In most common designs you se 2 Op-Amps in series for line-level output.
The first does some amplification and the second is a buffer which does not apply any amplification.
What is the reason for doing it this way and not use a single OpAmp for amplification and buffer ... the output should be able to handle the load and feedback.
Oystein
The first does some amplification and the second is a buffer which does not apply any amplification.
What is the reason for doing it this way and not use a single OpAmp for amplification and buffer ... the output should be able to handle the load and feedback.
Oystein
thermal feedback/cross coupling on a single chip and common power supply pin impedance coupling can limit the perfromance of dual op amps
if you want to optimize input and output characteristics for the application it is easier to choose different op amps having better specs for each function and combine them in a composite amplifier circuit with improved performance over either of the component amps
http://waltjung.org/PDFs/ADI_2002_Seminar_Ch6_Audio_Drivers_I.pdf
some of my composite amp sims:
http://www.diyaudio.com/forums/showthread.php?postid=512806#post512806
if you want to optimize input and output characteristics for the application it is easier to choose different op amps having better specs for each function and combine them in a composite amplifier circuit with improved performance over either of the component amps
http://waltjung.org/PDFs/ADI_2002_Seminar_Ch6_Audio_Drivers_I.pdf
some of my composite amp sims:
http://www.diyaudio.com/forums/showthread.php?postid=512806#post512806
The short answer is "it depends"... If the gain stage is only driving another stage that's on the same board (i.e. you know the input impedance, etc.) you might be better off without a buffer--there's something to be said for a cleaner signal path when it makes sense. But if the gain stage is driving say an external output, and you don't know what is going to be connected externally, then a buffer is often a good idea.
A different issue, highlighted by Jcx, is how to best use dual (or quad) op-amps. For the highest performance designs you should isolate the channels between different devices. A buffer may have to work harder driving low impedance or reactive loads (i.e. long cable runs) and you don't generally want that happening on the same piece of silicon as a high gain stage. So dual and quad op-amps are best used in one-channel-per-device applications where each op-amp is performing a similar function. A good example are high order filters with multiple stages. It's generally OK to combine filter stages for a given channel into a single device.
It's sort of like the difference between a receiver where all functions are crammed into a single box and having individual components--signal source, preamp, power amp. Generally the individual components will give better performance as each can be optimized for its function and well isolated from the others.
As Jcx mentioned, it's often ideal to optimize the op-amp by function. A high gain stage may be best implemented with a very low noise wide bandwidth opamp. A driver, OTOH, may require higher drive current output. But, personally, I think the latest high-end op-amps often do all of these things well. A good example is the National LM4562/LME49720. It has very low noise, very wide bandwidth and relatively high drive capability. So sometimes you *can* have your cake and eat it too
It should be said lots of very well respected commercial designs combine channels and different functions into dual or quad single devices with good results. Hence how much *audible* difference it really makes is open to debate. But if you want the best possible design, you should ideally isolate the channels and possibly conflicting functions among different physical devices.
One of the things us DIY guys enjoy is we can easily justify the cost of a few extra op-amps for a one-of personal project with the knowledge that it's more of a "no compromise" design than is generally considered acceptable in commercial products. With high-end audio op-amps (like the LM4562) at around $3 each most commercial designs tend to use as few as possible.
But, that said, don't add buffers where they're not needed. Having a bunch of extra transistors in the signal path won't help matters if the previous stage can easily drive the (known) load.
A different issue, highlighted by Jcx, is how to best use dual (or quad) op-amps. For the highest performance designs you should isolate the channels between different devices. A buffer may have to work harder driving low impedance or reactive loads (i.e. long cable runs) and you don't generally want that happening on the same piece of silicon as a high gain stage. So dual and quad op-amps are best used in one-channel-per-device applications where each op-amp is performing a similar function. A good example are high order filters with multiple stages. It's generally OK to combine filter stages for a given channel into a single device.
It's sort of like the difference between a receiver where all functions are crammed into a single box and having individual components--signal source, preamp, power amp. Generally the individual components will give better performance as each can be optimized for its function and well isolated from the others.
