I'm impressed with Digi-key's service too. I can't believe it but they dispatch from a Singapore location and deliver within as little as 4 days to a rural district in Oz and I'm 500 km from the airfreight terminal. Free too if the order is worth more than US $50. Even local distributors and postal services can't seem to beat that.
Member
Joined 2009
Paid Member
I came to realize, like JLH, that Class AB can be made so darn good that Class A is no longer the first choice.
It was with class AB that i came to realize that my system sounded much better than my headphones. But i still miss the class A experience, the enjoyment. I think I need to have a good class A and be able to compare side by side with whatever i design or build in class B. May be it is the sound of hot mosfet that i like so much
Right now i'm listening to Jewel's Foolish Games, kind of musics that i listened with class A amps, because they excel only with slow and simple music. I think Jewel in my system sounds more class A than what i can remember before, except for one thing: 'STRAIN'. With class B i periodically feel the strain during the music. Same thing i can hear with P3A. Higher bias will make it sound like class A, except that it is different with pure mosfet class A where increasing bias doesn't have trade off in any aspect of the sound. For sure i can understand when Rod mentions that class A P3B does not sound better than P3A.
Often, sound effects like "strain" are due to higher harmonics that can be even stronger than H3. You can play with simulation to predict what should be happening but nothing beats a sound card/software or real audio spectrum analyser to see what is happening in reality at various frequencies.
The VAS is sensitive to design values, parts types and tweaks in both class A and AB amplifiers. With class A, output stage distortion is usually low and quite predictable with good sim. models but class AB can be anything at all once you start playing with bias and component values/types where they affect crossover behaviour of the output stage.
Adjustments or tweaking of your particular build and modifications are much easier to perform and understand if you know what is happening. It's not necessarily expensive to do either.
The VAS is sensitive to design values, parts types and tweaks in both class A and AB amplifiers. With class A, output stage distortion is usually low and quite predictable with good sim. models but class AB can be anything at all once you start playing with bias and component values/types where they affect crossover behaviour of the output stage.
Adjustments or tweaking of your particular build and modifications are much easier to perform and understand if you know what is happening. It's not necessarily expensive to do either.
I trust simulation in combination with DC measurement and prototype listening (of course, combined with knowledge). Some people do not 'trust' LTP. I can understand that. And the way i use/treat LTP is different with almost everyone else (which you can see from their schematic anyway). This is one of the weakest links in P3A. It can be improved, or simply replaced (like Bigun did), but how about the CFP output? If the crossover issue is as audible as i currently think, it is incurable. But i'm still looking for a way to minimize the problem (one thing left is to use the output emitter resistor trick).
BTW, which class AB that has little high order harmonics, comparable to class A such as JLH??
Any output stage that "crosses over" will generate a spray of high harmonics according with the discontinuity to conductance as the output current transfers from one device to the other as it passes abruptly through the non-linear zero volt point. Bias and high GNFB is the way to reduce it to acceptable levels and class A is the only way to eliminate it in conventional power amplifiers. There are some effective ways to mask it though.
The LTP principally removes only even harmonic distortion. So there already, we have an emphasis on residual odd-harmonic distortion. That will mean some hardness to the sound unless the THD is so low that it is virtually inaudible. However, the usual solution has been to mask it by increasing H2, H4 etc. with tweaks that increase even-order harmonics, such as using a quasi-complementary or otherwise unbalanced output stage topology. Have a look at the NAD 310 output stage to see an example with another approach to this and surprisingly good sound quality. Note the singleton input stage as well: nad 310 schematic | Jacques | Flickr
Original Naim products followed the path of simply tweaking the LTP so that it didn't fully cancel even harmonics and the quasi-comp. output stage which augmented them - problem solved by DIY!
The LTP principally removes only even harmonic distortion. So there already, we have an emphasis on residual odd-harmonic distortion. That will mean some hardness to the sound unless the THD is so low that it is virtually inaudible. However, the usual solution has been to mask it by increasing H2, H4 etc. with tweaks that increase even-order harmonics, such as using a quasi-complementary or otherwise unbalanced output stage topology. Have a look at the NAD 310 output stage to see an example with another approach to this and surprisingly good sound quality. Note the singleton input stage as well: nad 310 schematic | Jacques | Flickr
Original Naim products followed the path of simply tweaking the LTP so that it didn't fully cancel even harmonics and the quasi-comp. output stage which augmented them - problem solved by DIY!
