Hi everyone,
The thread entitled "What do you think of this schematic" posted by PMM began with the schematic of an Aussie kit amp published in the 80s by AEM magazine. Unfortunately the thread got way off track and didn't explore what I think would be a very interesting subject - hence this thread.
In the designer's words, this amplifier was designed to provide the very best subjective performance, while offering superb objective performance. Hence it was titled the "Ultra-Fidelity" amplifier, project AEM6000 - commonly referred to as the "6000".
The author stated that he believes that the use of fully symmetrical stages in the input of an amp is inappropriate and that an asymmetric stage offers superior subjective and objective performance for this stage. However for the next stage (voltage amplifier), he believes the opposite: that symmetrical VAS stages are superior. This is why the 6000 was designed as such, and I've not seen any other amp with this topology. I think this amp is really exceptionally good sounding, and so does PMM.
The generic "Lin configuration" that Self discusses in his "Distortion in Power Amps" series always sounds similar to me - I have heard good and bad amps that use this topology, but they always have to my ears a rather forward midrange, bright but very detailed treble and hard but not extended bass. Based on Hugh Dean's comments about his AKSA amps, it is a matter of tweaking this circuit and getting the fine details correct to make it sound great - I don't dispute this and haven't heard these amps, but I have built many variations of this topogy and to me it always has similar sound characterisitcs. This is a fully asymmetrical topology.
Then you have fully symmetrical topologies such as the Leach amp, which to my ears sound more correct and civilised, but also less interesting than the Self topology (more laid back mids I suspect are the main reason for this).
The 6000 amplifier seems to offer the best of both worlds subjectively (and objectively according to the designer, David Tillbrook), and I wonder whether it is due to the mixture of symmetric and asymmetric stages.
Does anyone have any experience/comments on this subject?
The thread entitled "What do you think of this schematic" posted by PMM began with the schematic of an Aussie kit amp published in the 80s by AEM magazine. Unfortunately the thread got way off track and didn't explore what I think would be a very interesting subject - hence this thread.
In the designer's words, this amplifier was designed to provide the very best subjective performance, while offering superb objective performance. Hence it was titled the "Ultra-Fidelity" amplifier, project AEM6000 - commonly referred to as the "6000".
The author stated that he believes that the use of fully symmetrical stages in the input of an amp is inappropriate and that an asymmetric stage offers superior subjective and objective performance for this stage. However for the next stage (voltage amplifier), he believes the opposite: that symmetrical VAS stages are superior. This is why the 6000 was designed as such, and I've not seen any other amp with this topology. I think this amp is really exceptionally good sounding, and so does PMM.
The generic "Lin configuration" that Self discusses in his "Distortion in Power Amps" series always sounds similar to me - I have heard good and bad amps that use this topology, but they always have to my ears a rather forward midrange, bright but very detailed treble and hard but not extended bass. Based on Hugh Dean's comments about his AKSA amps, it is a matter of tweaking this circuit and getting the fine details correct to make it sound great - I don't dispute this and haven't heard these amps, but I have built many variations of this topogy and to me it always has similar sound characterisitcs. This is a fully asymmetrical topology.
Then you have fully symmetrical topologies such as the Leach amp, which to my ears sound more correct and civilised, but also less interesting than the Self topology (more laid back mids I suspect are the main reason for this).
The 6000 amplifier seems to offer the best of both worlds subjectively (and objectively according to the designer, David Tillbrook), and I wonder whether it is due to the mixture of symmetric and asymmetric stages.
Does anyone have any experience/comments on this subject?
Hi Owdeo,
I guess I can't add much to this argument. I still hold that the devil is in the details, particularly the stability issues.
I also believe there is something to be gained through close control of the open loop gain. Feedback factor is definitely important; too much gives too much damping which kills decay and thus affects vitality. Too little makes it loose, like a bass amp in the bottom end, and allows a little too much of the voltage amp non-linearities through. It is definitely a balancing act, the management of compromise. In this regard an audio amp is like most other machines.......
I have done very little with fully complementary designs, so can't comment too much. By and large I feel their complexity is not justified by improved sound, and complementary voltage amplifiers seem to me to be inappropriate because the OLGs will not be the same. This mandates a lot of emitter degeneration on each voltage amplifier, which greatly reduces the Zout of the VAS. This is very important since the VAS must cope with wildly varying impedances at the input to the output stage as it cycles from positive to negative half cycles through the crossover point.