As Jcx mentioned, it's often ideal to optimize the op-amp by function. A high gain stage may be best implemented with a very low noise wide bandwidth opamp. A driver, OTOH, may require higher drive current output. But, personally, I think the latest high-end op-amps often do all of these things well. A good example is the National LM4562/LME49720. It has very low noise, very wide bandwidth and relatively high drive capability. So sometimes you *can* have your cake and eat it too
It should be said lots of very well respected commercial designs combine channels and different functions into dual or quad single devices with good results. Hence how much *audible* difference it really makes is open to debate. But if you want the best possible design, you should ideally isolate the channels and possibly conflicting functions among different physical devices.
One of the things us DIY guys enjoy is we can easily justify the cost of a few extra op-amps for a one-of personal project with the knowledge that it's more of a "no compromise" design than is generally considered acceptable in commercial products. With high-end audio op-amps (like the LM4562) at around $3 each most commercial designs tend to use as few as possible.
But, that said, don't add buffers where they're not needed. Having a bunch of extra transistors in the signal path won't help matters if the previous stage can easily drive the (known) load.
Thanks jcx and RocketScientist,
Your explanations helped me a lot.
I think it is a simple statement that a buffer is an amplifier with very low (zero) amplification (1:1)
But as always anyone who ask a question ( like me ) has a reason but in most of the cases the reason itself is unknown.
Reason:
I am planning to build myself a pre-amplifier based on a passive design with a resistor ladder with relays as is done by Jos van Eijndhoven: http://www.vaneijndhoven.net/jos/index.html
A complete passive design has its advantages but also some drwabacks - especially on the output side.
The Op-Amp design is meant to make it an active output with only a little amplification (or maybe none amplification) but mainly to lower the output impedance and to take care of any "load" induced by the cable and power amp.
I wil not use a regular Op-Amp but a discrete soultion as offered by Burson Audio. This takes away some design guessing from my side as the transistors are matched and idle current set properly.
I already used their Op-Amps in my cd-player with very good result which has a dual Op-Amps for amplification and buffer.
As the amplification in my pre-amp will be low and the cable to the power amp short I im confident that a single Burson unit per channel can do the job very well.
In an earlier Op-Amp discussion here in diyAudio it was already stated that the Burson units are very well suited as buffers and to trow in a little amplification should not hurt the performance.
Any other comments or design advise are very welcome
Your explanations helped me a lot.
I think it is a simple statement that a buffer is an amplifier with very low (zero) amplification (1:1)
But as always anyone who ask a question ( like me ) has a reason but in most of the cases the reason itself is unknown.
Reason:
I am planning to build myself a pre-amplifier based on a passive design with a resistor ladder with relays as is done by Jos van Eijndhoven: http://www.vaneijndhoven.net/jos/index.html
A complete passive design has its advantages but also some drwabacks - especially on the output side.
The Op-Amp design is meant to make it an active output with only a little amplification (or maybe none amplification) but mainly to lower the output impedance and to take care of any "load" induced by the cable and power amp.
I wil not use a regular Op-Amp but a discrete soultion as offered by Burson Audio. This takes away some design guessing from my side as the transistors are matched and idle current set properly.
I already used their Op-Amps in my cd-player with very good result which has a dual Op-Amps for amplification and buffer.
As the amplification in my pre-amp will be low and the cable to the power amp short I im confident that a single Burson unit per channel can do the job very well.
In an earlier Op-Amp discussion here in diyAudio it was already stated that the Burson units are very well suited as buffers and to trow in a little amplification should not hurt the performance.
Any other comments or design advise are very welcome
If you decided to go with a discrete solution it should not be an opamp. Opamp itself is suboptimal for audio. It becomes optimal when and only when cost and size criteria are thrown into the optimization equation. But discrete opamamp does not have such qualities like low cost and small size, so it is suboptimal.
Dear Wavebourn,
Cost itself is not the main factor here.
The reason I will build one myself is quality and flexibility for a limited investment.
The Burson modules takes away some uncertain factors, easy setting of the amplification and this peace of mind may cost a few Euro's extra ..
I am sure that if you go all the way anyone can built an even better pre-amp for the money but I am happy with the balance between result / cost / time spent.
Cost itself is not the main factor here.
