Sound Quality Vs. Measurements

[Joachim Gerhard]

In the conclusion they reason why it should sound good :http://www.ti.com/lit/an/snaa047a/snaa047a.pdf
By the way settling time is given at 1.4usec in the data sheet ( not shown here ).
That is nothing speacial.

 

[Joachim Gerhard]

In the conclusion they reason why it should sound good :http://www.ti.com/lit/an/snaa047a/snaa047a.pdf
By the way settling time is given at 1.4usec in the data sheet ( not shown here ).
That is nothing special.

 

[jcx]

if you read a psychoacoustics textbook you might find that temporal masking is strong - lots of differing experiments with noise, tone probes, but it looks like the envelopes are generally worse than a 1 ms time constant decay curve - thats milliseconds not micoseconds for post temporal masking

makes settling time of electronics seem like a distant worry - at least look at loudspeaker drivers, cone bending modes, crossovers...

 

[Kastor L]

Ok, I think this needs some clarification.

I can see quite a few here and many others elsewhere have a certain reflex answer about op-amp's, revolving around this sentence "implementation is everything".

While this is certainly true to an extent, it has a certain dismissive character to it at the same time.

While we can stack IC's, increase current output, bias to Class-A, use the correct decoupling capacitors, use them in the correct circuit and so forth, X component will always be X component and it has certain inherent limitations within itself.


If one were to construct a competition and say "who can make the best sounding kit" or "who can make the best measuring kit with x measuring equipment and x measurement parameters", using x, y, z, q, r parts, then the results may vary quite a lot.


If this forum was overclock.net or diylaptop perhaps there would be very similar discussions.

You use x, y, z, q, r parts to make a laptop, if the operating system software is not handling them correctly and the internal hardware architecture is not very well designed, the laptop may not perform very well.

However, if that competition was using x, y, z, q, r laptop parts only from 2001, even a wizard with the highest skill could not make it perform like a decent 2014 laptop.

This is due to the fact that all components have certain inherent limitations within themselves.

However, the catch here is that laptops have not achieved any kind of upper limit yet.

A lot of people believe that audio has achieved that upper limit.

Thus, the expression is actually "Implementation is everything, I think audio has surpassed all limits and achieved perfect transparency".

If you believe that, let's say, LME49990 sounds better than LME49720 in an ideal system for each of them, then the expression is actually.

"Components are everything, I don't think audio has surpassed the limits, it has not achieved any kind of perfect transparency".

My comments are not personal to anyone in this thread, I'm just writing 'to the internet' in general.

Just my views.


I think if you believe in the former or the latter depends to a certain extent if you are mathematically inclined or musically inclined.

Luckily, society has a fair evenness of these sharp types, hopefully it advances in all directions.

It may also depend on if you are a leveller or a sharpener, these are two new terms which I learnt today!

http://www.stereophile.com/asweseeit/are_you_a_sharpener_or_a_leveler/index.html

 

[Kastor L]

I'd say that settling time is, in a sense, the opposite side of the slew rate, but in my view, an extremely important factor.

Rise time usually tells how fast will n amplifier go from 10% to 90% of its output; settling time tells you the exact opposite, how long will it take to go down from 90% to 10% of its output.

Most older op amps will have a reasonable rise time, but their settling times will be rather long, roughly ranging from 2,000 to 1,500 nS. If we want true signal symmetry, we would want out settling time to no longer than our rise times; we don't want anything happening once the signal has passed, do we?

If memory serves, Analog Devices was the first on the market with op amps which had reduced settling times, with their OPA 275 op amp being ver popular in audio at one point (and used by Rotel and many others). Over time, its settling time of 400 nS was cut down to below 100 nS, and for settling not to the industry standard 0.1%, but 0,01%.

Most of today's crop of new generation op amps now have similar numbers, and from most vendours, I think.

So, in terms of capability, it is definitely there, but whether it will be made to work properly is still for the designer to make happen.

if you read a psychoacoustics textbook you might find that temporal masking is strong - lots of differing experiments with noise, tone probes, but it looks like the envelopes are generally worse than a 1 ms time constant decay curve - thats milliseconds not micoseconds for post temporal masking

makes settling time of electronics seem like a distant worry - at least look at loudspeaker drivers, cone bending modes, crossovers...

O.k., thank you for the info.

Can you two please help me check I am reading this correctly......

http://www.ti.com/lit/ds/symlink/opa2111.pdf
Settling time 0.01% = 10 uS...... = 10,000 ns

http://www.analog.com/static/imported-files/data_sheets/AD811.pdf
http://www.analog.com/static/imported-files/data_sheets/AD828.pdf
Settling time 0.01% = 80ns

10,000 / 80 = 125.