Cheers,
Hugh
I guess I can't add much to this argument. I still hold that the devil is in the details, particularly the stability issues.
I also believe there is something to be gained through close control of the open loop gain. Feedback factor is definitely important; too much gives too much damping which kills decay and thus affects vitality. Too little makes it loose, like a bass amp in the bottom end, and allows a little too much of the voltage amp non-linearities through. It is definitely a balancing act, the management of compromise. In this regard an audio amp is like most other machines.......
I have done very little with fully complementary designs, so can't comment too much. By and large I feel their complexity is not justified by improved sound, and complementary voltage amplifiers seem to me to be inappropriate because the OLGs will not be the same. This mandates a lot of emitter degeneration on each voltage amplifier, which greatly reduces the Zout of the VAS. This is very important since the VAS must cope with wildly varying impedances at the input to the output stage as it cycles from positive to negative half cycles through the crossover point.
Cheers,
Hugh
Hugh hasn't designed any(?) symmetrical amps but I have _only_ designed symmetrical and the reason why is simply I like the looks of symmetry. I gather also that distortion is "kinder" and more symmetrical.
In contradiction to Mr Dean, I have no problems in using many parts. I'm pleased with my results. That's enough for me.
In contradiction to Mr Dean, I have no problems in using many parts. I'm pleased with my results. That's enough for me.
Per Anders
I don't quite understand what you mean by "kinder". A symmetrically designed amplifier does reduce second harmonics but instead it introduces more of third harmonics. This is because of the exponential nature of the transfer functions for both MOSFETs and bipolar devices.
Third harmonic is known to sound poorly compared to a similar amount of second harmonic. You can easily see it with your eyes if you sum two waves in pspice. The third harmonic affect the appearance of signal much more than the second harmonic.
I don't quite understand what you mean by "kinder". A symmetrically designed amplifier does reduce second harmonics but instead it introduces more of third harmonics. This is because of the exponential nature of the transfer functions for both MOSFETs and bipolar devices.
Third harmonic is known to sound poorly compared to a similar amount of second harmonic. You can easily see it with your eyes if you sum two waves in pspice. The third harmonic affect the appearance of signal much more than the second harmonic.
Hugh and I have had a little discussion about his AKSA amp. The amp may be a little bit "retro" but Hugh have choosed this (with good justifications) and I haven't heard it so I won't dismiss it unheard.
This amp is rather slow and some people claim that an amp should be rather fast with fairly high slew rate. I belong to the high slew rate wing, surprised?
This could be a really burning topic, is it?
Note: Hugh wants his amp to be slow by purpose, his design choice.
BTW: AKSA, does it mean something?
This amp is rather slow and some people claim that an amp should be rather fast with fairly high slew rate. I belong to the high slew rate wing, surprised?
This could be a really burning topic, is it?
Note: Hugh wants his amp to be slow by purpose, his design choice.
BTW: AKSA, does it mean something?
traderbam
I have actually read in an article by Robert R.Cordell ("Another view of TIM" I think) that a normalised slew rate of 0,02V/us/V is sufficient in order to cover all music examples. This figure was determined after measuring the FFT of a large amount of LPs if I remember correctly. In a 100W/8ohm amp this would mean that about 0,8V/us would be enough.
LPs probably contain much less information at high frequencies than CDs and especially DVDs or SACDs so triple the figure and end up at 0,06V/us/V which will mean that a 200W/8ohm amp will need 3,6V/us.
It would be interesting to analyze a SACD with a FFT
I have actually read in an article by Robert R.Cordell ("Another view of TIM" I think) that a normalised slew rate of 0,02V/us/V is sufficient in order to cover all music examples. This figure was determined after measuring the FFT of a large amount of LPs if I remember correctly. In a 100W/8ohm amp this would mean that about 0,8V/us would be enough.
LPs probably contain much less information at high frequencies than CDs and especially DVDs or SACDs so triple the figure and end up at 0,06V/us/V which will mean that a 200W/8ohm amp will need 3,6V/us.
It would be interesting to analyze a SACD with a FFT
Slew rate is depending of output power. The amp should be far away from slew rate limiting at all audio levels. Why do we need 192 kHz sampling and and slow amps? My theory is that the speaker should be the worst link in the audio chain.traderbam said:
To be honest: not fun to make slow amps! It's a lot more challenging to design fast ones. Have you followed the CFB thread?
