Re, post 1393
It would seem we don't know THE right way to build an amp, but we know SOME very good ones. Each has different tradeoffs, and what clues from an imperfect signal cause us to interpret what we hear as pleasing or correct are not uniform. So, different proponents and detractors who due to their personal perceptions, hold their views passionately.
Science fact is within a range of acceptable statistical deviation. It is not the yes or no we would like it to be. Is light a particle of a wave? The answer is a clear "it depends."
This site is a free and open public forum. It has many of us wallowing in our stupidity, insightful views from those with the real practical experience, and everyone else. Part of being educated is both recognizing the difference as well as respecting it.
Please, let the debates continue! Besides like one of the great Peanuts cartoons had Linus say "Some of those old wives were pretty sharp"
It would seem we don't know THE right way to build an amp, but we know SOME very good ones. Each has different tradeoffs, and what clues from an imperfect signal cause us to interpret what we hear as pleasing or correct are not uniform. So, different proponents and detractors who due to their personal perceptions, hold their views passionately.
Science fact is within a range of acceptable statistical deviation. It is not the yes or no we would like it to be. Is light a particle of a wave? The answer is a clear "it depends."
This site is a free and open public forum. It has many of us wallowing in our stupidity, insightful views from those with the real practical experience, and everyone else. Part of being educated is both recognizing the difference as well as respecting it.
Please, let the debates continue! Besides like one of the great Peanuts cartoons had Linus say "Some of those old wives were pretty sharp"
Could you list some of these papers? The researchers with whom I'm familiar generally are quite explicit about these variables.
even personal preference can be measured. science can account for that "subjective" deviation of preference from mean. can't see any reason why it shouldn't.
well, it would seem very ambiguous and hard to define. but at the same time I see lots of people that would apparently die defending their beliefs. kind of a contradiction.
I've said that myself in the past. a forum is what it is and nothing more. but there are times when one simply can't ignore that there are some people obsessively repeating over and over and over and over the same things without even trying to care about scientific proof.
yes, my signature reflects my current view on some of the discussions here.
I can't quote exact names right now but if you insist I can search.
just yesterday I was reading this paper on the correlation of sensation of harshness/preciseness with spectral distribution. guess what, the only subject used was the writer himself. also I can't remember many papers where the used equipment was mentioned. for all I know, they may have been using Logitech speakers.
now on a more philosophical note. I know people that sold their expensive, new speakers and switched to speakers that were made in the early '90s. I can't think of many people that sold their current production cars to return to some '90s model, except for nostalgic reasons. same thing with TVs or whatever. audio doesn't seem to obey the same rules and that is for obvious reasons, IMO.
A couple cites would be nice so that I can get an idea of what you're talking about. Thanks!
I doubt if this exists. Is there A right way to build a bridge, or a car? All engineering is compromise. That is why I find it amusing when people talk about the 'ultimate' amp/DAC/speaker etc.tvrgeek said:
Yes, it depends on what question you ask. If you ask "which slit did it go through?", the answer is "both; it is a wave". If you ask "where on the screen did it end up?", the answer is "here; it is a particle". Physicists have learnt to live with this. Others have not, so try to get round the results of experiments.
You are banging again in wide open gates.
All my power amp and mic pre designs since I come back to audio design in 2006 combine tubes and semiconductors. Some exceptions are SS only designs, like condenser mike with no tubes, crossover, mixer, other toys... However, before that (in 70-80'th) I used SS only because I had belief that transistors are more convenient, and tubes are obsoleted. Now I firmly believe that tubes will never be obsoleted. They have advantages for analog audio applications, especially when properly assisted by semiconductors.
What is usually called "Hybrid", is some weird combination of bad tube preamp with no feedback, high output resistance, and bad SS output followers presenting non-linear load to that poor tube. I don't play such silly games, you may search for my Tower-III thread here, it is totally different approach. But the word "Hybrid" is discredited, unfortunately, so it is hard to convince people even to consider to check how well hybrid amp can sound, when properly cooked by talented chef.
It would be fun to raise a pint or two with many on this page. 
The gentleman from Digi-Key responded to my unpleasant letter about On-Semi. My suggestion was to take a look at his supply chain. I doubt someone is counterfitting a 7 cent transistor, but selling seconds at full price seems very likely. Either that, or On-Semi (Sanyo) builds garbage intentionally. Digi-Key is stand up. I wonder if On-Semi is.