I think we're talking about CFP, not conventional one. You mentioned bias and high GNFB. Did you mean high bias of the driver, or correct bias of the CFP? What is your stands about the audibility of gm doubling? If everything else equal (esp. stability) what is your stand about the importance of high transconductance and its linearity, in relation with output resistance?
CFP, EF & Quasi-complementary are all conventional types of power amplifier topology. Here in the solid state forum, "unconventional" would usually imply class D or more unusual forms of amplification that are occasionally discussed.
GNFB is an acronym for global negative feedback. It seems to only be used in audio design and engineering discussion. Up to a point, the more feedback, the lower the distortion generated by the amplifier. It has nothing in particular to do with class A, AB or topology. The specific case of crossover distortion does not arise in class A so we might expect it to have lower overall distortion and less need for feedback. I hope this clarifies some of what I have been saying.
To answer your question about high bias and Gm doubling, I think we have to distinguish between the much higher bias used for class A and the lower level consistent with exceeding optimal bias in class AB, or B as Doug Self defines it. The 2 issues may seem to collide but I go along with his published work as proven science. Distortion In Power Amplifiers - Home Warranty Appliances - See Fig 14
Here, most of us playing with amplifiers at low power levels with no effective measurement capability, may not notice the small amounts of distortion produced by Gm doubling and may be happy to ignore it in a trade-off for a better sounding amplifier overall. That's just speculation, but it's what I opted to do myself on a few projects, some of which I've regretted
GNFB is an acronym for global negative feedback. It seems to only be used in audio design and engineering discussion. Up to a point, the more feedback, the lower the distortion generated by the amplifier. It has nothing in particular to do with class A, AB or topology. The specific case of crossover distortion does not arise in class A so we might expect it to have lower overall distortion and less need for feedback. I hope this clarifies some of what I have been saying.
To answer your question about high bias and Gm doubling, I think we have to distinguish between the much higher bias used for class A and the lower level consistent with exceeding optimal bias in class AB, or B as Doug Self defines it. The 2 issues may seem to collide but I go along with his published work as proven science. Distortion In Power Amplifiers - Home Warranty Appliances - See Fig 14
Here, most of us playing with amplifiers at low power levels with no effective measurement capability, may not notice the small amounts of distortion produced by Gm doubling and may be happy to ignore it in a trade-off for a better sounding amplifier overall. That's just speculation, but it's what I opted to do myself on a few projects, some of which I've regretted
Reason I asked for confirmation is because you stated "class A is the only way to eliminate it (xo distortion) in conventional power amplifiers". Then your statement included CFP. And i'm wondering why class A version of P3A didn't sound better than P3A according to Rod, while Self has highly praised class A with his EF Blameless. I still remember the sound of P3B. It sounded more like class A than like P3A. May be this is just about taste and preference. I myself have never been in love with BJT class A.
Here, most of us playing with amplifiers at low power levels with no effective measurement capability, may not notice the small amounts of distortion produced by Gm doubling and may be happy to ignore it in a trade-off for a better sounding amplifier overall. That's just speculation, but it's what I opted to do myself on a few projects, some of which I've regretted
That was the question. How audible the gm doubling is. Self talked seriously about gm doubling as if he had good ears. He showed the effect of gm doubling, in amp with less than 0.002% thd, which is my assumed audibility threshold. My amp is simulated around 0.003% and i 'think' that i can hear gm doubling effect.
I also asked your opinion regarding output resistor. Self seems to believe that 0.1 is the best bet. This is high output transconductance (Self uses the word 'efficiency') and when there is overbias, the gm doubling effect will only be half than if R is 0.22. Self also uses higher driver current than Rod's P3A. So, where do you stand? Rod is the minority here. There are many implementation of BJT CFP in DIY and commercial amps. Which one do you think is the best one? This is to find out the best implementation of CFP and whether it is sufficiently okay to go all out with it. Self said 'it is hard to see why this topology is not more popular'.
Oh, and Bigun's opinion about class AB is i think against Self.
