several decades (or more) of work that has gone before us.

JR

PS: I too find $20 an AES paper expensive, but as a long time member I read many papers when they were fresh. I concede these days I don't follow it as closely.

 

[Svart]

It's a debate that will never end. Audiophools tend to forget that the audio they listen to through their 20K$ stereo was likely run through tons and tons of opamp ICs.

IC's like every other parts chosen, have their place in the design world. I've been doing listening studies with IC's in my console and find that for high audio frequencies I like the 5532 and bipolar caps coupling and for low frequencies I like the 2134 and no cap coupling. Each has their sound and their place in the world. As for outboard units, for insert gear a 5532 would be fine, but for main output gear before final mixdown, I find that the discrete stuff wins out.

Legos man legos.

 

[Martin B. Kantola]

Know things have progressed since, but I'm still hung up on PRR's excellent post, sorry... It triggered a lot of thoughts for me.

[quote author="PRR"]In Audio, we need a reasonably stable gain, usually low.[/quote]

This is what I was thinking. Lower gain, fewer stages needed. Fewer stages, fewer parts? I don't need 120 dB of gain in my Type II amp.

[quote author="PRR"]The current source in the long-tail is needed for low-gain followers, not for high-gain or inverters. And arguably the diff-input is a gilded-lily for practical audio use. f you can drive a load, you can NFB to the input emitter, lose a noise source, and cap-couple your DC troubles away. [/quote]

Here I'd like to belive I find some support for my initial suggestion. A buffer is such a different task than a high-gain stage, and as I wrote we could drop the inverted input. In fact, the buffer (Type I amp) could use a differential output stage instead. Then it can also double as an inverter should we need one.

[quote author="PRR"]If you must design ONE building-block to cover MANY jobs (chip-making demands such silly notions), then yes the op-amp is the block to build. That argument holds no water when I solder-up one or two boosters.[/quote]

What about designing THREE simple building blocks then? Anybody interested?

[quote author="PRR"]But too many people (even some very wise people) can't hear and won't listen.(...)Learning how to hear, what we can hear, what pleases the ear, rather than focusing on mere parts.[/quote]

Amen. Measuring is not wrong, but can be very difficult when listening is an easy, fast and accurate way of working. Not scientific, but I don't care if it helps me acheive the goal.

[quote author="PRR"]If you must DIY non-heroic Audio opamps: Doug Self covers the waterfront.[/quote]

Yes, I feel I must, but I'll need some help. The Doug Self teachings are great reading, thanks for the suggestion!

Hope it's OK to quote a little from
http://www.dself.dsl.pipex.com/ampins/discrete/discrop.htm

"Opamps are mostly restricted to supply voltages of +/-18 or +/-20 Volts. Hybrid-construction amplifiers, typically packaged in TO3 cans, will operate from rails as high as +/-100 V, but they are very expensive, and not optimised for audio use in parameters like crossover distortion. Discrete opamps provide a viable alternative. // A load requires more drive current, because of its low impedance, than an opamp can provide without overheating or current-limiting; eg any audio power amplifier // The best possible noise performance is required. Discrete bipolar transistors can outperform opamps, particularly with low source resistances, say 500 Ohms or less. // The best possible distortion performance is demanded. Most opamps have Class-B or AB output stages, and many of them (though certainly not all) show clear crossover artefacts on the distortion residual. A discrete opamp can dissipate more power than an IC, amd so can have a Class-A output stage, sidestepping the crossover problem completely."

As earlier said, really wish to do this SE/Class-A all the way if possible.

Martin

 

[burdij]

There are three principles that lead to good amplifiers. One is a controlled amounts of negative feedback. The principle effects of this is to give accurate gain control and to allow the servo loop established around the amplifier to reduce distortion components. This was discovered by Black. The second is a low amount of gain per stage. This sets the stage for the reduction in the amount of TIM distortion. Humans are, for whatever reason, extremely sensitive to this type of distortion. This was expounded upon by Marshall Leach and others. In fact, there have been some interesting animal studies concerning this type of distortion including:

http://www.ncbi.nlm.nih.gov/entrez/...ve&db=PubMed&list_uids=10511624&dopt=Abstract

The third principle is elimination of crossover distortion, particularly for low level signals. This effects the "recovery" of low level detail in the audio waveform.

What criteria does this suggest, then for the simple discrete amp design?

1. Unfortunately, to be usefull, it appears to need both non-inverting and inverting inputs unless one wants to always design with an even number of gain stages so then, only an inverting input would be necessary.

