What to look for in a power amp?

[Gregadd]

Don-of course one should look for low output impedance. But the Audiholics article suggests that is overrated. That is where the big bucks come in. Trying to design the perfect output transformer or eliminating it all together. Morever a lot of the problems with boomy bass is just plain old inertia.

Your other point indicates why you must reagrd speaker/ amp as a sysytem.



Myles why don't we save NFB until we get to distortion.

 

[DonH50]

I haven't read the articles yet, sorry. Yes, it is a system, and for some speakers low amp output impedance isn't a big deal (my Maggies come to mind). Inertia is certainly part of it, but rather than get carried away too close to bedtime, I suspect we are thinking the same thing and the rest is semantics.

Regarding Myles' comment, NFB comes into play here not because of its distortion-reducing properties, but because it also reduces the effective output impedance (and thus raises the damping factor) by the feedback factor. I'm pretty sure that's what Myles was thinking (yes?)

Weren't Futterman amps OTL? I heard them but rarely and didn't really care for them that much -- they seemed to have lost some of that tube "magic". They were also reputed to have some reliability issues, though that may have been overstated. I built a tube preamp that was differential and had direct-coupled outputs; again, even with relatively low NFB, the result was a more "SS" sound though it had headroom no SS preamp of the time could touch. I sold it for a pretty penny... Bought my books and a bit mor for a semester!

Interesting discussion! - Don

 

[Gregadd]

There are three different approaches

VTL known for their excellent bass, They use some of the best transformers. www.vtl.com Luke Manley ushered in the whole era of high powered tube amps. he showed that tube amps could have good bass.
Moscode-No transformer using mosfets for the output stage. www.moscode.com George gives a good non-technical discussion about the problems with transformers
Atma-Sphere- True OTL design. www.atma-sphere.com Ralph seems to know everything about amps.
See there website for details.

All can and do work.

 

[MylesBAstor]

I haven't read the articles yet, sorry. Yes, it is a system, and for some speakers low amp output impedance isn't a big deal (my Maggies come to mind). Inertia is certainly part of it, but rather than get carried away too close to bedtime, I suspect we are thinking the same thing and the rest is semantics.

Regarding Myles' comment, NFB comes into play here not because of its distortion-reducing properties, but because it also reduces the effective output impedance (and thus raises the damping factor) by the feedback factor. I'm pretty sure that's what Myles was thinking (yes?)

Weren't Futterman amps OTL? I heard them but rarely and didn't really care for them that much -- they seemed to have lost some of that tube "magic". They were also reputed to have some reliability issues, though that may have been overstated. I built a tube preamp that was differential and had direct-coupled outputs; again, even with relatively low NFB, the result was a more "SS" sound though it had headroom no SS preamp of the time could touch. I sold it for a pretty penny... Bought my books and a bit mor for a semester!

Interesting discussion! - Don

Yes. As Ralph pointed out to me on Audiogon a while back in reference to the MBL speakers:

"the mbl has an impedance peak of about 8-9 ohms in the midrange driver. The designer is expecting the amp to reduce power by 3 db through this range. An amplifier with feedback will do it, one without will not. A good number of tube amps tend to sound shrill on this speaker even though otherwise they have plenty of power......
The problem here is that IMO, amps with feedback sound somewhat shrill out of the box, IOW its my opinion that a speaker that requires this will never sound like real music.

IMO the issue with feedback boils down to open loop propagation delay in the amplifier- IOW its a timing issue. The feedback signal simply does not arrive back at the input in time to make the correction. With a steady-state signal, the amp locks in pretty well over a few iterations, but with a constantly-changing waveform the amp will be chaotic. This is an interesting subject and I agree- a topic for another thread."

In reference to TIM, Ralph also pointed out, "obviously amps designed to overcome that 'issue' were horrendous."

 

[Gregadd]

Again its been a long time. I Transient intermodulation disitortion was a hot topic. This article talks about its relationship to the use of excessive amounts of NFB.
http://forums.klipsch.com/forums/storage/3/1107724/01162904.pdf

 

[DonH50]

@Myles: Use "excessive" feedback and I'd agree. Virtually all amps have a little feedback, and in fact even open-loop designs end up with a little local FB whether we want it or not.