I don't think that many designers, DIY'ers really have made a calculation of which figures different parameters should have, for example 0.05% dist, 35 kHz BW etc. I think we want something better regardless of needs.
How many think that MP3 is alright? How many think that CD is OK? How many think that CD isn't any good and want SACD, DVD etc?
The avatar, you have a vivid imagination! Most living stones are green, brown or beige but Lithops optica is the only one which is purple. I have no intention to change it at the moment, not tired of it yet.
peranders
Designing a 300V/us amplifier is like designing a 2000W amplifier,
You will feel pretty confident that the speed is enough in the first case as well as the power in the second case.
As long as those goals does not affect the sound quality I agree that it is more fun to design fast and powerful amplifiers but in most cases those goals will affect the sound quality.
Take for example the fastest amps you can find and you will see that they use current feed back. A current fed back amplifier normally have poor PSRR compared to a voltage fed back amplifier. Sure, you can put a linear regulator on the power rails to the input stage but the second harmonic is also much higher than with a differential stage.
Some might say that the distorsion is low enough anyway but in that case you can say the same about the speed.
Designing a 300V/us amplifier is like designing a 2000W amplifier,
You will feel pretty confident that the speed is enough in the first case as well as the power in the second case.
As long as those goals does not affect the sound quality I agree that it is more fun to design fast and powerful amplifiers but in most cases those goals will affect the sound quality.
Take for example the fastest amps you can find and you will see that they use current feed back. A current fed back amplifier normally have poor PSRR compared to a voltage fed back amplifier. Sure, you can put a linear regulator on the power rails to the input stage but the second harmonic is also much higher than with a differential stage.
Some might say that the distorsion is low enough anyway but in that case you can say the same about the speed.
Patrik, you sound like a boring engineer 
but you are right. "More" isn't allways better. You have somewhere a technical optimum.
You are rather new here so you have maybe missed, putting CD's in the freezer, peeling of covers of caps, using 17 um laminate for multi ampere amps and 10 mil /17 um for signal traces, etc, etc, etc? (The last one is a voodoo tweak from a known member here. )
The world is crazy
but you are right. "More" isn't allways better. You have somewhere a technical optimum.You are rather new here so you have maybe missed, putting CD's in the freezer, peeling of covers of caps, using 17 um laminate for multi ampere amps and 10 mil /17 um for signal traces, etc, etc, etc? (The last one is a voodoo tweak from a known member here.
)The world is crazy
I have a voltage feedback power amp with a slew rate of about 200v/us (the slew rate of the voltage gain section is about twice this value). This is a proof-of-concept prototype - neither the schematic nor the board layout have been optimized, so I expect that with more development effort the performance can be improved.
Given the amount of RF noise that is omnipresent in today's world, an audio amplifier has to deal with much higher bandwidth energy than the music signals themselves.
IME, amplifiers work best when: 1. there is a low-pass filter at the input. 2. the physical size of the circuitry is small, so that the amplifier acts like a smaller and therefore less sensitive antenna for RF. 3. The amplifier has enough bandwidth and slewrate that any HF energy that gets through 1 and 2 can be treated just like any other signal. I find that when the above measures are not done, the amplifier is more prone to IM distortion, and the results of that are sum-and-difference signals that are not related directly to the musical signal at all. When this happens, you may be able to hear the distortion as a separate mask over the music signal, or the amplifier may simply sound nasty.
I have noticed that frequently amplifier manufacturers (including myself) tend to get into trouble when they first demonstrate their prototypes in a different country. I believe that this is because the RF band is allocated differently according to local laws (not to mention hotel rooms with cable transmissions running through the walls on poorly-shielded cable).
hth, jonathan carr
Given the amount of RF noise that is omnipresent in today's world, an audio amplifier has to deal with much higher bandwidth energy than the music signals themselves.
IME, amplifiers work best when: 1. there is a low-pass filter at the input. 2. the physical size of the circuitry is small, so that the amplifier acts like a smaller and therefore less sensitive antenna for RF. 3. The amplifier has enough bandwidth and slewrate that any HF energy that gets through 1 and 2 can be treated just like any other signal. I find that when the above measures are not done, the amplifier is more prone to IM distortion, and the results of that are sum-and-difference signals that are not related directly to the musical signal at all. When this happens, you may be able to hear the distortion as a separate mask over the music signal, or the amplifier may simply sound nasty.