Thus, the AD828 is one hundred and twenty five times faster in ST than the other?

If yes, I think I'll order these and try to blind test them.

 

[jcx]

why wouldn't you start with a 200 kHz single pole filter - as is often seen in front of amps for RF filtering - I believe even John Curl and his buddies give a 200KHz pole a OK on audibility grounds

 

[fas42]

"Components are everything, I don't think audio has surpassed the limits, it has not achieved any kind of perfect transparency"

I would put it "Implementation plus components are everything; very rarely does audio even start to approach the limits; transparency has always been achievable but in the past implementation was extremely important because components were often so-so, the current crop of components make implementation less critical, but it still must be scrupulously attended to" ...

 

[Kastor L]

why wouldn't you start with a 200 kHz single pole filter - as is often seen in front of amps for RF filtering - I believe even John Curl and his buddies give a 200KHz pole a OK on audibility grounds

I have heard audio chips which pick up radio frequency information, I suspect it doesn't affect sound quality, it simply introduces some extra RF noise in the information.

LT1115 in Virtual Ground picks up my power supply noise, likewise it doesn't really seem to affect sound quality, it's just some annoying buzzing on top of everything else.

 

[1audio]

I think that settling time is usually defined as how long it takes for the output to remain within a certain error band after the transition.

You are right. DVV's remarks are a little unclear, likening it to rise time.
Two good references, first for simple clarity Settling time - Wikipedia, the free encyclopedia second because it goes deep into what settling time means in the context of amplifiers : http://www.analog.com/static/imported-files/application_notes/466359863287538299597392756AN359.pdf Note, its from 1970.

Settling time is a specific factor that cannot be determined from rise time, slew rate, power bandwidth or GBW product. All impact in various ways. Measuring accurately, especially for a non-inverting amp, is quite difficult. Its useful in knowing under dynamic conditions in what circumstances the output departs from the desired error band. Back to the -80 dB distortion number, that would be .01% accuracy. If nothing changes in the system in a time shorter than 50 uS (1/20 KHz) then a 50 uS settling time would be fine. I would suspect 5 uS to .01% may be adequate for a full or 1/2 scale transition but would want to look carefully to see if its enough. And the reciprocal may not adequately represent the required time for accuracy.

What you discover is that everything impacts that settling time, including thermals, stored charges, resistor voltage coef. and tempco's etc. Sometimes a slower opamp can have a faster settling time and other occasions the faster may never settle to usable accuracy.

 

[Kastor L]

I would put it "Implementation plus components are everything; very rarely does audio even start to approach the limits; transparency has always been achievable but in the past implementation was extremely important because components were often so-so, the current crop of components make implementation less critical, but it still must be scrupulously attended to" ...

I haven't seen any evidence that audio has achieved a perfect, divine, crystal-like transparency in any fashion.

I suspect that a very large extent of playback audio equipment is lightly shaded.

Keep in mind, everything which we actively see and hear, is around 200ms prior to reality. We are not passive receptors, we actively project reality unto ourselves.

Pre-attentive processing - Wikipedia, the free encyclopedia

Flash lag illusion - Wikipedia, the free encyclopedia

Flash-Lag Effect

http://www.youtube.com/watch?v=DUBM-GG0gAk

Therefore, with theoretical transparency factor 9000, I suppose it's theoretically viable, that if we trick our senses and delete the annoyance information, that we could theoretically trick ourselves into hearing a crystal-like illusory reality.

Note, this indicates that the defined limits of human hearing are actually the perceived limits, of course our pre-attentive processing is picking up a little more information in SNR, time, tonal colour et cetera than the information in which we are actively hearing.


The rest of that which you write is true, in 2014 it's much easier to theoretically achieve transparency factor 9000 than it was in 2001.

In any case, I favour the "transparency has no limits" school of thought in audio, that's all really!

The colour school seems never-ending, hype inclined, expensive, tiring and annoying.

Just my views. Have a nice day.

 

[dvv]

The 2N3055 is not a very good part. It is slow, limited in safe area, etc. It was the first NPN silicon power transistor that became available in the 1960's, so it was used in early equipment. Even Ampex moved to the 2N3773 by 1968 to get more safe area. I think that the 'textbook' example is accurate. The same thing happens here because of SPICE emulations, that use obsolete parts as well, and everybody learns while using them. There is better out there, but for a motor drive servo, it might be enough. I designed my first phase locked servo with a 2N3055 in 1968, and used 741's by the handful in 1969, at these slow speeds they were OK.