The reason I will build one myself is quality and flexibility for a limited investment.
The Burson modules takes away some uncertain factors, easy setting of the amplification and this peace of mind may cost a few Euro's extra ..
I am sure that if you go all the way anyone can built an even better pre-amp for the money but I am happy with the balance between result / cost / time spent.
Just to offer another perspective here, and not wanting to start any heated debates 
. . . but I'll put the LM4562/LME49720 properly implemented up against *any* discrete or hybrid design in an appropriate audio application. National designed the parts using a very high-end sound room and system including Wilson Watt speakers. They are also used in many commercial products that have been rave reviewed by the most Golden of the Golden Ears.
Certainly some discrete circuits have "euphonic distortion" that some might prefer. But I'd argue it's hard to build or buy a more accurate audio op-amp than the LM4562 and its siblings. They are relatively new on the scene, and the only op-amps I'm aware of that were specified, designed and tested by a team of genuine audiophiles (some of whom are here on diyAudio).
So you can spend more, but I'm not sure you'll *get* more. In fact, you might get less. But I also respect part of DIY audio, for some, is about being esoteric above all else. And I can understand a $3 op-amp not having sufficient esoteric appeal. If it helps any, they're closer to $20 in the metal cans
. . . but I'll put the LM4562/LME49720 properly implemented up against *any* discrete or hybrid design in an appropriate audio application. National designed the parts using a very high-end sound room and system including Wilson Watt speakers. They are also used in many commercial products that have been rave reviewed by the most Golden of the Golden Ears. Certainly some discrete circuits have "euphonic distortion" that some might prefer. But I'd argue it's hard to build or buy a more accurate audio op-amp than the LM4562 and its siblings. They are relatively new on the scene, and the only op-amps I'm aware of that were specified, designed and tested by a team of genuine audiophiles (some of whom are here on diyAudio).
So you can spend more, but I'm not sure you'll *get* more. In fact, you might get less. But I also respect part of DIY audio, for some, is about being esoteric above all else. And I can understand a $3 op-amp not having sufficient esoteric appeal. If it helps any, they're closer to $20 in the metal cans
And I thought that those national parts were the ones with euphonic distortion.
The way I see signal amplification is , what comes in, is what must come out at higher level and I find a couple of AD video chips doing much better than the national chips in this regard. This is pretty easy to test at these signal levels.
I wont mention how well one can do with discrete design
homemodder said:
And I thought that those national parts were the ones with euphonic distortion.
The way I see signal amplification is , what comes in, is what must come out at higher level and I find a couple of AD video chips doing much better than the national chips in this regard. This is pretty easy to test at these signal levels.
I wont mention how well one can do with discrete design
I'd be curious to know in what measurable way, at frequencies and speeds of concern in audio, the AD chips outperform the National chips? Sure there are parts with higher gain bandwidth, slew rates, etc, but the speed of the LM4562 already far exceeds what's required for just about any rational audio application (and certainly for a unity or low gain preamp buffer as being discussed in this thread). All the faster parts I'm aware of are usually significantly inferior in ways that arguably do matter like noise and/or THD.
The National parts excel not only on the test bench, but also in listening tests. If AD has parts with superior overall measurements for audio use, I'd like to know which parts and what comparison measurements you have done?
And with discrete designs you're limited by how closely you can match discrete devcies, how closely they will thermally track each other, etc. Those issues alone put the discrete designer at a significant disadvantage to a monolithic IC designer. And the IC designer gets to create his own transistors optimized for each aspect of the circuit rather than having to compromise with what's available off the shelf. And 99+% of the people on this forum have nothing like the instrumentation resources at their disposal the big semiconductor companies have. Few here, that I know of, even have an AP system.
So, personally, I'll put my money on the chip designer likely getting a better result--especially when the parts are being specified and tested by serious audiophiles as was the case at National for the LM4562 and the later LME devices.
RocketScientist said:
Hmmm.... My ancient TOA console uses this "relatively new on the scene" opamps. Also, it uses relatively new 4558 opamps that I turned into even relatively newer ones biasing their output in class A, the difference is night and day. Plain they are horrible, when biased they are nice. But anyway opamp topology is suboptimal for sound amplification. However, you may prefer technotronic distortions that are hard to measure, while others may still prefer euphonic distortions that are inaudible.
