Insider Information -the LM4562 family of devices
Feb 18, 2009 4:56:13 GMT
Post by Deleted on Feb 18, 2009 4:56:13 GMT
LM4562 family of I.Cs. Inside information from an ex NS Engineer.
Excerpts from the thread at DIYAudio : www.diyaudio.com/forums/showthrea....25&pagenumber=1
VFB=Voltage Feedback CFB=Current Feedback
The LME49713 CFB is a slightly better sounding part than the LME49710 VFB but they are both better than any other opamp on the market at this time.
The THD+N spec of 0.00005% or so is REAL! But the process the parts were built on and the engineer who designed them is what sets them apart.
(I specified the LME49713) These will be the reference opamps for audio for many years to come...
and this is coming from someone who is out of work because of National and wondering where I will end up?
In my D/A design I use the metal can LME49713's in most places but I do use the LME49710 metal can in one place where VFB opamps work better than CFB ones.
By the way the metal can parts sound better than the dip parts but measure the same on the AP2 Cascade (192).
and...
The LME49810 was designed as higher output current driver for live music amplifiers and has 50 ma of output current.
To get the extra output current (the LME49811 has 7ma) an A/B output stage was added to that part (the 811 has a class A output stage).
There were also some minor changes made to the input stage on the LME49811 (one was a suggestion by Bob Pease which is a great story
I might tell everyone sometime). I prefer the sound of the LME49811 to the 810 (I did listen to both parts on the same circuit board with minor changes)
but they are very close to each other. If you need more than 300w I would use the 810. For 300w or less the 811 is the way to go.
NS Sound Room : www.edn.com/blog/1700000170/post/790011279.html?q=LME49860
The LME49811 is a mono version of the LM4702 with a few minor layout changes that Bob Pease suggested.
The LME49710 is a whole other animal! Both parts sound great and all my listening comparisons were done
on the same ckt board inside of the same mono block amplifiers for both parts.
The pole being higher in freq for the 811 is because an extra A/B output stage was added to the 810 to get the 50 ma of output current, which of course changes the loop response. I believe that is the reason I slightly prefer the sound of the 811 to the 810 with identical bipolar output stages.
I don't like sound of FET outputs so let's not even talk about the 830.
Someday I could relate a long story of a test at Wilson audio with coupling cap removal in a 4702 amplifier but suffice to say
even the very best/most expensive caps in the world are very, very audible in the signal path!!!!!
Servos are the way to go where there is a DC problem but better to make a pure signal path with no offsets... which I know is hard to do.
The extra compensation circuit for the 4702 and 49811 is actually a noise gain compensation circuit not an extra compensation circuit.
I did not know that either and one of our Masters Stanford grads pointed out to several of us the difference.
Young college kids actually can be of use sometimes! It is mandatory for that 75pf and 3.3k resistor to be there!!!
You will have osc problems coming out of clipping if it is not there.
Also the 4702 and 49811 share the same soft clipping circuit that is virtually a Baker Clamp but the clamp in the 49811 IS a real Baker Clamp.
Sonically both parts are very easy on the ears when in clipping. So use the 810 if you need the clipping flag or need the 50ma to get more than 300 watts.
Use the 811 for the highest sound quality if the power needs are 300 watts or less.
oops! Post #22
One error I just saw in my reply needs to be corrected. The 810 has the real baker clamp not the 811.
The LME49811 and LM4702 have something very close to but not pure baker clamps. All the parts sound very good when driven into clipping compared to other solutions.
Of course it is ALWAYS better to have tons of headroom never go into clipping! But that does tend to increase costs just a bit!
Why do you want to use a single supply? It involves DC blocking caps, which will screw up the signal quality.
Usually single supplies are used on portable devices where signal quality is not a priority and single batteries are being used. (Also large consumer companies also do not want to spend more than a few cents on opamps in their hand held devices.)
The metal can devices (in split supply DC circuits) exhibit a more open and spacious sound quality with a little more detail (sorry for the non-technical audiophile terms).
In our single blind tests everyone choose the metal can devices over the identical internal die dip packages every time!
(Not a ton of tests though. Usually 5 to 10 subjects.) We used two identical D/A preamps I designed loaded with the two different packages (HA vs MA) and then measured them on the AP2/192 and saw no differences with the standard tests. More investigation is warranted and hopefully Bob Pease will attack this question at some point. I would really like to know for sure (measurements!) why the metal can devices sound a bit better…even in the power supplies!
I was told a little about the topology but yes my lips are sealed. Suffice to say there were some new wrinkles put into the 49710 VFB and 49713 CFB opamps that I believe contribute to their great sonic qualities. But then again the process these parts were built on, that makes the NPN and PNP parts identical in terms of carrier flow and dimension, may be the main reason for the great sound of this series of parts.