TIM, yeah... The first approaches were to (a) eliminate virtually all NFB and/or (b) make extremely wide band amps. Neither approach was a proper solution sonically, as we all found out!

@Gregadd: Good summary of types. I have a Counterpoint in its box in the basement, another hybrid design (tube in MOSFET out). I thought the Krell and Levinson amps of the day were much "tighter", but the Counterpoint provide better/deeper/whatever bass than my tube amp, and when I bi-amped it provided a solid low end while still blending with my tube upper amp.

Neat stuff! - Don

 

[Gregadd]

Let's look at some numbers. i hope to post some pulse response and square wave response. Don- I know you have some good graphs.

Specs for the Atyma-Sphere MA MkII
220 watts/channel into 4, 8 or 16 ohm load before clipping.
Power bandwidth: 2 Hz - 85 kHz within 1/2 db (open loop)
Frequency response (1 watt, open loop): l Hz - 100 kHz within 3 dB
IM Distortion: less than .05% at full power
THD: less than 0.5%
Square Wave Tilt: unmeasurable
Phase shift: less than 1 degree @ 20 kHz
Power supply current: 80 Amps
Input impedance: 100K single-ended, 200K balanced- selectable to 600 ohms
Output section risetime: 600 V/micro-second
Output Impedance: ~1.75 Ohms (open loop)
Power Supply Storage: 1200 Joules in output section, 100 Joules in Driver supply.
Power consumption: 800 watts/chassis
Tube complement: 20 6AS7 output tubes and 6 6SN7 driver tubes.
Dimensions: 28" l x 17" w x 10" h
Weight: 103 lbs./chassis

This is Class A,Triode OTL, wide band,fast rise time, etc. You may disagree with Ralph but he is pushing the envelope.

 

[Gregadd]

transistors:
Amplifier Classes:
Most mobile amplifiers use complementary transistor pairs to drive the speakers. In this configuration there is a transistor (or group of transistors) which conducts current from the positive power supply voltage for the positive half of the audio waveform and a different transistor (or group of transistors) which conducts current from the negative power supply voltage for the negative half of the waveform. There are some amplifiers which use the same transistor(s) to drive both the positive and the negative halves of the waveform.

NOTE:Amplifiers in classes A, B, and AB operate their output transistors in a 'linear' mode. Class 'D' amplifiers operate their outputs in 'switch' mode.

Mode examples:

Linear mode:
Imagine that you are the amplifier's output device(s) and you must support a 10 pound iron weight (the speaker load). The most difficult method (linear mode) would be to hold the weight straight out in front of you. This would very roughly simulate the linear mode architecture. Your muscles would start to ache in a short amount of time. Think of this pain as the power dissipation in output transistors.

Switch mode:
In this example, you can support the weight in one of two positions. In the first position, you can hold the iron weight directly over your head with your elbows locked so that you're not really using very much effort to support the weight. In the second position, you would let the weight hang down by your side. This would also use very little effort from your muscles. If you held it directly over your head half of the time and by your side for the other half of the time, it's position would 'average' out to be the same as if you held it out straight in front of you like in the previous (linear mode) example. This would roughly simulate the switch mode which we will discuss later in this page. You can see that with this method (switch mode), there would also be little pain (power dissipation) involved in supporting the weight.
--------------------------------------------------------------------------------


CLASS 'A'
Many class A amplifiers use the same transistor(s) to reproduce both the top and bottom halves of the audio waveform. In this configuration, the output transistor(s) always has current flowing through it, even if it has no audio signal (the output transistors never 'turn off'). The current flowing through it is D.C. A pure class 'A' amplifier is very inefficient and generally runs very hot even when there is no audio output. The current flowing through the output transistor(s) (with no audio signal) may be as much as the current which will be driven through the speaker load at FULL audio output power. Many people believe class 'A' amps to sound better than other configurations (and this may have been true at some point in time) but a well designed amplifier won't have any 'sound' and even the most critical 'ear' would be hard-pressed to tell one design from another.
NOTE: Some class A amplifiers use complimentary (separate transistors for positive and negative halves of the waveform) transistors for their output stage.


