MBK said:
Intermittent bugs such as those you describe have a dispropotionate effect on the performance of your amp.....ergo......distortion residual must be time aligned with your output signal, and observed for some time with different resistive loads, including the reactive variety, or combinations thereof.
traderbam said:
How true. One of the dangers of relying purely on listening, without measurement or diagnostic aids, is that you never get down to why something sounded the way it did. Then you fall into the trap of attributing it to the first thing that comes to hand.
For example - selection of transistors in a differential input pair in a power amp. It's natural to try a few different types, find one that sounds good, and proclaim that type of transistor a Good Thing. However, it may be that that particular circuit relies on good matching between the input transistors (and isn't particularly sensitive to other parameters), and what you've actually done is stumbled across a well-matched pair by accident.
If so, taking the trouble to find this out will let you get good results from cheap transistors, and avoids you claiming things about transistor types which aren't reproducable elsewhere.
Cheers
IH
IanHarvey said:
I agree to some extent but there is more to it IMO. Even with good measurement equipment you may still not be able to pin down why it sounded that way. Too many parameters to measure - which one did matter in some particular case? This falls again in the conundrum of finding out what matters most and what doesn't. Plus, there's those "measures bad - sounds good" cases.
To find out the parameters that the circuit is most vulnerable to I guess you need a mixture of theoretical intuitions and reasoning as to why this or that should matter most. Measurement can then confirm it. And for that good intuition the only thing that will help is broad background and broad experience. This is what I feel I miss the most in my own DIY fiddling.
Measurement alone can keep you busy forever if you don't have a good clue beforehand. That's what I mean with "to know what to measure and how to do it".
About learning to design
Ian's example is an excellent one. It is really easy to draw the wrong conclusion from an observation. Considering many possibilities and testing out the simple ones first is a good methodology in my experience.
It is also critical to get the basics right before you attack the more complicated things. For example, you must get the grounding arrangements right at the start - otherwise you will spend your life messing with transistors and feedback systems and all sorts and get absolutely nowhere and become confused and frustrated because you can't find any consistency.
MBK is right that knowing what to measure and how to measure it requires a combination of experience and education and as such takes time to acquire. But you can work it out if you are diligent, systematic and learn and believe in basic engineering principles. Resist those who claim there is some etherial force at work that makes things sound good. No. Leave the art and esoterics to the musicians.
If you really want to achieve perfection I recommend starting with something extremely simple, something where the system is relatively simple and then play with it to gain intuition. The add to it in small steps and repeat playing with it. At each step try to understand what the engineering explanation is for the change in performance. Gradually build up the circuit and your knowledge.
If you start with someone else's complex circuit, like Slone's, you will find it very difficult to understand how to design things. The system is already very complex and changing a single component will not easily lead to a single cause of performance change. Rather, I would start with one transistor and a pair of headphones in class A and go from there. Use quality parts: polyester caps, tantalum electrolytics, metal film resistors - all easily available and quite adequate.
To start off with, you need to know Ohm's Law, you need to understand reactance and impedance, how to calculate power and how to calculate the static and dynamic currents in a circuit. If you use loop feedback you need to learn about stability. All of this can be gained from basic texts like Horowitz and Hill. Just one thing to note, an it's kind of important, very few electronics books deal with non-linearities and none deal with how non-linearities affect the sound of music. This is what you have to learn for yourself.
Beware of amateur publications: it is really hard to separate the wheat from the chaff, and beware of AES journal-like stuff because it is all about the academic pursuit of obscure concepts in excruciating detail. The very best source of info is the person who designed the amp you think sounds great - but guess what - they probably won't tell you 'cause they are making money in a competitive market and don't want to give away their hard-earned secrets.
Ian's example is an excellent one. It is really easy to draw the wrong conclusion from an observation. Considering many possibilities and testing out the simple ones first is a good methodology in my experience.
It is also critical to get the basics right before you attack the more complicated things. For example, you must get the grounding arrangements right at the start - otherwise you will spend your life messing with transistors and feedback systems and all sorts and get absolutely nowhere and become confused and frustrated because you can't find any consistency.
MBK is right that knowing what to measure and how to measure it requires a combination of experience and education and as such takes time to acquire. But you can work it out if you are diligent, systematic and learn and believe in basic engineering principles. Resist those who claim there is some etherial force at work that makes things sound good. No. Leave the art and esoterics to the musicians.