I think you are attempting to compare amplifiers and designers with very different aims. First, Self is a professional, commercial designer of top notch professional audio gear. He made a name as a designer of first-rate mixing consoles where issues of noise and distortion must be minimised at all cost. In recent times, he advised that he is able to design for max. 0.0005% power amplifier THD which is incredibly good for relatively simple, Blameless type amplifiers. He doesn't consider "sound quality" as a valid aim for amplifier design - it's all about zero distortion = perfect audio for most engineers in fact. All his articles and books refer to measurements and perhaps simulation when theory is discussed - nothing written is assessed by ear unless it's just a personal comment.
If we want to hear nice musical effects, they should be added in a preamp style sound effect box with a "niceness knob" in his words.
Rod Elliott's ESP designs are certainly low distortion but as a priority, they are intended to sound good in their intended home use. The two aims may seem to be the same but in reality, as DIY builds, Blameless designs couldn't sound more different.
You may simulate 0.002% for P3a but Rod measures it at typically 0.04% from 1-80W.
I measured it with a QA400 soundcard at 0.05-0.06% from 1W to 50W. That's in fair agreement and what I would expect for the design, typical parts and assembly on matrix board that I used, which was nowhere near as good as Rod's current PCB. That's at least 20 times the distortion of the Blameless designs.
Does it matter that P3a distortion is higher than possible with ultra-low distortion designs? Absolutely! Sound quality is mostly about Hiraga's preferred spread of distortion products or 'harmonic profile', hence zero distortion translates to virtual zero sound quality, in blunt terms.
I'm not going to answer all queries because Self detailed his findings very well and once again, his aim was to determine where lowest THD occurred. I think his results speak for themselves but others have different aims and will persist with higher THD for various reasons, principally for sound quality but also because DIY traditions and reliability in home applications may be just as important to them.
If we want to hear nice musical effects, they should be added in a preamp style sound effect box with a "niceness knob" in his words.
Rod Elliott's ESP designs are certainly low distortion but as a priority, they are intended to sound good in their intended home use. The two aims may seem to be the same but in reality, as DIY builds, Blameless designs couldn't sound more different.
You may simulate 0.002% for P3a but Rod measures it at typically 0.04% from 1-80W.
I measured it with a QA400 soundcard at 0.05-0.06% from 1W to 50W. That's in fair agreement and what I would expect for the design, typical parts and assembly on matrix board that I used, which was nowhere near as good as Rod's current PCB. That's at least 20 times the distortion of the Blameless designs.
Does it matter that P3a distortion is higher than possible with ultra-low distortion designs? Absolutely! Sound quality is mostly about Hiraga's preferred spread of distortion products or 'harmonic profile', hence zero distortion translates to virtual zero sound quality, in blunt terms.
I'm not going to answer all queries because Self detailed his findings very well and once again, his aim was to determine where lowest THD occurred. I think his results speak for themselves but others have different aims and will persist with higher THD for various reasons, principally for sound quality but also because DIY traditions and reliability in home applications may be just as important to them.
Very good explanation and assessment there, Ian. Thanks. I'm a bit hooked by this CFP topology, as i have been hooked by latfet. I think it is the sound of 'linearity' that i heard and liked. That's why i didn't try hard enough to reduce local FB in the CFP. Regarding 'niceness' versus low distortion, i have many designs that can achieve both, so i can't see a reason to sacrifice any of these attributes. I'm just looking for simple and cheap amps for my three-way speaker.
One thing that i like with this CFP topology is that the bass can be made sooo good, that i have to retire any ideas related to amplifiers with cap at the output such as the NAD 310 (i don't remember if the similar AKSA-Ranchu design has output cap or not, if i had to pursue this topology). And if i later decide not to go with CFP, i can use the amp for my two-inch speaker (I found the amp can be made to sound better than other amps with small speaker).
One thing that i like with this CFP topology is that the bass can be made sooo good, that i have to retire any ideas related to amplifiers with cap at the output such as the NAD 310 (i don't remember if the similar AKSA-Ranchu design has output cap or not, if i had to pursue this topology). And if i later decide not to go with CFP, i can use the amp for my two-inch speaker (I found the amp can be made to sound better than other amps with small speaker).