Mr. PP, Guess you never have had the pleasure of driving an old Morgan or TVR. You drive a shoebox Chevy for nostalgia, you drive an old British roadster for the thrill. 135 in a 64 Plus-4 is far more adrenalin than any new car. Alas, my knees can't deal with a clutch anymore.
So we still don't know which objective measure correlates to some perceived deficiency. Thinking outside the box per say, what effect would result in playing with the resonance or Q between the amp output filter and the speaker? We add capacitance to cancel driver inductance, but is that a good thing? We add indictance to keep the amp happy with the capacitive load. But the speaker load is unknown to the amp designer, so how do we pick the best output filter? Could this be one of the reasons otherwise very good amps get different results on different speakers? Or is it just something to make ugly square waves in magazine reviews and has no effect on the sound? It is tough to design half a system and tweeters do some really nasty things, so I worry about the interaction.
The gentleman from Digi-Key responded to my unpleasant letter about On-Semi. My suggestion was to take a look at his supply chain. I doubt someone is counterfitting a 7 cent transistor, but selling seconds at full price seems very likely. Either that, or On-Semi (Sanyo) builds garbage intentionally. Digi-Key is stand up. I wonder if On-Semi is.
Mr. PP, Guess you never have had the pleasure of driving an old Morgan or TVR. You drive a shoebox Chevy for nostalgia, you drive an old British roadster for the thrill. 135 in a 64 Plus-4 is far more adrenalin than any new car. Alas, my knees can't deal with a clutch anymore.
So we still don't know which objective measure correlates to some perceived deficiency. Thinking outside the box per say, what effect would result in playing with the resonance or Q between the amp output filter and the speaker? We add capacitance to cancel driver inductance, but is that a good thing? We add indictance to keep the amp happy with the capacitive load. But the speaker load is unknown to the amp designer, so how do we pick the best output filter? Could this be one of the reasons otherwise very good amps get different results on different speakers? Or is it just something to make ugly square waves in magazine reviews and has no effect on the sound? It is tough to design half a system and tweeters do some really nasty things, so I worry about the interaction.
Tvr, you are just naive. Look at the complete specs for the 2N2222, and learn and grow. Second, the POSITION that this part is used can be filled by just about 100 different parts without any trouble at all.
Your second-guessing WHY we add Zobel networks and series inductors is below 'sophomoric', we are fully understanding why we NEED to add these extra components, and we wish that we did not have to.
For the record, the REASON a difficult load looks so bad with a given amp sometimes, has mostly to do with DAMPING FACTOR, not Zobel networks, etc. However, I must say that ANY amp using over 2uH inductance is potentially problematic, like the Hafler 120 that you are fixing. Bad design! Even Bob Cordell would agree with me on this, (I think).
Your second-guessing WHY we add Zobel networks and series inductors is below 'sophomoric', we are fully understanding why we NEED to add these extra components, and we wish that we did not have to.
For the record, the REASON a difficult load looks so bad with a given amp sometimes, has mostly to do with DAMPING FACTOR, not Zobel networks, etc. However, I must say that ANY amp using over 2uH inductance is potentially problematic, like the Hafler 120 that you are fixing. Bad design! Even Bob Cordell would agree with me on this, (I think).
Last edited:
will do so tomorrow, right now I've just finished emptying a couple of beers and I'm lazy
no I've never driven a TVR or a Morgan. but I'm not sure the analogy stands.
I'm not into cars but I know a thing or two about motorcycles (and own one btw).
look at this video of a BMW S1000RR doing 50 to 199 mph. please tell me that you could get that kind of thrill factor from a bike 20 years ago
2011 BMW S1000RR Top Speed - 50 to 190mph - YouTube
I can't see any reasons why interactions can't be scientifically studied, measured, so on and so forth
Last edited by a moderator:
Harman has done a lot of careful work in subjective evaluation of loudspeakers, with trained and untrained listeners, using statistical tools and double-blind testing, and the effort continues. Floyd Toole (retired but consulting) and Sean Olive (currently directing the research) have published extensively. A good summary of many of the conclusions from years of work can be found in Floyd's book Sound Reproduction (978-0-240-52009-4).