2. The gain block must have a fairly high slew rate at a moderately high gain so that it can achieve a good GBW for a significant number of harmonics of the fundamental of each intermingled wave in the wave "packet".

3. Class A, all the way.

 

[JohnRoberts]

[quote author="burdij"]There are three principles that lead to good amplifiers. One is a controlled amounts of negative feedback. The principle effects of this is to give accurate gain control and to allow the servo loop established around the amplifier to reduce distortion components. This was discovered by Black. The second is a low amount of gain per stage. This sets the stage for the reduction in the amount of TIM distortion. Humans are, for whatever reason, extremely sensitive to this type of distortion. This was expounded upon by Marshall Leach and others. In fact, there have been some interesting animal studies concerning this type of distortion including:

http://www.ncbi.nlm.nih.gov/entrez/...ve&db=PubMed&list_uids=10511624&dopt=Abstract

The third principle is elimination of crossover distortion, particularly for low level signals. This effects the "recovery" of low level detail in the audio waveform.

What criteria does this suggest, then for the simple discrete amp design?

1. Unfortunately, to be usefull, it appears to need both non-inverting and inverting inputs unless one wants to always design with an even number of gain stages so then, only an inverting input would be necessary.

2. The gain block must have a fairly high slew rate at a moderately high gain so that it can achieve a good GBW for a significant number of harmonics of the fundamental of each intermingled wave in the wave "packet".

3. Class A, all the way.[/quote]

I don't share your exact design philosophy or observations as stated.

"Transient-evoked otoacoustic emissions" in guinea pigs if anything argues for less slew rate and bandwidth not more. Less click, less IMD in their/our ears.

Rather than low gain in every stage managing slew rate induced effects, this battle is won or lost in the very first stage. Marshal Leach's paper on this is quite clear http://www.aes.org/e-lib/browse.cfm?elib=3265. ( at only 3 pages long this is even less of a value at $20 for non members.) Of course all stages need to be linear and capable of delivering adequate power bandwidth, but that must be combined with proper treatment in the front end. When out of band signals are properly LPF by the front end they will not generate IMD and there will not be any signal edge rates left capable of causing slew limiting.

We are in agreement about many of the pieces. Of course crossover distortion is bad and Class A while brute force, can be very linear but even that has it's tradeoffs. Negative feedback when properly used will deliver very linear signal paths. I'm not aware of any advantage to low amounts of gain per stage, especially if it's a choice between two low gain or one higher gain stage. When these gain stages are inside an overall negative feedback loop the extra delay from the additional stage will make stabilization more difficult and could compromise GBW available.

Good Luck

JR

 

[clintrubber]

[quote author="Martin B. Kantola"]... when listening is an easy, fast and accurate way of working. Not scientific, but I don't care if it helps me acheive the goal.[/quote]
It'll have the 'problem' of selecting the source material. What sounds good on a soft triangle may obviously not sound good on a full mix and vice versa.

But problems are to be solved; your remark reminds me I should compile some samples of different material. Separate drums, complete loops, elec & ac. bass notes, some mixes etc for evaluation. Some DI'd clean gtr & bss as well for re-*mping tests as well etc

And we shouldn't share these on purpose, since it'll be better that everyone has a highly personalized collection of snippets.
Nobody does the same kind of music and has the same goal in mind, so lets incorporate that. I'm not interested in what the best opamp is on the voice of a underage girl hopping around in videoclips. :wink:

Regards,

Peter

 

[recnsci]

[quote author="burdij"]There are three principles that lead to good amplifiers. One is a controlled amounts of negative feedback. The principle effects of this is to give accurate gain control and to allow the servo loop established around the amplifier to reduce distortion components. This was discovered by Black. [/quote]
Accurate gain is not mandatory for every audio chain ( and is even
less of a problem for DIY people). How will NFB affect distortion
is case dependant.


[quote author="burdij"]The second is a low amount of gain per stage. This sets the stage for the reduction in the amount of TIM distortion. [/quote]
What do you mean by low amount of gain per stage? Stages inside global
NFB? And this is somehow related to TIM in non-case-dependant fashion?
Hard slew in LTP based amp, for example, is strictly dependat on
capacitances and bias currents. Which in lot of cases is gain independent.