John, I'd like to amlify your statement that 2N3055 was a very good part and change that to an excellent part. However, with a note: before anyone knew it, everybody and their dog were making it, and as it usually happens, not all were the same.

I think that everybody who started in stereo in the mid 60ies had something to do with it at one time or another. My own was by owning a reVox A78 integrated amplifier, which used a single pair in a SEPP configuration per channel, and delivered a healthy 40/50 WRMS into 8/4 Ohms. Frankly, that was a good sound at the time, but I hasten to add mine were RCA devices.

The British audio industry stuck with them until the mid 80ies (e.g. Arcam).

Much the same can be said about 741 - everybody used it because at the time it was what could have been expected, they were easily available and they didn't cost an arma and a leg.

But that was then, and now, it's a whole new ball game. We have much better power deviecs, and op amps come in all colours, sizes and formats, from cheap and cheerful, to expensive but really good. I do believe that in this day and age, using 2N3055/2N2955 is a bit off, given the possible alternatives.

 

[dvv]

Kastor, two notes:

1. My explanation was a very general one only, so if you want the fine line, you will have to read up on it in more detail, and

2. Each and every component has its limits, the ones which are inherent in its internal topology. All these specs define only one aspect of it, but the end result will always be defined by its overall performance and by the way it has been implemented. This primarily applies to using it within its performance envelope, i.e. nor asking it to do something that it cannot do, or something that brings is very near to its absolute limits.

Sometimes, you might see comments like op amps are limited because their dynamic rangle of the output is limited by the usual +/- 15...18V PSU lines. At first glance, this may appear to be quite reasonable, but do give it a second glance and nore that many of them will swing +/- 24V peak to peak, or about 8.5 Vrms. Now ask yourself - assuming just one half of that coming out of them, what's my power amp doing?

I don't think there's a power amp made by anyone, at any power or price, which will not go well into clipping with such an input signal. Most power amps in the 100-200W/8 Ohms range need less than 2V of input to deliver full power. So what do you need more than 8.5V capability in the real world?

This is a one sided view, there are many more issues to deal with. I have taken just one aspect of op amps in the example above, and we both know how many there are by just looking at a reguklar data sheet.

Do take a look at Samuel Groner's site SG-Acoustics · Samuel Groner · Home , he did a comprehensive overview of many of the populatr op amps.

 

[dvv]

Demian, I likened the settling time to the rise time only by saying that this was the opposite effect, going the opposite direction, from 90% back down to 10%.

That simply means that such a possibility, or capability, exists, but by no means does it guarantee such behavior in any and all cases. Too many factors are involved which, as you say, may influence both the rise and settling time.

With op amps, a simplistic approach simply does not work every time. I have seen a lot of folks locally who read the data sheet of some op amp, saw that it had a slew rate of say 350 V/uS, and lately 2,000+ V/uS, and blindly took out their much slower op amps from whatever, threw the new ones in and then wondered what went wrong. They equated slew rates with sound quality - and paid the price of it.

Hence my addition to the comments "depending on the implementation". Sometimes a direct swop works, but other times, it doesn't.

 

[fas42]

Therefore, with theoretical transparency factor 9000, I suppose it's theoretically viable, that if we trick our senses and delete the annoyance information, that we could theoretically trick ourselves into hearing a crystal-like illusory reality.

Well, it's not "perfect, divine, crystal-like" - it just always sounds real, no matter how loud it is, to the ear - which can be adjusted by moving your head closer to the speakers. Of course it's a trick, the "annoyance information" is still being reproduced, but it's low enough in level that your consciousness is not aware of it, and in fact you can't make yourself aware of artifacts even if you deliberately concentrate on doing so, your mind has accepted the illusion as real.

 

[Kastor L]

Well, it's not "perfect, divine, crystal-like" - it just always sounds real, no matter how loud it is, to the ear - which can be adjusted by moving your head closer to the speakers.

What do you mean by it sounds real? You mean, close to real? I've never in my life heard a speaker reproduction in my life that sounds real.

What do you mean by the head closer to the speakers effect? LoL

_________

Just continuing my previous post above about the sensory information.

If someone hits a baseball towards you, you are actively imagining that baseball, you are seeing into the future, so you can catch it.

This is not theory, just look at the flash latency effect.

If we were passively seeing it, we wouldn't have the reaction time to catch it, by the time we reach out for it it would have just passed our hand already.

So now the confusing part......