Re your personal question, johnego; no, I don't believe that anyone can hear simple distortion as low as 0.003%. Controlled tests have proven that time and again. I struggle to hear 0.1% distortion of a 400Hz sinewave but I am not a kid anymore and now my hearing doesn't extend beyond about 12 kHz. Still, I believe we can hear some side effects of low level distortion in music that we can't detect in other forms of audio.
Re. active speakers. Rod's recommendation to use P3a is likely spot-on for small to medium boxes where power requirements for bass speakers aren't close to safe limits. He has a lot of experience with them and I take his personal choice as a wise one. The catch is mounting all the electronics and wiring in a secure but accessible way, either in the boxes or separately.
Re. active speakers. Rod's recommendation to use P3a is likely spot-on for small to medium boxes where power requirements for bass speakers aren't close to safe limits. He has a lot of experience with them and I take his personal choice as a wise one. The catch is mounting all the electronics and wiring in a secure but accessible way, either in the boxes or separately.
Every individual has different taste and distortion preferences. Amplifier can be in A class and still have high harmonic distortion. Some people prefer odd order harmonics over even order harmonics. Some people don't like crossover distortions but to me they are not that offensive.
I remember some 30+ years ago that in HiFi Choice magazine there was test of Audio Innovations tube amp (I think it was Audio Innovations 400 integrated). It was push pull EL84 based meaning that it had huge amounts of all kinds of distortions, both harmonic and crossover. The test results were: Lab - poor, Sound - excellent! Many solid state amps with great specs were unable to achieve Sound - excellent result.
Distortion performance of amp is highly overestimated in it's importance. Doug Self proved that Japanese high speed output transistors have lower distortion than American or European parts. Sakis too prefers the sound of P3A with Japanese high performance output transistors. But my own personal experience, from many DIY projects, is that I like the sound of low speed parts like TIP35/36C, MJL 2119x, etc., better .
I remember some 30+ years ago that in HiFi Choice magazine there was test of Audio Innovations tube amp (I think it was Audio Innovations 400 integrated). It was push pull EL84 based meaning that it had huge amounts of all kinds of distortions, both harmonic and crossover. The test results were: Lab - poor, Sound - excellent! Many solid state amps with great specs were unable to achieve Sound - excellent result.
Distortion performance of amp is highly overestimated in it's importance. Doug Self proved that Japanese high speed output transistors have lower distortion than American or European parts. Sakis too prefers the sound of P3A with Japanese high performance output transistors. But my own personal experience, from many DIY projects, is that I like the sound of low speed parts like TIP35/36C, MJL 2119x, etc., better .
It is low enough that you don't need feedback to lower THD like in class B. GNFB has high contribution to the problematic sound quality of class B amplifiers.
Depends on your expectation. I'm not sure i know how crossover distortion is perceived by ears. But i know that i have never been satisfied with class B amplifiers.
I think i know why. You need to have skill to manage the issues that come with wide bandwidth amplifiers.
Step one:
Ask people in class AB or even class B what is their beloved music ....I bet that they will say mostly typical rock music
Step two:
Verify that these listeners listen from any amplifier except Japanese made.It should /would be either US made or UK made ( or school its the same )
Step three:
See,study, learn and understand the spectral package of harmonics and distortions appeared in class AB and class B amplifier compare and visualize spectral package of harmonics and distortions appeared in music programs that include electric guitars and associated effects like overdrive, gain ,distortion, extra distortion,ultragain with twin superdrive and distortion
and you will see some things in common....There you go champ !
Now you know why some people will say that the best amplifier they ever listened is a Musical Fidelity ....
PS
When you are done enter soulful house music ..... much smoother
Last edited:
I'm not sure if it is the distortion that is perceived, may be there is something else. My common sense agrees with you but i was able to abx two files with added distortion less than that???
I think the problem will be... to redesign the amp for 36v supply (currently 30v). I have simulated that 30v tends to be better than 36v. It is because i have many similar 36v transformers but i think only one transformer with 30v. I need at least two, one for each speaker. I hope it wouldn't be difficult and i could 'bring forward' the best result that i found with 30v supply.
There you go champ !
Now you know why some people will say that the best amplifier they ever listened is a Musical Fidelity ....
I have heard/built only two MF amps. Both are class A.