CLASS 'B'
A class 'B' amplifier uses two transistors (or two groups of transistors). One transistor (or group of transistors) is used to reproduce the top half of the waveform. A second transistor (or group of transistors) is used to reproduce the bottom half of the waveform. In a class 'B' amplifier, there is typically no idle/bias current flowing through the output transistors when there is no audio. In most cases, if the amplifier has no bias potentiometers and it's not a class D amplifier, it's a class 'B' amplifier.

CLASS 'AB'
Class 'AB' amplifiers use two groups of transistors like class 'B' amplifiers. In most respects, class 'AB' and class 'B' amplifiers are very similar. As we said earlier, a class 'A' amplifier is very inefficient. This is not good for a car audio amplifier. Some people believe that class 'B' amplifier can never produce clean audio because their output transistors aren't biased 'on'. A class 'AB' amplifier is generally considered to be the best compromise. A class 'AB' amplifier is a class 'B' amplifier which has a small amount of 'bias' current flowing through the output transistors at all times. This eliminates virtually all of the crossover distortion that's possible with class 'B' amplifiers. The bias current is flowing because the output transistors are always conducting current (even without an audio signal). This differs from a pure class 'A' amplifier in the amount of current flow. A pure class 'A' amplifier typically has an enormous amount of current flowing through its output transistors with NO audio signal. A pure class 'B' amplifier has NO current flowing through its outputs with no audio signal. A class 'AB' amplifier is much more efficient than the class 'A' but without the possible distortion of the class 'B'. MANY of the car audio amplifiers which claim to be a class 'A' amplifier are just a high bias class 'AB' design. These amplifiers are only class 'A' at very low power output levels. At higher power levels, one of the output transistors will switch off while the other output transistor is conducting. I don't want you to think that I am telling you that there are no class 'A' amplifiers. There are a few high quality mobile amplifiers which are a true class 'A' design.

CLASS 'D'
We said that class 'A' amplifiers were VERY inefficient. Class 'AB' amplifiers are also inefficient but are more more efficient than class 'A' amplifiers. Class 'AB' mobile amplifiers are generally 60% efficient when driving a 4 ohm load at maximum power (just before clipping). The reason that these amplifier configurations are inefficient is because there is a difference of potential (voltage) across the output transistors and current flowing through the output transistors. When you have voltage across the device and current flow through the device, there will be power dissipation in the form of heat. The Ohm's Law formula P=I*E expresses this clearly. The power needed to produce this heat is wasted power. When there is (virtually) no voltage drop across a device (whether it's a large piece of wire or a transistor), there can be a significant amount of CURRENT flow through the device with (virtually) no power dissipation. This means that there is virtually no heat given off (highly efficient). The inverse is also true. If you have a significant amount of VOLTAGE across the device (transistor, wire...) but no current flow through the device, again, there will be no wasted power. If you look at the formula P=I*E again, you can clearly see that if you reduce either I or E (I is current, E is voltage) to a value near 0, the power dissipation will be very low.
OK, now to the point. A class 'D' amplifier, which may also be known as a switching amplifier or a digital amplifier, utilizes output transistors which are either completely turned on or completely turned off (they're operating in switch mode). This means that when the transistors are conducting (switched on) there is virtually no voltage across the transistor and when there is a significant voltage across the transistor (switched off), there is no current flowing through the transistor. This is very similar to the operation of a switching power supply which is very efficient...
http://www.bcae1.com/ampclass.htm

 

[Steve Williams]

Greg

I appreciate the time you took to get this together for everyone

 

[Orb]

@Myles: Use "excessive" feedback and I'd agree. Virtually all amps have a little feedback, and in fact even open-loop designs end up with a little local FB whether we want it or not.

TIM, yeah... The first approaches were to (a) eliminate virtually all NFB and/or (b) make extremely wide band amps. Neither approach was a proper solution sonically, as we all found out!