This stuff is a lot of work. I've chatted with Sean about adding amplifiers to the components under test, and he's conceded that it needs to happen sooner or later. Floyd used to say that he was focused on the 3dB effects, which loudspeakers and rooms can readily provide
A lot of people don't like the conclusions reached from the research. I find that I can make judgments about loudspeakers pretty readily, but source components and amplifiers I usually have to live with for a time --- unless there is something terribly wrong. So rapid A/B/X comparisons of decent electronic components are less useful, for me at least.
I'm puzzled by the intensity with which some audiophiles claim the virtues of certain components (cables in particular come to mind). But although I think my ears, except for age-related HF rolloff, are pretty decent, I rarely can hear the differences stated, even when I would like to, and even when I know to what I'm listening. But I'm not prepared to discount all subjective accounts just because they seem implausible.
A while back I tried to test the waters for some electronics I was designing, and made a unity gain buffer employing some of my favorite topologies. I persuaded an audiophile with a very fancy system to allow me to insert the buffer in his signal chain. This was someone who insisted he could barely tolerate solid-state. The preamp feeding the buffer had 6AS7s in the output and some fairly substantial output coupling caps.
The general assessment was a surprised one: the man agreed that I hadn't ruined the sound of his system, EVEN though he knew that he was listening through the stage. But after he listened for a while, he said "I think I may hear just a tiny bit less bass extension." I thought how can that be, as the whole thing is d.c.-coupled. But --- further investigation revealed that the input impedance of the amp to which the preamp output was normally directly connected was slightly higher than the d.c. termination input resistor on my board, and this translated to a quite-low-frequency difference in rolloff and phase response. I was mightily impressed at the man's powers of perception.
Brad Wood
This stuff is a lot of work. I've chatted with Sean about adding amplifiers to the components under test, and he's conceded that it needs to happen sooner or later. Floyd used to say that he was focused on the 3dB effects, which loudspeakers and rooms can readily provide
A lot of people don't like the conclusions reached from the research. I find that I can make judgments about loudspeakers pretty readily, but source components and amplifiers I usually have to live with for a time --- unless there is something terribly wrong. So rapid A/B/X comparisons of decent electronic components are less useful, for me at least.
I'm puzzled by the intensity with which some audiophiles claim the virtues of certain components (cables in particular come to mind). But although I think my ears, except for age-related HF rolloff, are pretty decent, I rarely can hear the differences stated, even when I would like to, and even when I know to what I'm listening. But I'm not prepared to discount all subjective accounts just because they seem implausible.
A while back I tried to test the waters for some electronics I was designing, and made a unity gain buffer employing some of my favorite topologies. I persuaded an audiophile with a very fancy system to allow me to insert the buffer in his signal chain. This was someone who insisted he could barely tolerate solid-state. The preamp feeding the buffer had 6AS7s in the output and some fairly substantial output coupling caps.
The general assessment was a surprised one: the man agreed that I hadn't ruined the sound of his system, EVEN though he knew that he was listening through the stage. But after he listened for a while, he said "I think I may hear just a tiny bit less bass extension." I thought how can that be, as the whole thing is d.c.-coupled. But --- further investigation revealed that the input impedance of the amp to which the preamp output was normally directly connected was slightly higher than the d.c. termination input resistor on my board, and this translated to a quite-low-frequency difference in rolloff and phase response. I was mightily impressed at the man's powers of perception.
Brad Wood
headphones which don't stage between my ears
I have specific request;
a headphone amp topology which sound stages like a pair of speakers; that is the sound stage is not between the ears. Maybe it has already been done,,,
will do so tomorrow, right now I've just finished emptying a couple of beers and I'm lazy
no I've never driven a TVR or a Morgan. but I'm not sure the analogy stands.
I'm not into cars but I know a thing or two about motorcycles (and own one btw).
look at this video of a BMW S1000RR doing 50 to 199 mph. please tell me that you could get that kind of thrill factor from a bike 20 years ago
2011 BMW S1000RR Top Speed - 50 to 190mph - YouTube
I can't see any reasons why interactions can't be scientifically studied, measured, so on and so forth
I have specific request;
a headphone amp topology which sound stages like a pair of speakers; that is the sound stage is not between the ears. Maybe it has already been done,,,
I've fixed the misquote.