I like to think though that the process, design, layout and metal can packages all contribute to the great sound of these parts.
(And… let’s not forget the lowly apps guy that specified the 49713 CFB part for audio and pushed hard to get both parts made in metal cans!)
Since these LME parts have wider bandwidth they can have problems in some circuits and with bad power supplies.
Just putting bypass caps right on the power supply pins usually won't help with the osc problem.
Look at the output of your power supplies and see what is going on there. Sometimes a very low resistance series resistor before the bypass cap on the opamp power supply pins can help.
Layout can also be an issue, so cutting the power supply traces to the opamap and routing a pair of power supply wires directly to the opamp can also help.
What does the input to the LME49720/4562 look like? Sometimes a better opamp will pass through problem signals that the older,
lower bandwidth, device filtered out. Also don't put caps across the opamp, except maybe a couple of pf silver mica,
I never like slowing the opamp down to solve an osc problem, i.e. bandaid!
The LME49713 is a current feedback opamp that was designed at National per my specifications for audio.
CFB opamps are usually used in video circuits but they have some unique properties that make them a great choice for audio.
In my listening tests in the sound room at National the LME49713 was my favorite part but in my latest D/A design
(I have designed several D/A preamps at a bunch of audio companies over the last 20 years) I use both CFB and VFB opamps in the signal path, depending upon the circuit requirements. For example a VFB part works much better in a Diff to single ended circuit because the input structures of the plus and minus inputs are identical where as in a CFB opamp the two input structures are quite different.
Also CFB opamps are a bit harder to use as they require a fixed feedback resistance value for optimal performance.
In the case of the LME49713 that value is around 1.2k. The following is an excerpt from the 49713 data sheet that talks about the fixed value for Rf...
"The value of the Rf, is also a dominant factor in compensating
the LME49713. For general applications, the LME49713 will
maintain specified performance with an 1.2kÙ feedback resistor.
Although this value will provide good results for most
applications, it may be advantageous to adjust this value
slightly for best pulse response optimized for the desired
bandwidth. In addition to reducing bandwidth, increasing the
feedback resistor value also reduces overshoot in the time
domain response."
I should also mention that in my audio designs I prefer to use inverting circuit topologies wherever possible.
the LME49600 would be a great choice for your headphone buffer as that is what the 49600 was designed for.
It is built on the same process as the other LME parts and has the same sonic qualities.
I might also suggest the LME49713 as a second possibility as it has 100ma of output current which means you don't need to use a buffer at all to drive your headphones.
The metal can 49713 with a clip on heatsink would give you even a little more!
With regards to the opamps stay with what you have unless you need devices tested for +/-22VDC operation.
Do get the metal can LME49720's though if you want the best sound quality. (the 49720 and 4562 are identical parts).
If you can use singles the LME49713 CFB opamp makes a better I/V part.
Yes the 4562 and LME49720 are identical. National changed the numbering system to go from 4 to 5 digits, and since they would not give us a new name for the high performance series of parts we just changed the LM to LME (E for excellence).
Also the High voltage parts are identical but tested and guaranteed to higher standards. Testing always costs more...can anyone say MIL Spec!
2. Process variations are a fact of life and even parts on the same wafer can have slightly different characteristics.
Leakage is something that can vary on parts on the same wafer in the fab and they will pass all the tests AND leakage can affect sound quality!
The LM4702 and LME49811's are tested for leakage. I still think they are the best power amplifier front end parts in the world.
I will put them up against almost any discrete front end designs. I know of one ultra high end company that has done extensive testing on the 4702 parts, and feel it is almost identical to the sound of their 25K$+ mono block amplifiers with discrete front ends.
I heard an amplifier in their sound room that used a single stereo LM4702 in a mono balanced design at their factory over a year ago and it blew my mind.
We did that demo with the demo D/A preamp I designed for National. (They were very impressed with the D/A preamp with all LME parts).
The new ARCAM AVR-600 surround sound processor / Rcvr (HDMI-1.3!) also uses the LM4702's.
3. Bob Pease and I both feel the same way about Spice models for analog circuits... Almost worthless!
Engineers who never actually build anything use Spice models. The link for the spice model for the LM4565 series of parts is:
www.national.com/analog/audio
Still the best thing to do is just Build your circuit and test it out. The models can give you a general idea if your circuit will be functional, but in the end building it up is the only way to know what your circuit will really do what you want it to do!
I was unable to get the 49713 made in the dip package but the metal can can easily have its leads bent to fit into a DIP pattern without using a conversion board.
You can also use a SOIC to dip conversion board but the metal can will sound a bit better. Much more expensive for the metal can though...but for sound quality improvement worth it!
Take a good pair of needle nose and push on each pin make sure they are all in.