@Gregadd: Good summary of types. I have a Counterpoint in its box in the basement, another hybrid design (tube in MOSFET out). I thought the Krell and Levinson amps of the day were much "tighter", but the Counterpoint provide better/deeper/whatever bass than my tube amp, and when I bi-amped it provided a solid low end while still blending with my tube upper amp.

Neat stuff! - Don

What I find interesting about NFB is how it can polarise even the best engineers in the audio industry; between those who prefer none or as little as possible to those who present it should be high to be good.
The Nelson Pass article gave us an indication on his feelings and studies in the other thread.
Just going to post a very high level (more for us none audio engineers here but provides a little insight into NFB when also considered with the Nelson link):
Its multiple chapters that are not together so need to use next; Unhappy about Negative Feedback?, and Negative Feedback Guidelines (2), and some other topics not on NFB but of interest and the contents of the presentation is on around the 5th next; including digital/EMI-Class D/filters/etc.
http://hypex.nl/docs/Bruno Masterclass/slides.htm
One caveat though, this is a presentation so its much harder to follow as we do not have all the notes and cannot be read or followed as well as say a paper/article.

Myles and Don; whats your take on the implementation of NFB in the various products released, because it seems to me that not only do we have zero feedback or high feedback (as both nelson and Bruno advise if going this route), but quite possibly those that are in inbetween?
I appreciate power amps make this more complicated for damping/distortion but was thinking more in general as well, and if you feel that this can affect the sound quality if deviating from the very low or high feedback design topologies.

Thanks
Orb

 

[Orb]

Don-of course one should look for low output impedance. But the Audiholics article suggests that is overrated. That is where the big bucks come in. Trying to design the perfect output transformer or eliminating it all together. Morever a lot of the problems with boomy bass is just plain old inertia.

Your other point indicates why you must reagrd speaker/ amp as a sysytem.



Myles why don't we save NFB until we get to distortion.

Interesting point on the output impedance that got me to look at some measurements out there from some review publications.
Interestingly it does seem if the output impedance is relatively low (0.3 that can also include some tube amps) and linear in behaviour then with a simulated real load it looks to be inaudible in terms of difference from those that are around 0.1ohm, by inaudible I am going by being under 0.25db as a guideline.
The simulated load shows a jump when its around 0.5ohm output impedance.
So in general it could be argued for most SS and a few tube designs it may be overrated, but for the others such as most tube amps and a still a few SS it would still be very applicable.
A big BUT though and this is quite interesting, the Rogue tube amp while having same output impedance into its ultralinear and triode taps shows a noticable and quite possibly audible difference.
Unfortunately the triode tap was 8ohm while the ultralinear was 4ohms.
I may miss it but I do not see an explanation why JA went with 8ohm (most stable/best results??)
Anyway this suggests the key is for a low-linear output impedance for it to be considered overrated - IMO but I am no audio engineer

Ideally then it would make sense to find more tube amps that can switch to ultralinear and see what their performance behaviour is.
Anyway as a simple comparison (if you got others to add that may change this perspective thats cool and would be interesting) here is the Rogue and Audio Research tube amps:
http://stereophile.com/tubepoweramps/rogue_audio_m-180_monoblock_power_amplifier/index6.html
http://stereophile.com/tubepoweramps/807ar/index4.html

I may try to find more measurements from other products to see if there is a correlation between the output impedance, simulated load, and changing to ultralinear.

Cheers
Orb

 

[Gregadd]

Greg

I appreciate the time you took to get this together for everyone

And thanks to Don for keeping me technically correct.

 

[Gregadd]

Audio Power Amplifier Fundamentals- http://www.rocketroberts.com/techart/amp.htm

 

[microstrip]

Interestingly it does seem if the output impedance is relatively low (0.3 that can also include some tube amps) and linear in behaviour then with a simulated real load it looks to be inaudible in terms of difference from those that are around 0.1ohm, by inaudible I am going by being under 0.25db as a guideline.