John,
It was my understanding we add the output inductor to help maintain stability when driving a capacitve load. The smaller the better so long as we don't have an oscillator. For real world problems, I assume it winds up way over the calculated perfect for safety margins. As every speaker is different, so would be the load. So that was my question, if for specific amp-speaker system, should this be optimized. I gather you are saying no.
On the speaker end we add a zobel to deal with the rising inductance of the driver so the crossover load is predictable and does something close to what we asked. I have had better results by using steeper crossovers, but every speaker builder has different ideas.
From Self and others, I thought any DF over 50 or so really had no advantage. Connectors, cables and crossovers can easily be half an ohm. Of course there may be some other effect than just output impedance that a high DF is indicative of I am unaware of.
Besides use of poor small signal transistors by today's standards, not matching the diff pair or current mirror, and putting 1.9W through a 2W resistor, I am all ears on other places where the 120 design can be improved within the basic topology and constraints of the design. I am repairing to have a baseline so as I work on the other channel, I have something to compare to. It seems to be pretty textbook in most respects, so a decent place to start. LTP input, cascode VAS, MOSFET output. It is a 40 year old design after all. It is a place where I can try and understand the practical ramifications of the Miller effect and Early effect. These were mentioned in tech school, but not studied. So I know what they are, but not how to compensate for them. I understand we want increase feedback above some set frequency to 100% and that phase shift in the feedback is what turns an amp into an oscillator, but not how to decide on what frequency and what phase compensation I do not get yet. I have read the chapters, but I need some bench time to look and play with it. I can see what the cap between the diff input does, I don't understand why it is wanted. Things like that. I can't put a scope on a textbook. One thing to test is that the ground for the input stage is back through the signal ground wiring to the input RCS and then back up to the central ground. I was going to see if there was any improvement in the power for the input if I ran a separate ground from the 100uf cap filters on the input rails to ground instead of to the signal input ground. Would low frequency distortion improve if the input blocking cap was much larger as Cordell suggests? Only measuring will tell. reading is one thing, doing is quite another.
It was my understanding we add the output inductor to help maintain stability when driving a capacitve load. The smaller the better so long as we don't have an oscillator. For real world problems, I assume it winds up way over the calculated perfect for safety margins. As every speaker is different, so would be the load. So that was my question, if for specific amp-speaker system, should this be optimized. I gather you are saying no.
On the speaker end we add a zobel to deal with the rising inductance of the driver so the crossover load is predictable and does something close to what we asked. I have had better results by using steeper crossovers, but every speaker builder has different ideas.
From Self and others, I thought any DF over 50 or so really had no advantage. Connectors, cables and crossovers can easily be half an ohm. Of course there may be some other effect than just output impedance that a high DF is indicative of I am unaware of.
Besides use of poor small signal transistors by today's standards, not matching the diff pair or current mirror, and putting 1.9W through a 2W resistor, I am all ears on other places where the 120 design can be improved within the basic topology and constraints of the design. I am repairing to have a baseline so as I work on the other channel, I have something to compare to. It seems to be pretty textbook in most respects, so a decent place to start. LTP input, cascode VAS, MOSFET output. It is a 40 year old design after all. It is a place where I can try and understand the practical ramifications of the Miller effect and Early effect. These were mentioned in tech school, but not studied. So I know what they are, but not how to compensate for them. I understand we want increase feedback above some set frequency to 100% and that phase shift in the feedback is what turns an amp into an oscillator, but not how to decide on what frequency and what phase compensation I do not get yet. I have read the chapters, but I need some bench time to look and play with it. I can see what the cap between the diff input does, I don't understand why it is wanted. Things like that. I can't put a scope on a textbook. One thing to test is that the ground for the input stage is back through the signal ground wiring to the input RCS and then back up to the central ground. I was going to see if there was any improvement in the power for the input if I ran a separate ground from the 100uf cap filters on the input rails to ground instead of to the signal input ground. Would low frequency distortion improve if the input blocking cap was much larger as Cordell suggests? Only measuring will tell. reading is one thing, doing is quite another.
- Status
- Not open for further replies.