Mark-audioman54
Excerpts from the thread at DIYAudio : www.diyaudio.com/forums/showthread.php?s=&threadid=132471&perpage=25&pagenumber=1
Excerpts from the thread at DIYAudio : www.diyaudio.com/forums/showthrea....25&pagenumber=1
VFB=Voltage Feedback CFB=Current Feedback
The LME49713 CFB is a slightly better sounding part than the LME49710 VFB but they are both better than any other opamp on the market at this time.
The THD+N spec of 0.00005% or so is REAL! But the process the parts were built on and the engineer who designed them is what sets them apart.
(I specified the LME49713) These will be the reference opamps for audio for many years to come...
and this is coming from someone who is out of work because of National and wondering where I will end up?
In my D/A design I use the metal can LME49713's in most places but I do use the LME49710 metal can in one place where VFB opamps work better than CFB ones.
By the way the metal can parts sound better than the dip parts but measure the same on the AP2 Cascade (192).
and...
The LME49810 was designed as higher output current driver for live music amplifiers and has 50 ma of output current.
To get the extra output current (the LME49811 has 7ma) an A/B output stage was added to that part (the 811 has a class A output stage).
There were also some minor changes made to the input stage on the LME49811 (one was a suggestion by Bob Pease which is a great story
I might tell everyone sometime). I prefer the sound of the LME49811 to the 810 (I did listen to both parts on the same circuit board with minor changes)
but they are very close to each other. If you need more than 300w I would use the 810. For 300w or less the 811 is the way to go.
NS Sound Room : www.edn.com/blog/1700000170/post/790011279.html?q=LME49860
The LME49811 is a mono version of the LM4702 with a few minor layout changes that Bob Pease suggested.
The LME49710 is a whole other animal! Both parts sound great and all my listening comparisons were done
on the same ckt board inside of the same mono block amplifiers for both parts.
The pole being higher in freq for the 811 is because an extra A/B output stage was added to the 810 to get the 50 ma of output current, which of course changes the loop response. I believe that is the reason I slightly prefer the sound of the 811 to the 810 with identical bipolar output stages.
I don't like sound of FET outputs so let's not even talk about the 830.
Someday I could relate a long story of a test at Wilson audio with coupling cap removal in a 4702 amplifier but suffice to say
even the very best/most expensive caps in the world are very, very audible in the signal path!!!!!
Servos are the way to go where there is a DC problem but better to make a pure signal path with no offsets... which I know is hard to do.
The extra compensation circuit for the 4702 and 49811 is actually a noise gain compensation circuit not an extra compensation circuit.
I did not know that either and one of our Masters Stanford grads pointed out to several of us the difference.
Young college kids actually can be of use sometimes! It is mandatory for that 75pf and 3.3k resistor to be there!!!
You will have osc problems coming out of clipping if it is not there.
Also the 4702 and 49811 share the same soft clipping circuit that is virtually a Baker Clamp but the clamp in the 49811 IS a real Baker Clamp.
Sonically both parts are very easy on the ears when in clipping. So use the 810 if you need the clipping flag or need the 50ma to get more than 300 watts.
Use the 811 for the highest sound quality if the power needs are 300 watts or less.
oops! Post #22
One error I just saw in my reply needs to be corrected. The 810 has the real baker clamp not the 811.
The LME49811 and LM4702 have something very close to but not pure baker clamps. All the parts sound very good when driven into clipping compared to other solutions.
Of course it is ALWAYS better to have tons of headroom never go into clipping! But that does tend to increase costs just a bit!
Why do you want to use a single supply? It involves DC blocking caps, which will screw up the signal quality.
Usually single supplies are used on portable devices where signal quality is not a priority and single batteries are being used. (Also large consumer companies also do not want to spend more than a few cents on opamps in their hand held devices.)
The metal can devices (in split supply DC circuits) exhibit a more open and spacious sound quality with a little more detail (sorry for the non-technical audiophile terms).
In our single blind tests everyone choose the metal can devices over the identical internal die dip packages every time!
(Not a ton of tests though. Usually 5 to 10 subjects.) We used two identical D/A preamps I designed loaded with the two different packages (HA vs MA) and then measured them on the AP2/192 and saw no differences with the standard tests. More investigation is warranted and hopefully Bob Pease will attack this question at some point. I would really like to know for sure (measurements!) why the metal can devices sound a bit better…even in the power supplies!
I was told a little about the topology but yes my lips are sealed. Suffice to say there were some new wrinkles put into the 49710 VFB and 49713 CFB opamps that I believe contribute to their great sonic qualities. But then again the process these parts were built on, that makes the NPN and PNP parts identical in terms of carrier flow and dimension, may be the main reason for the great sound of this series of parts.
I like to think though that the process, design, layout and metal can packages all contribute to the great sound of these parts.