1 - If we are imagining the baseball, then how did we catch it? LoL

2 - If the baseball flies past our ear and we hear the swish sound, then the swish sound is 200ms into the future as well, correct?

Hmm, anyway, here's an interesting paper "This demonstrates that compared to non-musicians, musicians are superior in pre-attentively extracting more information out of musically relevant stimuli."
http://www.ncbi.nlm.nih.gov/pubmed/10363945

The above is basically a vague equivalent of saying "this demonstrates that compared to non-audiophiles, audiophiles are superior in extracting more information out of audio equipment relevant stimuli"

Another "Pre-attentive auditory deficits, which represent a core feature of schizophrenia"
http://www.biologicalpsychiatryjourn...630-2/abstract

 

[jan.didden]

But settling time is not defined as the inverse of the rise time. The inverse of the rise time is the fall time.
Settling time starts with the rise of the signal - generally, at the top of the wave, there is a few cycles of resonance until the level 'settles' at the final value. Generally, settling time ends at the time that the resonances die down to a certain percentage of the final value.
For instance, 0.1% settling time is the time it takes for the level to 'settle' at 0.1% of the final value.

Jan

 

[Kastor L]

Kastor, two notes:
Do take a look at Samuel Groner's site SG-Acoustics · Samuel Groner · Home , he did a comprehensive overview of many of the populatr op amps.

Thanks DVV for your explanations.

Yes, I remember he measured the distortion of many popular op-amp's including discrete.

I remember, I think, the AD797 had the lowest distortion of all that he measured.

The discrete op-amps that he measured like SUN, EARTH, MOON, Burson and some other company had the highest distortion.

I don't know, I don't think that's an error of discrete, just the discrete designs he had in his tests?

I wonder, if someone made an exact AD797 circuit in discrete form, if it would perform higher or lower in distortion than the normal AD797.

 

[nigel pearson]

I wonder why would anyone, even Nige, use 2N3055/2N2955 in the day and age when Toshiba's 2SC5200/2SA1943 cost literally a few pennies more? Come on, Nige, this 2014, not 1970.

On Joachim's comment on more output current - I have yet to hear ANY op amp which did not sound better with a small discrete current booster stage. If you add to this a driver-output stage combo, you have a very viable pure class A headphone amp of possibly impressive qualities.

Use say AD 847 dual op amp, one half for the audio work and the other half as a servo, and the have say BC 639/640 or MPSA 056/06 drive say BD 139/140, or MJE 15030/15031, and I can't imagine a dynamic headphone this couldn't drive to total deafness if required - but into pure class A deafness.

That's not the point. I have often found a good engineer can make excellent out of mediocre ( chefs also ). We are judging the engineer not the components. Without any disrespect to anyone choosing a Miss World is one thing. The person is more complex and not just a pretty face. Equally a transistor is mostly a transistor these days , the personality or correct personality of the circuit is what I want. I suspect most 2N3055 are nothing of the sort. They are something better sold as 2N3055 ( TIP 3055 was 8 MHz in the A+R A60 in the time of Noah ) . That is the point of the competition. I suspect some would even take numbers off and rubber stamp fake ones. Cute ? If you read the JLH site the 3055 was hard to beat. If running the competition I would totally accept a JLH with excellent PSU and suspect to my ears a Krell killer. I am into mid range and layers with bass to please. Some people love bass mostly. I can be that person if the mid is correct. No, I am that person. Accepting the JLH would especially allow a novice with great ears to win. I would use the Krell as a reference amp. I would use Dynaco A25 and Quad ESL 63 as references. I would be happy to be told the latter not acceptable. I would also fully expect the Krell to have a very difficult fight. People can not ignore it's reputation and be sympathetic to the competitors. I would fully expect the winner to be better. I would not be on the judging panel.

 

[nigel pearson]

Here is the inside of NE5534. It is a very complex device. Motorola went out of their way to say how simple the MC33079 is. It is usefully better in every spec except output current. Look at the graphs and even the 5534 strong points are equaled. Simple with equal spec will sound better.

Just road map examples. You must go to the town and do your own research.
The BC337 is about 4 times better on noise. The input is via the compensation pins which MC33078/9 can not do that.

 

[Joachim Gerhard]

I ones made an inverting phono with only one stage.
More or less a Phonoclone.
4 x NE5534 in parallel and RIAA shunt.
It worked surprisigly well and i build it from the waiste bin.
The NE5534 has around 70dB gain at 20kHz so that is possible.
With 60dB gain at 1kHz 30dB are left for feedback at 20kHz.
Not much but it worked.