[quote author="burdij"]The third principle is elimination of crossover distortion, particularly for low level signals. This effects the "recovery" of low level detail in the audio waveform.[/quote]

Well this is an issue that deserv lenghty discussion. That Hawksford guy
details low level distrotion mechanism based on quantised nature of
current (finite number of carriers), which would broaden issue
to just about any topology imaginable. Anyway, I would bet my life that
nicely designed AB out will outperform badly designed SE out in
"detail" department.

[quote author="burdij"]
1. Unfortunately, to be usefull, it appears to need both non-inverting and inverting inputs unless one wants to always design with an even number of gain stages so then, only an inverting input would be necessary.
[/quote]

If you are deadbent on having just one ampblock for every
function in circuit. You can make very nice SE blocks that would be
inverting or noninverting. Just look how many different ampcards
early Neve SE designs had. And, more or less, these early designs
were often feedback-to-emiter topology. That is simply SE version
of todays CFB opamps. So you have two inputs. Its just that one is
low impedance.


[quote author="burdij"]
2. The gain block must have a fairly high slew rate at a moderately high gain so that it can achieve a good GBW for a significant number of harmonics of the fundamental of each intermingled wave in the wave "packet".[/quote]
I didnt understand this one. Could you explain ?


[quote author="burdij"]
3. Class A, all the way.[/quote]
Probably, class AB would be better only in PSU-and-heatsinks ecconomy
department. But again Class A is not equal Good Sound. Fuzz Face
is SE class A.

cheerz
urosh

 

[burdij]

I believe the article cited talks about non-harmonic noise interaction within the cochlea of the ear caused by an, as of yet, unknown effect.

Leach describes an amplifier with low TIM as one with less than about 20db of gain per stage, set by local feedback loops. There are several articles authored by him and others on TIM effects. One problem with TIM is measuring it. It is a case of the measuring instrument's TIM performance. Kind of like measuring .001% harmonic distortion with a harmonic distortion analyzer (like mine) with .01% minimum resolution. And as far as I know IMD and TIM are two different things, that in fact, IMD is an in-band phenomenon of the interaction or modulation of high frequency in-band components by lower frequency in-band components.

Here is a short excerpt from Maxim App. Note 2119 regarding the two common methods for measuring IMD:

.

The SMPTE (Society of Motion Picture and Television Engineers) standard specifies a two-sinewave test signal consisting of a low-frequency, high-amplitude tone linearly combined with a high-frequency sinewave at 1/4 the amplitude of the low-frequency tone. The SMPTE specification calls for 60Hz and 7kHz as the two sinewaves. When a non-linear device is subjected to a two-tone test signal, intermodulation products appear as sidebands around the high-frequency tone. The percentage intermodulation distortion is defined as the percentage of amplitude modulation, represented by the second and third order pair of sidebands, of the high-frequency signal. Second order sidebands around the high frequency tone are spaced at a frequency equal to the low-frequency tone (FH ± FL). Third order sidebands are spaced at twice the low-frequency tone or FH ± 2FL. (FH and FL correspond to the highfrequency and low-frequency tones, respectively.2)

CCIF intermodulation distortion testing consists of using two equal-amplitude, high-frequency signals spaced closely in frequency. Signals used for test data presented include 13kHz and 14kHz. The audio analyzer measures only the amplitude of the difference tone or low-frequency product resulting from the two high-frequency test signals. Here, the percentage intermodulation distortion is defined as the percentage of amplitude modulation which the low order or difference signal (FH - FL) represents on the high-frequency signals.

burdij wrote:

2. The gain block must have a fairly high slew rate at a moderately high gain so that it can achieve a good GBW for a significant number of harmonics of the fundamental of each intermingled wave in the wave "packet".

I didnt understand this one. Could you explain ?

As far as this comment goes, it all has to do with the Fourier Series. If you don't reproduce a goodly number of the harmonics of the fundamental, including their phase relationships (since there is a complex component involved with the AC waveform), you can't possibly result in an accurate "reproduction" of the input waveform including its transient phenomenon. An important concept in FM electronic music synthesis, I believe.

This stuff isn't my idea, by the way, this has been around since well, 1927, I suppose. I was just trying to get a list of some achievable goals for the design of a simple discrete gain stage, not get mired in the psychoacoustic mumbo jumbo. Let's face it, most things can look very good when run through test equipment, but, as we have all experienced, can vary widely when subjected to the harsh criticism of the listener.

 

[mikep]

[quote author="burdij"]One problem with TIM is measuring it. It is a case of the measuring instrument's TIM performance. [/quote]

it will show up in a null test. use transient-filled music as the source and the TIMD should sound like splats or little bursts of noise.