If you really want to achieve perfection I recommend starting with something extremely simple, something where the system is relatively simple and then play with it to gain intuition. The add to it in small steps and repeat playing with it. At each step try to understand what the engineering explanation is for the change in performance. Gradually build up the circuit and your knowledge.
If you start with someone else's complex circuit, like Slone's, you will find it very difficult to understand how to design things. The system is already very complex and changing a single component will not easily lead to a single cause of performance change. Rather, I would start with one transistor and a pair of headphones in class A and go from there. Use quality parts: polyester caps, tantalum electrolytics, metal film resistors - all easily available and quite adequate.
To start off with, you need to know Ohm's Law, you need to understand reactance and impedance, how to calculate power and how to calculate the static and dynamic currents in a circuit. If you use loop feedback you need to learn about stability. All of this can be gained from basic texts like Horowitz and Hill. Just one thing to note, an it's kind of important, very few electronics books deal with non-linearities and none deal with how non-linearities affect the sound of music. This is what you have to learn for yourself.
Beware of amateur publications: it is really hard to separate the wheat from the chaff, and beware of AES journal-like stuff because it is all about the academic pursuit of obscure concepts in excruciating detail. The very best source of info is the person who designed the amp you think sounds great - but guess what - they probably won't tell you 'cause they are making money in a competitive market and don't want to give away their hard-earned secrets.
Re: About learning to design
Agree. Then again. So I started using op amp based circuits including the power stage to keep with systems where the design basics had been worked out by the manufacturer, right? Add water and mix, right? ... ... until ... ... until I realized that say "Understanding the inverting op-amp input" is not a textbook paragraph
... but a book!! ... in 35 volumes!!!
At this point I'd be happy enough if I truly understood what maketh a decent power supply.
[note: It's a joke. A joke! Errr. Not.]
traderbam said:
Agree. Then again. So I started using op amp based circuits including the power stage to keep with systems where the design basics had been worked out by the manufacturer, right? Add water and mix, right? ... ... until ... ... until I realized that say "Understanding the inverting op-amp input" is not a textbook paragraph
... but a book!! ... in 35 volumes!!! At this point I'd be happy enough if I truly understood what maketh a decent power supply.
[note: It's a joke. A joke! Errr. Not.]
"Beware of amateur publications: it is really hard to separate the wheat from the chaff, and beware of AES journal-like stuff because it is all about the academic pursuit of obscure concepts in excruciating detail. The very best source of info is the person who designed the amp you think sounds great - but guess what - they probably won't tell you 'cause they are making money in a competitive market and don't want to give away their hard-earned secrets."
Other than the fact that some of the greatest articles on audio design were written for the Audio Amateur by Nelson Pass, Erno Borbely, Richard Marsh, Walt Jung, Ben Duncan, and our own DIY member Jan Didden. Many of these articles were written by people outside the audio industry with new approaches and test equipment outside the realm of the standard audio test equipment used by most designers. It's kind of hard to layout the cash for things like a $60 K network analyzer. I had access to an Agilent 4284A Precision LCR Meter at my last telecom job that was quite useful but was about $15 K. I was so sad to see my test equipment and I part company that I could have written a country western song about it that would make any engineer cry.
Knowledge of topologies, passive components, and listening to what other designers are doing is essential but the good fundamentals of engineering will take you to the point and beyond that the tweeker will reach and in less time.
Skills and experience in both camps are very useful. Perhaps a new term like EnginEARing is in order to describe the combination of the two approaches.......
Fred
P. S. Nelson and Erno are sharing secrets like mad. A removed top and 30 minutes with a good amplifier will also tell you a hell of a lot, but don't Steve McCormak or he will start potting things or removing part numbers.
Other than the fact that some of the greatest articles on audio design were written for the Audio Amateur by Nelson Pass, Erno Borbely, Richard Marsh, Walt Jung, Ben Duncan, and our own DIY member Jan Didden. Many of these articles were written by people outside the audio industry with new approaches and test equipment outside the realm of the standard audio test equipment used by most designers. It's kind of hard to layout the cash for things like a $60 K network analyzer. I had access to an Agilent 4284A Precision LCR Meter at my last telecom job that was quite useful but was about $15 K. I was so sad to see my test equipment and I part company that I could have written a country western song about it that would make any engineer cry.