I have noticed that frequently amplifier manufacturers (including myself) tend to get into trouble when they first demonstrate their prototypes in a different country. I believe that this is because the RF band is allocated differently according to local laws (not to mention hotel rooms with cable transmissions running through the walls on poorly-shielded cable).
hth, jonathan carr
"By and large I feel their complexity is not justified by improved sound, and complementary voltage amplifiers seem to me to be inappropriate because the OLGs will not be the same. "
So deliberately choosing a 'slow' cap for the bootstrap current source makes the OLG the same for positive and negative going signals?
Right.
So deliberately choosing a 'slow' cap for the bootstrap current source makes the OLG the same for positive and negative going signals?
Right.
jcarr
Not quite sure I understand. The best way to make sure that HF-disturbances aren't interfering with the audio signal would be to limit the bandwidth of the gainstage, wouldn't it?
peranders
I don't know if you have seen these wonderful little resistors called MINIMELF? They are packaged in a cylindrical case and they sound great because of their low noise and high bandwidth. You can buy them i Farnell and they don't cost more than regular thickfilm resistors. A bit harder to solder though
Not quite sure I understand. The best way to make sure that HF-disturbances aren't interfering with the audio signal would be to limit the bandwidth of the gainstage, wouldn't it?
peranders
I don't know if you have seen these wonderful little resistors called MINIMELF? They are packaged in a cylindrical case and they sound great because of their low noise and high bandwidth. You can buy them i Farnell and they don't cost more than regular thickfilm resistors. A bit harder to solder though
Pabo wrote
Regarding RF I fully agree with Jonathan Carr:
It is definitely better when your amp is able to handle the RF at it's input properly rather than trying to get araound the problem by making any amp stages sluggish.
Since a passive lowpass is always more linear (given the right choice of components and layout) and always better able to handle RF signals than any active circuitry, the input filter is a good aproach for me.
I don't remember ever having built or owned any power amp without one !!!!!
Who ever heard a broadcasting station through his audio equipment knows what the the effect of improper handling of RF signals means (rectification/demodulation within active stages). Even if you dont' hear them distinctively doesn't mean that they are not there and don't impair the sound.
Regards
Charles
Regarding RF I fully agree with Jonathan Carr:
It is definitely better when your amp is able to handle the RF at it's input properly rather than trying to get araound the problem by making any amp stages sluggish.
Since a passive lowpass is always more linear (given the right choice of components and layout) and always better able to handle RF signals than any active circuitry, the input filter is a good aproach for me.
I don't remember ever having built or owned any power amp without one !!!!!
Who ever heard a broadcasting station through his audio equipment knows what the the effect of improper handling of RF signals means (rectification/demodulation within active stages). Even if you dont' hear them distinctively doesn't mean that they are not there and don't impair the sound.
Regards
Charles
There was an app note by AD on this. A lot of RFI rectification seems to take place in the input transistors, where it produces an error voltage. Running the LTP at higher current makes it less susceptible. High speed bipolar and FETs were less susceptible than low speed, precision devices.
Limiting the VAS bandwidth is probably not a good idea, as it will allow the LTP to go into saturation on a RFI transient. And input filter to throw away the unneeded bandwidth before the active circuit is always a good idea.
Regards,
Eric
Limiting the VAS bandwidth is probably not a good idea, as it will allow the LTP to go into saturation on a RFI transient. And input filter to throw away the unneeded bandwidth before the active circuit is always a good idea.
Regards,
Eric
janneman
I state that you are wrong.
Simply draw the transfer functions of a PNP and a NPN and then combine these curves.
You end up with something that looks like a third order function.
The output voltage of a gain stage is equal to the input voltage multiplied with the transfer function so a third order transfer will introduce or increase the third harmonic.
If you don't believe me, simply simulate!
I state that you are wrong.
Simply draw the transfer functions of a PNP and a NPN and then combine these curves.
You end up with something that looks like a third order function.
The output voltage of a gain stage is equal to the input voltage multiplied with the transfer function so a third order transfer will introduce or increase the third harmonic.
If you don't believe me, simply simulate!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.