Super Op-Amps and why they mostly suck for audio
Jun 10, 2006 18:49:25 GMT
Post by rickcr42 on Jun 10, 2006 18:49:25 GMT
All about perspective and why the part was designed from the beginning.
Older Op Amps look like shit on paper in direct comparison to modern devices until you actually pay attention to those specs that impress so many,especially the specs whores who see the big picture but seem to always miss out on the details of end use.
At first glance modern DSL drivers would seem to be the ideal audio chip having light speed,huge bandwidth and gobs of available current at the output for driving long lines or headphones (or with parallel devices even loudspeakers) yet across the board they sound like crap !
Why ?
The market.Modern Parts that sound good are mostly entirely by accident because the intended use is in the mhz range with particularly bad performance characteristics at audio frequencies and especially so the lower registers where not only does distortion rise but don't expect those "gobs" of power to be any where near musical in nature !
Cell phones,network drivers,GPS,automotive "on board" controls,portable media players with crap headphones.....THIS is the target market with the last segment having zero to do with quality audio and everything to do with battery life for the ipod generation who want their music,movies,pictures NOW DAMMIT with no care about quality over and above TV resolution and sound.
so how to know when a chip is good for audio and when not without purchasing than evaluating ? How to know which parts will go fkn nut boy when used outside the bandwidth area they are designed for without headaches ?
Sometimes the data sheet tells the tale,sometimes not,sometimes you just have to fill in the blanks mentally with the information at hand then hope you are right-a crapshoot in short
Now look at an old standby known to work well for audio (if you truly can handle deadly accurate instead of just thinking you can because the term sounds good )
www.analog.com/UploadedFiles/Data_Sheets/336938187AD797_e.pdf
first of all the package is user freindly,the 8-Pin DIP poackage fully meant for humans to use instead of for machine insertion as with ALL modern devices.These parts maybe can be used but by design are fully meant to be placed then soldered by a machine which is strike 1 for use
Next look at the noise and distortion specs.First at the top of the page then at figures 2 and 3.
See anything familiar ? How about easy for a blind man to see just how the part will react to MUSIC and not RF signals ! In fact we can see a rise in THD over 30khz which right there goes 100% against the data of modern parts because of intended end usage.
we can also see a rise in noise in the below 500hz area but so ? Even this rise is miniscule for one and for two self noise is not as important below 3khz anyway if listening is the test
Dynamic performance :
The gain bandwidth product is enough for any conceivable audio product.No limitaons here folks.
Slew rate-20V/uS.This is almost laughable when compared to modern chips but the truth is you do not need more and in fact if you buy into the arguments on S.I.D. (slewing induced distortion,references see W.Jung and T.Holman) then you are in deep doo doo if you follow an ultrafast audio stage with one having less speed unless you use a low pass filter to slow down the preceeding stage
Settling time-finally a spec that has meaning for audio chips and why you never see this mentioned in most modern op-amp device data sheets.
Slew rate and bandwidth are a measure of how fast the part can respond to changes at the input without overshoot of lopping off part of the signal.too slow means it can not keep up,can not respond in time to what it is asked to do so in place of an accurate representation of the input but at a higher level what you get in return is a distorted "representation" of the input at a louder volume.
Settling time is the ability to STOP once there is no longer a signal at the input rather than continue to produce artifacts of the original long after the signal itself has stopped.These "sounds" have nothing to do with the music but everything to do with the actual chip.whatever harmonics the part tends to will be the last sound you hear for every single signal that is at the input so by nature is what determines the "voicing" of the part.Long settling times means more chip sound,faster setlling times means more accurate sound which truth be told most really can not handle.Reality can suck in audio and especially so if one section is accurate to the point of highlighting limitations elsewhere.
The way human nature works is-when theyt insert something that highlights system limitations they blame the "new" for the bad sonics rather than try and take the rest of the system up a notch to meet the challenge of TRUE accuracy.the same clowns who despise tone controls or filters because of some mythical accuracy usually can not handle "here it is jerky,straight in yo face the way it is without the sugar !"