Knowledge of topologies, passive components, and listening to what other designers are doing is essential but the good fundamentals of engineering will take you to the point and beyond that the tweeker will reach and in less time.
Skills and experience in both camps are very useful. Perhaps a new term like EnginEARing is in order to describe the combination of the two approaches.......
Fred
P. S. Nelson and Erno are sharing secrets like mad. A removed top and 30 minutes with a good amplifier will also tell you a hell of a lot, but don't Steve McCormak or he will start potting things or removing part numbers.
Yogi Beraisms anyone...........
"observed for some time with different resistive loads, including the reactive variety"
I would like to see one of those! My friend Jay O. Mega says it would be too complex for me and resisted telling me why for some reason. I will not be impeded and will get the angle on this no matter what the magnitude of the task. I thought you were serious until I read the post again but some of the other stuff gave it away as subtle example of techno-humor. I'm jealous even with the number of times that I posted humor that was taken for actual advice. I tip my hat to you and will endeavor to work harder in the face of such competition.
http://www.wordspring.com/yogiberaquotes.html variety"
Good one,
Fred
"observed for some time with different resistive loads, including the reactive variety"
I would like to see one of those! My friend Jay O. Mega says it would be too complex for me and resisted telling me why for some reason. I will not be impeded and will get the angle on this no matter what the magnitude of the task. I thought you were serious until I read the post again but some of the other stuff gave it away as subtle example of techno-humor. I'm jealous even with the number of times that I posted humor that was taken for actual advice. I tip my hat to you and will endeavor to work harder in the face of such competition.
http://www.wordspring.com/yogiberaquotes.html variety"
Good one,
Fred
I empathize, MBK.
The op-amp is a classic example of a very complex circuit which mfrs portray as very simple. Op-amps are not usually designed for audiophile applications, nor are they usually designed by experienced audio designers. Mfrs make most of their money by selling these things to other industries, where parameters like power consumption, power bandwidth, stability, noise, reliability, price, size, availability are primary motivators.
So the amateur should not make the mistake of believing something is fit for purpose just because it is made by a respected name in the components industry.
Besides, in order to reduce size and manufacturing cost, opamps are made on a wafer of silicon. Recall that silicon is a semi-conductor and thus is inherently NON-LINEAR. This means resistors in an opamp will be inferior to discrete carbon or metal film resistors. Similarly, silicon capacitors are non-linear and cannot be made very large due to wafer realestate consumption - only tens of pF. Try regulating a circuit with just a few pF to play with. Inductors are almost impossible to make in silicon wafers without huge area consumption and so they are avoided. You have no control over the matching of parts within an op-amp.
The good aspects of opamps is that they are small - reducing inductance - and certain performance parameters are fairly tightly controlled or tested-in by the mfrs. It is easy to make a functioning circuit first time because you have less components to deal with and connect up.
The op-amp is a classic example of a very complex circuit which mfrs portray as very simple. Op-amps are not usually designed for audiophile applications, nor are they usually designed by experienced audio designers. Mfrs make most of their money by selling these things to other industries, where parameters like power consumption, power bandwidth, stability, noise, reliability, price, size, availability are primary motivators.
So the amateur should not make the mistake of believing something is fit for purpose just because it is made by a respected name in the components industry.
Besides, in order to reduce size and manufacturing cost, opamps are made on a wafer of silicon. Recall that silicon is a semi-conductor and thus is inherently NON-LINEAR. This means resistors in an opamp will be inferior to discrete carbon or metal film resistors. Similarly, silicon capacitors are non-linear and cannot be made very large due to wafer realestate consumption - only tens of pF. Try regulating a circuit with just a few pF to play with. Inductors are almost impossible to make in silicon wafers without huge area consumption and so they are avoided. You have no control over the matching of parts within an op-amp.
The good aspects of opamps is that they are small - reducing inductance - and certain performance parameters are fairly tightly controlled or tested-in by the mfrs. It is easy to make a functioning circuit first time because you have less components to deal with and connect up.
There's also a LOT of chaff. Say, was it Marsh who wrote that article about shunt regulators in AA with a design that didn't/couldn't work? I remember a follow up article from someone who pointed out the fatal flaws (no feedback, no series resistance). Marsh's response was something to the effect of, "Well, MINE worked, but I've lost all my notes on it."
