Member
Joined 2009
Paid Member
I think the 'THD is bad' thing is overblown. It was a thing of the past - when solid state amplifiers took over from valves and proved to sound like **** people asked what do we need to fix ? Manufacturers found a metric that they could use that was simple and easy to define - THD - and I have the view that it did prove to be a useful tool in advancing the art. Amplifier designers had to pay attention to it, and that lead them to look inside the THD metric at the component parts, to worry about slewing and other things. It was a great for marketing and amplifier manufacturers who paid attention to their designs were rewarded. Didn't the Japanese take over most of the industry in the 80's. So I see THD as something that brought some good. Once the distortion issue has been 'solved' it ceased to be an effective metric. And today I don't think my kids would every look at the spec of their music system anyhow - they are going to ask 'what sounds good' and the arbiter of that is Joe Public through on-line reviews. So in the end it isn't too bad.
Now speaker manufacturers didn't adopt a simple distortion metric. There are bucket loads of really **** speaker drivers on the market. Not only do they have bad distortion but they have awful frequency responses which are often fudged in spec sheets through smoothing etc.
Now speaker manufacturers didn't adopt a simple distortion metric. There are bucket loads of really **** speaker drivers on the market. Not only do they have bad distortion but they have awful frequency responses which are often fudged in spec sheets through smoothing etc.
The Quad ESL is low thd. It has some weird beaming distortions which are unacceptable to some. THD is very useful if analysed. One way to do that is to relax the negative feedback loop. In a well designed amp everything is what it should be.Double the gain and double the thd is expected. There is a debate as to whether high order distortion is created by negative feedback. It looks to be true. More careful investigation mostly shows it is the result of lack of open loop gain when frequency limiting bites. Used carefully thd can tell us plenty. Most amplifiers have Ideal gain which I doubt is ever found in a shop bought example. Using thd you can upgrade a design. I would bet 99% of amplifiers are not as good as they should be. For very little money you can change what you buy. Douglas Self says we should be able to find evidence of all distortion mechanisms from thd. I agree. However you have to become experienced in looking and listening.
The Quad 405 had lack of current as a type of distortion. The.405.2 less so. It was a very week 100 watts. A parallel bridge 405 was very different.
The Quad 405 had lack of current as a type of distortion. The.405.2 less so. It was a very week 100 watts. A parallel bridge 405 was very different.
Reentrant distortion as best as I have worked out does not exist. It is just lack of correcting negative feedback. What happens is as the gain is reduced due to large negative feedback the frequency end closest to the Nyquist limit isn't corrected. This looks to be the distortion was invented by the negative feedback. It isn't. For example. if two pole compensation is added to the VAS the amount of negative feedback that can be used is greater without risk of instability given shrewd design. Suddenly the distortion is lower with no significant active component change.
Doubtless both arguments have justification. The point of interest is investigating the problem can give solutions.
Doubtless both arguments have justification. The point of interest is investigating the problem can give solutions.
Member
Joined 2009
Paid Member
There are a lot of generalizations floating around about negative feedback which obscure the reality. There were calculations many years back before we could run spice on a desktop computer and the simplifying assumptions lead to conclude that feedback 'creates' higher order harmonics thus requiring ever more feedback to 'fix' the problem it creates. This is just one example as there are many of them. I've just not built or heard enough amplifiers to be 100% sure, but my experience is that for long term listening of a broad range of music types I prefer low distortion high feedback amplifiers. For simple music (let's call it Jazz for now) there is a pleasure to be had with a no feedback amp with 'character' (aka distortion) and under these conditions the distortion is simply not evident to my ears. Add in complex music, then turn up the wick and the IM products rear their ugly heads.
As you say well worth debating. Mine is by observation using simple thd tests. Mostly to note changes. Even so we should really take a current waveform. If distortion is different voltage compared with current waveform it asks a question. Take the voltage waveform into the VAS for example. A sinewave becomes triangular. At the VAS. output the waveform is undistorted. Looking at the current waveform into the VAS it also is undistorted. Most people are happy to say that's the joy of transconductance. An I.to V converter. Yes and no. If like John Lindsley Hood you doubt that you can design better amplifiers.
Member
Joined 2009
Paid Member
Well, the VAS is not amplifying the signal, but the error signal created by the input stage which compares the input signal and output fed back to create something else. It is, in a way of thinking, pre-distorted to account for the distortion that will occur at the VAS and output stage (that's not strictly correct but trying to find the words here...)
For me THD is still a bad thing. Anything that is audible and negatively affect sound is bad. So we should know the other distortions beside THD, we should know our threshold expectations (for each distortion) and we should know how to set our priorities when we have to choose between these distortions.
By listening, it is easy to hear improvement when our system is still below 'standard'. But after we reach a certain quality threshold, it is not easy anymore to make improvement by listening.
So before reaching the threshold, THD is not a direct objective, hence not important. But after the threshold is surpassed, the THD is important as long as it is audible. To know the audibility threshold for THD is not easy because it is almost impossible to change only the THD part in an amplifier. But from experience i decided that as long as i can lower it to 0.002% (from simulation point of view) then i will do it because i think it is some kind of my audibility threshold.