So to put this in perspective,consider a car that can do 0-60MPH in 5 seconds,this is the "speed" componant of the specs.So you are at a red light,you decide to find out just how fast you can go and when the light turns green you NAIL IT !
YEEEEEHAAAAAAAWWWWW !
All of a sudden a car pulls out in front of you,your brain says "stop" and sends a signal to your foot to nail the brakes.If you have good reflexes,a good response time you avoid wrecking.If not tragedy..........
this is settling time
www.analog.com/UploadedFiles/Application_Notes/466359863287538299597392756AN359.pdf
fig.7-
0.1-10hz noise.You will NEVER see this spec in an ultra-high speed opamp data sheet.Since byt design they are never meant to see or respond to this area why bother woryying about it ?
PSRR (power supply rejection range) and CMRR (common mode rejection range)-meaningless for opamp comparison mostly since all monolithic chips beat the crap out of discrete designs and all are way more than adequate at audio relevant frequencies.Skip right over this section of a data sheet,I do ;D
fig.11
Extremely meaningfull data because it tells you what you can expect at the midband into a real world tough load and make no mistake,600 ohms is a bitch to drive unless the part has enough current to drive the input of the next stage
fig 12-
already touched on
fig.13-
irrrelevant but if you must,it does have a nice curve in the musically significant frequency band-all good
fig.14-
all about audio folks ! Look at the voltage levels and the frequency of the measurements.compare this data to any chip introduced in the last three years and you will find a handfull at best that have these specs in evidence.there are a few but considering the sheer amount of new product announcement over that time scary
Obviously there is a lot more to it or everyone would just grab a chip from the specs alone and have perfect audio which we all know is not the case butr read the data sheet for the AD797 over and over and over again and you will see a mental picture form that will enable at least culling out the truly bad parts once the qualifications for audio frequency use is understood.It is not always about the specs alone but the "why and what' of the specs can at least give a person some clue as to what can be expected in an actual circuit assuming a good layout and proper passive parts.
the rest is a combination of intuaition and time spent actually listening
Class dismissed ;D
.
Older Op Amps look like shit on paper in direct comparison to modern devices until you actually pay attention to those specs that impress so many,especially the specs whores who see the big picture but seem to always miss out on the details of end use.
At first glance modern DSL drivers would seem to be the ideal audio chip having light speed,huge bandwidth and gobs of available current at the output for driving long lines or headphones (or with parallel devices even loudspeakers) yet across the board they sound like crap !
Why ?
The market.Modern Parts that sound good are mostly entirely by accident because the intended use is in the mhz range with particularly bad performance characteristics at audio frequencies and especially so the lower registers where not only does distortion rise but don't expect those "gobs" of power to be any where near musical in nature !
Cell phones,network drivers,GPS,automotive "on board" controls,portable media players with crap headphones.....THIS is the target market with the last segment having zero to do with quality audio and everything to do with battery life for the ipod generation who want their music,movies,pictures NOW DAMMIT with no care about quality over and above TV resolution and sound.
so how to know when a chip is good for audio and when not without purchasing than evaluating ? How to know which parts will go fkn nut boy when used outside the bandwidth area they are designed for without headaches ?
Sometimes the data sheet tells the tale,sometimes not,sometimes you just have to fill in the blanks mentally with the information at hand then hope you are right-a crapshoot in short
Now look at an old standby known to work well for audio (if you truly can handle deadly accurate instead of just thinking you can because the term sounds good )
www.analog.com/UploadedFiles/Data_Sheets/336938187AD797_e.pdf
first of all the package is user freindly,the 8-Pin DIP poackage fully meant for humans to use instead of for machine insertion as with ALL modern devices.These parts maybe can be used but by design are fully meant to be placed then soldered by a machine which is strike 1 for use
Next look at the noise and distortion specs.First at the top of the page then at figures 2 and 3.