Jan's regulator designs were first-rate.
Are you guys kidding?
"Op-amps are not usually designed for audiophile applications, nor are they usually designed by experienced audio designers. Mfrs make most of their money by selling these things to other industries, where parameters like power consumption, power bandwidth, stability, noise, reliability, price, size, availability are primary motivators."
This is too much....... There is really not that much in priciple for a well designed op amp not to be good for audio. John Curl holds Scott Wurcer at Analog Devices to be a serious audio geek and says he has built discrete J fetss designs for his home audio system. Scott has designed some of ADs most interesting Op amps
From the OPA604 data sheet:
The OPA604 is a FET-input operational amplifier designed for enhanced AC performance. Very low distortion, low noise and wide bandwidth provide superior performance in high quality audio and other applications requiring excellent dynamic performance."
Go pull the top off a DVD player and tell me what logo you see on the op amps and how many, and tell me the audio market place is not a significant customer. Many of the op amps are quite good now and sonically bettered by someone with extremely good design skills in discrete design. I design with both approaches depending on the application.
Is it the British counterpart for April fools day today or something?
"Op-amps are not usually designed for audiophile applications, nor are they usually designed by experienced audio designers. Mfrs make most of their money by selling these things to other industries, where parameters like power consumption, power bandwidth, stability, noise, reliability, price, size, availability are primary motivators."
This is too much....... There is really not that much in priciple for a well designed op amp not to be good for audio. John Curl holds Scott Wurcer at Analog Devices to be a serious audio geek and says he has built discrete J fetss designs for his home audio system. Scott has designed some of ADs most interesting Op amps
From the OPA604 data sheet:
The OPA604 is a FET-input operational amplifier designed for enhanced AC performance. Very low distortion, low noise and wide bandwidth provide superior performance in high quality audio and other applications requiring excellent dynamic performance."
Go pull the top off a DVD player and tell me what logo you see on the op amps and how many, and tell me the audio market place is not a significant customer. Many of the op amps are quite good now and sonically bettered by someone with extremely good design skills in discrete design. I design with both approaches depending on the application.
Is it the British counterpart for April fools day today or something?
Marsh mellowed..........
I remember the circuit and error but I don't remember that response.......... I do seem to remember the fix being sent in by someone. Oh by the way Dick Marsh is a highly successful high end audio designer today with favorably reviewed designs and is well respected by a few designers I have talked to. I guess he over came that design problem.
It seemed to me to be a schematic error and an obvious one at that to me and a several others, and not a real design flaw. It is an easy thing to do and a lot of stuff is built before the schematic is drawn. I found a schematic error on a telecom board a couple of years ago (non and inverting op amp inputs swapped) and every one swore it couldn't happened with design reviews, computer data bases for the schematic and PCB layout and the board was working for god's sake! They said I must be mistaken. Other than the fact that I know how an op amp works......... The pulled a board from the lab and ohmed it out and guess what.......... the schematic was wrong. Guess what I have seen a circuit in several op amp application data sheets that both Nelson Pass and I scratched our heads over. I thought it should work he didn't. I ran Spice and Nelson was right. I found the original circuit and it included coupling caps the op amp version. If I give you the info will you call Analog Devices and tell them its wrong and tell me what they plan to do about it?
http://www.marshsounddesign.com
He also originated the MIT multicap. I can't say much for his taste in shirts though..........
http://www.marshsounddesign.com/contactset.htm
SY said:
I remember the circuit and error but I don't remember that response.......... I do seem to remember the fix being sent in by someone. Oh by the way Dick Marsh is a highly successful high end audio designer today with favorably reviewed designs and is well respected by a few designers I have talked to. I guess he over came that design problem.