I would like to add some lines on my experience. I own VTL MB750’s and Soundlab A1 px. The Soundlab’s have one a the worst impedance curves I have ever seen – about 50 ohm at 60 Hz , 5 ohm at 600 Hz and 2 ohm at 20 kHz. The curve looks like a magic mountain; phase goes from almost 90 to -90 degrees.
Happily the speaker has adjustable bass, medium and treble controls. Although the first two are in steps, treble is continuous. Every time I change amplifiers, I must adjust the treble level using a spectrometer built with a EMU 24bit 192 kHz USB Tracker pre and a ECM8000 microphone calibrated against B & K reference.
The other amplifier I use is a Classe CA201, especially in hot days in summer .
The response is almost flat for the VTL with minimum attenuation, but as expected for the Classe I have to attenuate treble by almost 4 dB.
Now the “magic” part.
(Disclaimer – these observations were made in a friendly environment, with a few audiophiles and music lovers, in a non scientific non controlled way, just by curiosity and for fun.)
Most of my friends found difficult to notice variations of up to 2 dB at 10 kHz in treble. However, changing the XLR cables in the output of the CD from Nordost Valhalla to Sunyata Antares changed the sound in a way everyone could notice. And, surely, the frequency response did not change when I changed the cables …
Even after equalizing the responses of the system with the VTL and Classe, the sound of the two amplifiers is quite different - and everyone prefers the tubes.

 

[Gregadd]

Not only does NFB have negative effects but our major problem was that it was used to mask a poor design. Intuitively we tend to think if some is good more must be better. If less is good,none must be perfect. This makes good ad copy. NFB bad. We don't use any. How about we start with a proper amp design and use judicious amounts of NFB. All out designs may elect to forgo NFB.

 

[Gregadd]

Tthis article is copyrighted. So I'll just provide a link.
What is negative feed back?
http://www.aikenamps.com/NegativeFeedback.htm

 

[Orb]

Heya microstrip and thats definitely interesting,
this brings up something I was going to ask Don/Amir in the measurements thread.
From what I understand it is possible for CD/transports to transmit different levels of ultrasonic to the preamp/intregrated.
I mention this because there is a British company (Chaper Audio) who make a CD player with whats called "DAC Modulator Mode", which enables you to trade S/N achieved in the audioband against unwanted ultrasonic noise.
So the higher the quantisation "width" the better the audio S/N, but resulting in more ultrasonic noise transmitted into preamp/integrated.

Just wonder how much of a possibility ultrasonic noise generated by a source can have on either sound quality or the down stream performance behaviour of the connected devices (preamp/poweramp).

Cheers
Orb

 

[microstrip]

Just wonder how much of a possibility ultrasonic noise generated by a source can have on either sound quality or the down stream performance behaviour of the connected devices (preamp/poweramp).

Cheers
Orb

Good point. I do not remember any more who wrote about that, and even more: in an amplifier using global NFB, the output of the amplifier is partially re-injected in the input. This way the speakers and speaker cables can act as an antenna and inject RF noise in the amplifier.
Remember the radio long wave band starts around 150 kHz and the radio clock signals are transmitted between 40 and 80 kHz.

 

[DonH50]

Thanks for the kind words!

Random thoughts...

NFB has a lot of positive effects, too. I would not base my choice of an amplifier on whether it does, or does not, have negative feedback. Every "open-loop" design I have seen does have a little bit of local NFB. IMO, the best-sounding amps I have heard start with a sound design (no pun intended), then add enough feedback to get that last little bit: lower distortion, lower and more controlled output impedance, stable bias over supply and temperature variation, etc.

Significant ultrasonic noise, or even worse borderline oscillation, is BAD! It can cause all sorts of problems in the system, from adding audio trash via rectification and mixing in various low-level stages, to frying tweeters. I had not heard about a transport doing it, but the clock could certainly cause problems if not well-filtered. I'd be a bit surprised if I saw high ultrasonic levels from a modern design. I have not measured any recently, however.

Just to throw it out, there is (also) feed-forward compensation in some designs, especially some current-mode amplifier designs.

I was listening with a friend of mine and could instantly tell the difference between a tube amp and SS, and after a short time, he could too!


FWIWFM - Don

 

[Gregadd]

Distortion- EE Times
Distortion in power amps- http://www.eetimes.com/design/audio...Part-I-the-sources-of-distortion?pageNumber=1