Actually, there is one and only one way to reproduce clean sound, while number of ways to distort it is countless. That's why more and more of measurement methods are being invented, but they still can't cover all errors that may be created. It reminds me a crowd of persons that instead of going to the goal always try to avoid obstacles, so always have to back up bumping into walls, trees, poles, each other, falling in holes, and so on, because concentration on problems (distortions), running from them backward, that crowd of smart persons don't see what is behind.
Gopher said:
I simply mean implemented for appropriate applications (like Oystein's preamp output buffer--the subject of this thread) with proper circuit layout, power supply, grounding, voltage and current levels, etc. Obviously you can mis-use any op-amp (and a lot of people do).
Put another way, the LM4562 would make a lousy 100 Mhz RF amplifier, it won't drive a 50 ohm load, and if you don't pay attention to proper PCB layout, you won't get 0.00003% THD+N out of it (or any other op-amp). But if you do properly implement it, I argue it will objectively outperform virtually anything else out there, at any price, for audio purposes.
Wavebourn said:
The LM4562 was released in 2006 and the LME versions in 2007. So if your TOA mixer is "ancient" it must use some other device. In terms of op-amps designed for audio use, all the others I'm aware of are older designs--often *way* older as is the case with the 4558.
As for measurements, I agree. That's one reason I like to play real music through my designs under actual operating conditions and use differencing to compare the input to the output. So far, nothing has outperformed the LM4562 in this regard. Playing real music, under actual operating conditions, it alters the signal less than anything else I have tried. I hope to someday present an AES paper on using differencing to quantify various audio distortions that are normally not captured with more conventional measurements. I want to extend the work by Bill Waslo, and others such as:
Detecting Changes in Audio Signals by Differencing
RocketScientist, I was thinking more about
1. power supplies - higher or lower voltage rails for best sound?
2. local decoupling - rail-to-rail and/or rail-to-gnd and what value elecs with or without bypasses?
3. class a biasing of o/p - what value standing current?
4. inverting or non-inverting configuration?
5. matching impedances at +ve and -ve inputs
That sort of thing. Presumably National's audiophiles looked at all that? Any insights?
Could make the basis for a good thread specific to these opamps. I'd agree they probably are the best available at present but they do need careful implementation to get of their best.
Maybe Mark Brasfield could contribute?
1. power supplies - higher or lower voltage rails for best sound?
2. local decoupling - rail-to-rail and/or rail-to-gnd and what value elecs with or without bypasses?
3. class a biasing of o/p - what value standing current?
4. inverting or non-inverting configuration?
5. matching impedances at +ve and -ve inputs
That sort of thing. Presumably National's audiophiles looked at all that? Any insights?
Could make the basis for a good thread specific to these opamps. I'd agree they probably are the best available at present but they do need careful implementation to get of their best.
Maybe Mark Brasfield could contribute?
Gopher said:
I think Mark is better qualified to comment on the above than I am. He's already commented on some things. He says the metal can (TO-99) devices sound best and he prefers an inverting configuration. I think he's also commented on bypass caps but I don't remember exactly what his preference was.
My personal experience with the LM4562 is by simply following National's guidelines the performance exceeds what an Audio Precision system can measure. As I just wrote in my post above, I also use audio differencing to evaluate performance with real music signals, and here again, the LM4562 yields a better result than anything else I've performed the same tests on.
Wavebourn said:
Thanks. There's no date on that datasheet that I could find. This would be a good question for Mark or one of the other National guys on here. It's possible the JRC part is a second source license of National's design or it could be a completely different part.
Here's the National 2006 press release which clearly says it's "new":
LM4562 Press Release
I agree with RocketScientist. I use the LM4562 (LME49720) and have found it less colored than other op amps. Certainly no euphonic distortions which may not suit all. I have no desire to try a discrete op amp as I cannot tell any audible difference between in and out with this opamp.
(For preamp OP I use a LME49600 open loop, works well.)
(For preamp OP I use a LME49600 open loop, works well.)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.