Member
Joined 2009
Paid Member
Yes. My latest amp THD is 0.0038% for 0.2v input and 0.0023% for 0.7v input (around 20v output). I have critically listened to this setup for 3 days now by listening various music and i'm very satisfied. Yes, it takes me days to listen to only one change in the schematic, to make sure that there will be no problem with any quality measures.
Because the THD is still above 0.002% i can feel it and i know it can sound better. My next change is to reduce the THD by lowering the output offset (currently around -15mV). Only way to do it without breaking other thresholds is to balance the LTP by compromising the input impedance. Currently the input resistor is 15k and my plan is to change this in two steps, that is 12k then 10k. Because currently i don't use buffer i need to be careful to not draw a wrong conclusion when the sound is not better when the circuit is in need of a buffer. I believe tho that distortion will be lower and that i will hear an improvement.
Yes we can make better amplifiers. There might be a new 'theory' that will allow designers to design better amplifiers. May be in a few years, may be not. It is possible because it is logical. It is impossible because in order to fully believe the theory one might need proof that it is audible. I don't believe they can get suitable respondent easily.
JLH was very good. The problem with him and many other veterans is that they don't have sufficient experience working with fast transistors or wide bandwidth amplifiers in order to have suitable theory or conclusion.
Hmmm, I'm not so sure.
After a change in regulator design in my active crossover, I have reversed my ranking of the Leach and the Mooly amp (in post 1109).
And this is the whole problem with subjective assessments. OTOH, as a final assessment only the subjective experience matters!
At what signal level?
At low signal levels the % speaker distortion will reduce.
I have owned LS50s and used them with a high pass filter at 120Hz. This will reduce the distortion significantly compared to run full range. Used this way and with a bit of eq to reduce the output ~2.5Kz, LS50s are rather better than ESL63s IMO.
In addition to this you need to consider that some speaker manufacturers encourage a bit of 2nd harmonic to make the bass from ported speakers sound tighter.
Last edited:
Pre distortion is a very interesting concept. It is something already gone correcting itself. In valves using pre distortion we can do away with that comprise. A pentode and triode have opposite errors. If the anode of a high power type feeds a power triode a more linear curve is had. If an active anode load used to the pentode 0.2% thd one watt without feedback of any sort. Not even local if the triode taken to be a device. Driving Quad ESL 63 the lack of damping factor not too obvious. A little less power than JLH.
Recently I was given some small speakers using three inch Tonnigen. titanium drivers. They sound great except they have the metal cone blandness. I added Motorola clone piezoelectric horns as super tweeters. A twelve ohm resistor across them and capacitor coupled at twelve kHz. The result is low distortion even though debatable that it could be heard. Even the phase matters. At my age remembering I started electronics in the 50s I clealy hear it. Not a perfect result, very close. In my friend's house it's the space she has. The point being filtering the bottom end of a speaker should give lower distortion.
Recently I was given some small speakers using three inch Tonnigen. titanium drivers. They sound great except they have the metal cone blandness. I added Motorola clone piezoelectric horns as super tweeters. A twelve ohm resistor across them and capacitor coupled at twelve kHz. The result is low distortion even though debatable that it could be heard. Even the phase matters. At my age remembering I started electronics in the 50s I clealy hear it. Not a perfect result, very close. In my friend's house it's the space she has. The point being filtering the bottom end of a speaker should give lower distortion.
Member
Joined 2009
Paid Member
One aspect of this amplifier that I find especially interesting is the RC 'phase' networks placed in the bases of the output stage. I've never seen this kind of compensation anywhere else and it was not very easy to determine why JV had used them at all. Michael (in post 98) first started to cast some light on this. I still don't understand why this solution, why not a different approach, why nobody else uses it and what are the implications on sound etc.
Last edited:
That's why i challenged you to design your own amp with this topology without copying Naim. With other topology i don't see the need for this trick (it has many compromises such as stability against capacitive load). But this topology is difficult. Question is, how will you create a good sounding amp with this topology without the trick. If you don't understand how it work, how can you use it? It makes your amp a mess.
As said it could be to match the reactive load of Quad ESL speakers. With modern transistors I doubt they do a good job. One possibility is it was a design idea of Alan Mornington West. Like with my boss I like to have things only I understand. In one circuit I build I use a reactor circuit. This is not unlike how the early Citroen 2CV suspension works except it's electronic. No one has ever asked me why!
Member
Joined 2009
Paid Member
I have no plans to design any more Class AB amplifiers. I do have a design not yet built and tested (TGM11) and another design that should be respun into a proper PCB because I remember it tasting rather good (TGM3) but I can't identify anything about the sound of my TGM8 that should be improved so I've called it a day for the forseeable future. I am going to look at some tube amps and some pre-amp projects.
Nevertheless, my TGM10 should be finished up and I like to learn and understand. And I don't understand these phase networks
Member
Joined 2009
Paid Member
The simulations referred to earlier were into an 8 Ohm purely resistive load - and the phase networks were necessary. I'm not sure speaker choice is a factor in the decision.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.