(And… let’s not forget the lowly apps guy that specified the 49713 CFB part for audio and pushed hard to get both parts made in metal cans!)
Since these LME parts have wider bandwidth they can have problems in some circuits and with bad power supplies.
Just putting bypass caps right on the power supply pins usually won't help with the osc problem.
Look at the output of your power supplies and see what is going on there. Sometimes a very low resistance series resistor before the bypass cap on the opamp power supply pins can help.
Layout can also be an issue, so cutting the power supply traces to the opamap and routing a pair of power supply wires directly to the opamp can also help.
What does the input to the LME49720/4562 look like? Sometimes a better opamp will pass through problem signals that the older,
lower bandwidth, device filtered out. Also don't put caps across the opamp, except maybe a couple of pf silver mica,
I never like slowing the opamp down to solve an osc problem, i.e. bandaid!
The LME49713 is a current feedback opamp that was designed at National per my specifications for audio.
CFB opamps are usually used in video circuits but they have some unique properties that make them a great choice for audio.
In my listening tests in the sound room at National the LME49713 was my favorite part but in my latest D/A design
(I have designed several D/A preamps at a bunch of audio companies over the last 20 years) I use both CFB and VFB opamps in the signal path, depending upon the circuit requirements. For example a VFB part works much better in a Diff to single ended circuit because the input structures of the plus and minus inputs are identical where as in a CFB opamp the two input structures are quite different.
Also CFB opamps are a bit harder to use as they require a fixed feedback resistance value for optimal performance.
In the case of the LME49713 that value is around 1.2k. The following is an excerpt from the 49713 data sheet that talks about the fixed value for Rf...
"The value of the Rf, is also a dominant factor in compensating
the LME49713. For general applications, the LME49713 will
maintain specified performance with an 1.2kÙ feedback resistor.
Although this value will provide good results for most
applications, it may be advantageous to adjust this value
slightly for best pulse response optimized for the desired
bandwidth. In addition to reducing bandwidth, increasing the
feedback resistor value also reduces overshoot in the time
domain response."
I should also mention that in my audio designs I prefer to use inverting circuit topologies wherever possible.
the LME49600 would be a great choice for your headphone buffer as that is what the 49600 was designed for.
It is built on the same process as the other LME parts and has the same sonic qualities.
I might also suggest the LME49713 as a second possibility as it has 100ma of output current which means you don't need to use a buffer at all to drive your headphones.
The metal can 49713 with a clip on heatsink would give you even a little more!
With regards to the opamps stay with what you have unless you need devices tested for +/-22VDC operation.
Do get the metal can LME49720's though if you want the best sound quality. (the 49720 and 4562 are identical parts).
If you can use singles the LME49713 CFB opamp makes a better I/V part.
Yes the 4562 and LME49720 are identical. National changed the numbering system to go from 4 to 5 digits, and since they would not give us a new name for the high performance series of parts we just changed the LM to LME (E for excellence).
Also the High voltage parts are identical but tested and guaranteed to higher standards. Testing always costs more...can anyone say MIL Spec!
2. Process variations are a fact of life and even parts on the same wafer can have slightly different characteristics.
Leakage is something that can vary on parts on the same wafer in the fab and they will pass all the tests AND leakage can affect sound quality!
The LM4702 and LME49811's are tested for leakage. I still think they are the best power amplifier front end parts in the world.
I will put them up against almost any discrete front end designs. I know of one ultra high end company that has done extensive testing on the 4702 parts, and feel it is almost identical to the sound of their 25K$+ mono block amplifiers with discrete front ends.
I heard an amplifier in their sound room that used a single stereo LM4702 in a mono balanced design at their factory over a year ago and it blew my mind.
We did that demo with the demo D/A preamp I designed for National. (They were very impressed with the D/A preamp with all LME parts).
The new ARCAM AVR-600 surround sound processor / Rcvr (HDMI-1.3!) also uses the LM4702's.
3. Bob Pease and I both feel the same way about Spice models for analog circuits... Almost worthless!
Engineers who never actually build anything use Spice models. The link for the spice model for the LM4565 series of parts is:
www.national.com/analog/audio
Still the best thing to do is just Build your circuit and test it out. The models can give you a general idea if your circuit will be functional, but in the end building it up is the only way to know what your circuit will really do what you want it to do!
I was unable to get the 49713 made in the dip package but the metal can can easily have its leads bent to fit into a DIP pattern without using a conversion board.
You can also use a SOIC to dip conversion board but the metal can will sound a bit better. Much more expensive for the metal can though...but for sound quality improvement worth it!
Take a good pair of needle nose and push on each pin make sure they are all in.
Mark-audioman54
Excerpts from the thread at DIYAudio : www.diyaudio.com/forums/showthread.php?s=&threadid=132471&perpage=25&pagenumber=1