It seemed to me to be a schematic error and an obvious one at that to me and a several others, and not a real design flaw. It is an easy thing to do and a lot of stuff is built before the schematic is drawn. I found a schematic error on a telecom board a couple of years ago (non and inverting op amp inputs swapped) and every one swore it couldn't happened with design reviews, computer data bases for the schematic and PCB layout and the board was working for god's sake! They said I must be mistaken. Other than the fact that I know how an op amp works......... The pulled a board from the lab and ohmed it out and guess what.......... the schematic was wrong. Guess what I have seen a circuit in several op amp application data sheets that both Nelson Pass and I scratched our heads over. I thought it should work he didn't. I ran Spice and Nelson was right. I found the original circuit and it included coupling caps the op amp version. If I give you the info will you call Analog Devices and tell them its wrong and tell me what they plan to do about it?
http://www.marshsounddesign.com
He also originated the MIT multicap. I can't say much for his taste in shirts though..........
http://www.marshsounddesign.com/contactset.htm
You are dissappointing me now Fred.
You are doing just what I said you shouldn't: taking the datasheet as read. And what evidence is there of the sound quality of Scott Wurcer's home brew amps? Interesting that he is designing his amps using dicrete parts.
Having just looked at the datasheet (on the web) there isn't much I can tell about the audio performance of the OPA604. There is no detailed schematic, there are no performance graphs, the single distortion spec is a THD at 1kHz with closed loop gain of 1 at 1kHz and a 1k resistive load.
I agree that some op-amps are very good. I never said none were good. However, I have never heard of a particularly good power op-amp (the theme of this thread).
I think you too quickly dismiss the inherent problems with op-amps, though. All silicon circuits really do constrain the designer. And the need for low power consumption specs leads to the use of class B circuits. You'd think the ultimate op-amp would at least be single-ended class A.
You are doing just what I said you shouldn't: taking the datasheet as read. And what evidence is there of the sound quality of Scott Wurcer's home brew amps? Interesting that he is designing his amps using dicrete parts.
Having just looked at the datasheet (on the web) there isn't much I can tell about the audio performance of the OPA604. There is no detailed schematic, there are no performance graphs, the single distortion spec is a THD at 1kHz with closed loop gain of 1 at 1kHz and a 1k resistive load.
I agree that some op-amps are very good. I never said none were good. However, I have never heard of a particularly good power op-amp (the theme of this thread).
I think you too quickly dismiss the inherent problems with op-amps, though. All silicon circuits really do constrain the designer. And the need for low power consumption specs leads to the use of class B circuits. You'd think the ultimate op-amp would at least be single-ended class A.
Well, the follow up (a full article, not a correction) appeared in AA 4/88, p46 (John Adelsbach, "Push-Pull Shunt Regulator Revisited"). The entire concept of the regulator was flawed- it wasn't a simple schematic mistake, unless that schematic mistake coincided perfectly with circuit board mistakes. Marsh claimed that the prototype worked, but was unable to provide any documentation or the prototype itself.
and on that note...
Has anyone realised the patent for the Quad 'current dumping' feed-forward correction system first used on the 405 is due to expire this year (or has already?).
If there was an output stage topology that should be ideal for DIY, this could be it. I have a sneaking suspicion it never quite realised it's full potential in Quad's hands.
Andy
Has anyone realised the patent for the Quad 'current dumping' feed-forward correction system first used on the 405 is due to expire this year (or has already?).
If there was an output stage topology that should be ideal for DIY, this could be it. I have a sneaking suspicion it never quite realised it's full potential in Quad's hands.
Andy
Keeping your OPampTIONS open.......
You are dissappointing me now Fred.
You are doing just what I said you shouldn't: taking the datasheet as read. And what evidence is there of the sound quality of Scott Wurcer's home brew amps? Interesting that he is designing his amps using discrete parts.
Having just looked at the datasheet (on the web) there isn't much I can tell about the audio performance of the OPA604. "
I don't what in particular you are referring by 'taking the data sheet as read." I assume you mean thinking the ap note circuit worked initially. At first inspection and with some simple nodal analysis it appeared to and with the coupling caps it would. I ran a Spice model and got the same results as Mr. Pass as to the problem. I was also curious enough to find the original design to see why the circuit had been included and where they went wrong.
I have never seen a Class B stage in any op amp and for most typical loads in a preamp they are class A or class AB. 5 volts into 10K is 0.5 mA and far below the bias current of most op amps used for audio. The AD 825 is biased at bout 6.5 mA and the open loop output impedance is 8 ohms indicating a good amount of bias through the out put stage. The AD811 has a supply current of 16.5mA at +/- 15 volts and requires a heat sink at these voltages. It is often used as an output stage in composite op amps.