See anything familiar ? How about easy for a blind man to see just how the part will react to MUSIC and not RF signals ! In fact we can see a rise in THD over 30khz which right there goes 100% against the data of modern parts because of intended end usage.
we can also see a rise in noise in the below 500hz area but so ? Even this rise is miniscule for one and for two self noise is not as important below 3khz anyway if listening is the test
Dynamic performance :
The gain bandwidth product is enough for any conceivable audio product.No limitaons here folks.
Slew rate-20V/uS.This is almost laughable when compared to modern chips but the truth is you do not need more and in fact if you buy into the arguments on S.I.D. (slewing induced distortion,references see W.Jung and T.Holman) then you are in deep doo doo if you follow an ultrafast audio stage with one having less speed unless you use a low pass filter to slow down the preceeding stage
Settling time-finally a spec that has meaning for audio chips and why you never see this mentioned in most modern op-amp device data sheets.
Slew rate and bandwidth are a measure of how fast the part can respond to changes at the input without overshoot of lopping off part of the signal.too slow means it can not keep up,can not respond in time to what it is asked to do so in place of an accurate representation of the input but at a higher level what you get in return is a distorted "representation" of the input at a louder volume.
Settling time is the ability to STOP once there is no longer a signal at the input rather than continue to produce artifacts of the original long after the signal itself has stopped.These "sounds" have nothing to do with the music but everything to do with the actual chip.whatever harmonics the part tends to will be the last sound you hear for every single signal that is at the input so by nature is what determines the "voicing" of the part.Long settling times means more chip sound,faster setlling times means more accurate sound which truth be told most really can not handle.Reality can suck in audio and especially so if one section is accurate to the point of highlighting limitations elsewhere.
The way human nature works is-when theyt insert something that highlights system limitations they blame the "new" for the bad sonics rather than try and take the rest of the system up a notch to meet the challenge of TRUE accuracy.the same clowns who despise tone controls or filters because of some mythical accuracy usually can not handle "here it is jerky,straight in yo face the way it is without the sugar !"
So to put this in perspective,consider a car that can do 0-60MPH in 5 seconds,this is the "speed" componant of the specs.So you are at a red light,you decide to find out just how fast you can go and when the light turns green you NAIL IT !
YEEEEEHAAAAAAAWWWWW !
All of a sudden a car pulls out in front of you,your brain says "stop" and sends a signal to your foot to nail the brakes.If you have good reflexes,a good response time you avoid wrecking.If not tragedy..........
this is settling time
www.analog.com/UploadedFiles/Application_Notes/466359863287538299597392756AN359.pdf
fig.7-
0.1-10hz noise.You will NEVER see this spec in an ultra-high speed opamp data sheet.Since byt design they are never meant to see or respond to this area why bother woryying about it ?
PSRR (power supply rejection range) and CMRR (common mode rejection range)-meaningless for opamp comparison mostly since all monolithic chips beat the crap out of discrete designs and all are way more than adequate at audio relevant frequencies.Skip right over this section of a data sheet,I do ;D
fig.11
Extremely meaningfull data because it tells you what you can expect at the midband into a real world tough load and make no mistake,600 ohms is a bitch to drive unless the part has enough current to drive the input of the next stage
fig 12-
already touched on
fig.13-
irrrelevant but if you must,it does have a nice curve in the musically significant frequency band-all good
fig.14-
all about audio folks ! Look at the voltage levels and the frequency of the measurements.compare this data to any chip introduced in the last three years and you will find a handfull at best that have these specs in evidence.there are a few but considering the sheer amount of new product announcement over that time scary
Obviously there is a lot more to it or everyone would just grab a chip from the specs alone and have perfect audio which we all know is not the case butr read the data sheet for the AD797 over and over and over again and you will see a mental picture form that will enable at least culling out the truly bad parts once the qualifications for audio frequency use is understood.It is not always about the specs alone but the "why and what' of the specs can at least give a person some clue as to what can be expected in an actual circuit assuming a good layout and proper passive parts.
the rest is a combination of intuaition and time spent actually listening
Class dismissed ;D
.