Scott Wurcer has designed several very good sounding op amps, has the respect of John Curl, and might design with jfets at home
because he is tired of playing with op amps at work. How many of you do for a hobby what you do for work? I would be quite happy to take design audio advice from him. Someone who can design op amps must know enough to design great discrete circuits. The converse is probably not true very often. The demands for low distortion and low DC offsets and drift are orders of magnitude better than many discrete designs can deliver due to the thermal coupling and close transistor matching of several transistors fabed on the same die. It is difficult with a discrete transistor design to get above a few tens of MHz before board parasitic will give you major headaches. i have seen op amp with gain bandwidth product of a GigaHertz!
The real exploitation of the virtues and minimization of the drawbacks backs of op amps comes from composite circuits. The first in loop external output stage I ever saw was the use of a jfet follower with jfet current source inside the feedback loop of an op amp. This was done by PS audio maybe 15 years ago. I have seen tubes, mosfets, bipolar transistors, video op amps inside the loop of op amps. Op amp based power supplies have been in tube amps and preamps for about 20 years and even used in some Audio Research tube products. I have lost count of the number of op amp based servos to DC couple discrete transistor designs.
I design with discretes but writing off op amps with the level of design and new composite circuits in use would be very foolish
and may be at the point where op amp based designs with enough attention to details could be better than many less rigorously discrete designs. Jocko and I were just looking at some of the new jfet op amp data sheets and notice the specs for even the non exotic designs would be state of the art a few years ago. The last five years have seen some amazing improvements to op amps and op amp circuits and a serious audio designer would ignore them at his peril since the gap is rapidly narrowing.
You are dissappointing me now Fred.
You are doing just what I said you shouldn't: taking the datasheet as read. And what evidence is there of the sound quality of Scott Wurcer's home brew amps? Interesting that he is designing his amps using discrete parts.
Having just looked at the datasheet (on the web) there isn't much I can tell about the audio performance of the OPA604. "
I don't what in particular you are referring by 'taking the data sheet as read." I assume you mean thinking the ap note circuit worked initially. At first inspection and with some simple nodal analysis it appeared to and with the coupling caps it would. I ran a Spice model and got the same results as Mr. Pass as to the problem. I was also curious enough to find the original design to see why the circuit had been included and where they went wrong.
I have never seen a Class B stage in any op amp and for most typical loads in a preamp they are class A or class AB. 5 volts into 10K is 0.5 mA and far below the bias current of most op amps used for audio. The AD 825 is biased at bout 6.5 mA and the open loop output impedance is 8 ohms indicating a good amount of bias through the out put stage. The AD811 has a supply current of 16.5mA at +/- 15 volts and requires a heat sink at these voltages. It is often used as an output stage in composite op amps.
Scott Wurcer has designed several very good sounding op amps, has the respect of John Curl, and might design with jfets at home
because he is tired of playing with op amps at work. How many of you do for a hobby what you do for work? I would be quite happy to take design audio advice from him. Someone who can design op amps must know enough to design great discrete circuits. The converse is probably not true very often. The demands for low distortion and low DC offsets and drift are orders of magnitude better than many discrete designs can deliver due to the thermal coupling and close transistor matching of several transistors fabed on the same die. It is difficult with a discrete transistor design to get above a few tens of MHz before board parasitic will give you major headaches. i have seen op amp with gain bandwidth product of a GigaHertz!
The real exploitation of the virtues and minimization of the drawbacks backs of op amps comes from composite circuits. The first in loop external output stage I ever saw was the use of a jfet follower with jfet current source inside the feedback loop of an op amp. This was done by PS audio maybe 15 years ago. I have seen tubes, mosfets, bipolar transistors, video op amps inside the loop of op amps. Op amp based power supplies have been in tube amps and preamps for about 20 years and even used in some Audio Research tube products. I have lost count of the number of op amp based servos to DC couple discrete transistor designs.
I design with discretes but writing off op amps with the level of design and new composite circuits in use would be very foolish
and may be at the point where op amp based designs with enough attention to details could be better than many less rigorously discrete designs. Jocko and I were just looking at some of the new jfet op amp data sheets and notice the specs for even the non exotic designs would be state of the art a few years ago. The last five years have seen some amazing improvements to op amps and op amp circuits and a serious audio designer would ignore them at his peril since the gap is rapidly narrowing.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.