Its a more complex subject then you suggest - but the foundation of any amplifier is its PSU.
Presuming a "flat" frequency response amplifier there are some important parameters of an amplifier design that effect the perceived Bass performance.
1. Open loop output stage impedance (this is the OPS inherent output impedance before the application of Globe feedback).
2. Power Supply Rejection Ratio - its often erroneously assumed that PSRR will be equal or greater then the amplifiers FB gain but this is not the case as PSU noise (Modulation) is often injected as Common mode and thus Invisible to the Feedback loop.
3. Open loop gain the more gain the lower the Closed loop output impedance
The Aleph30 and to a lesser degree the JLH 20W designs are poor in the above parameters making them very sensitive to there PSU performance this does not need to be the case.
Sonically the JLH20W is still a landmark design game changing for its time (WRT Transistors amplifiers of the day), but that was over 40 years ago - sadly few designs have advanced the art since (with a few notable exceptions such as the Quad Current Dumping)...
Its a more complex subject then you suggest - but the foundation of any amplifier is its PSU.
Presuming a "flat" frequency response amplifier there are some important parameters of an amplifier design that effect the perceived Bass performance.
1. Open loop output stage impedance (this is the OPS inherent output impedance before the application of Globe feedback).
2. Power Supply Rejection Ratio - its often erroneously assumed that PSRR will be equal or greater then the amplifiers FB gain but this is not the case as PSU noise (Modulation) is often injected as Common mode and thus Invisible to the Feedback loop.
3. Open loop gain the more gain the lower the Closed loop output impedance
The Aleph30 and to a lesser degree the JLH 20W designs are poor in the above parameters making them very sensitive to there PSU performance this does not need to be the case.
Sonically the JLH20W is still a landmark design game changing for its time (WRT Transistors amplifiers of the day), but that was over 40 years ago - sadly few designs have advanced the art since (with a few notable exceptions such as the Quad Current Dumping)...
Ah, that'll be the design flaw of the Musical Fidelity A-100 then. The fan-cooling had sucked household dust inside the hot bits, thereby progressively insulating them until they blew up from self-combustion. Clearly, there was no thermal switching ...Fan cooling appears to be the only practical means of realising class A amps of reasonable output without them becoming too massive or heavy in my experience. Noise issues can be dealt with by using oversize fans operating at a fairly low speed. The safety aspect needs thermal switches on the heatsink that will switch off the power if the fans fail and the temperature rises to dangerous levels.
1. Open loop output stage impedance (this is the OPS inherent output impedance before the application of Globe feedback).
IMaybe it's because they have higher o/p impedance and Robert is right!
Although it may seem counter-intuitive, in many applications (not in zero feedback designs obviously) your point 1 is not correct... I was quite surprised at the proof myself! It was shown by Prof. E. Cherry that in the case of both common collector (emitter follower) and common emitter output stages, that all being equal up to that point, when the feedback loop is closed the extra open loop gain of the CE OPS reduces the closed loop output impedance to exactly match that of the EF OPS.....
also from my experience I'd say they can indeed lack ultimate bottom end "thump" - but have always found there midband to be very open and a very spacious sound stage.
I've always assumed the lack of very Tight Bass to be due to the higher Open Loop output impedance in the Mosfet output stage which can be 10dB to 20dB worst then an equivalent Bipolar OPS.
Interesting, I have never thought my MOSFET amps lacked bass but they use an error correcting technique described by Hawksford, I am sure you have seen it but here it is anyway for any one else who is interested.
http://www.essex.ac.uk/csee/research/audio_lab/malcolmspubdocs/J3 Distortion correction PA.pdf
Bob Cordell did an application note for Siliconix, I think!, that used this technique.
Well indeed - the OPS EC loop goes a very long way to resolving the above problems... well trimmed and your looking at atleast 40dB error cancellation...
I guess this is where we will have to differ, having to rely on FB to reduce output impedance does not result in the sonic performance as an output stage with inherently the same output impedance as would be obtained by the use of global feedback.
Effects that come into play is the reverse EMF from the speaker driver which does not see a true Zero ohm source unheeded by the OPS poor impedance finds it way into the input stage of the amplifier thus upsetting the amplifier loop with an erroneous error signal (the phase of this back EMC is anyone's guess) - this is especially troublesome around the Class A/B crossover region where the output stage experiences severe impedance variations - the output stage is effectively Open loop (OK the loop gain is all over the shop) and feedback is at best struggling to act transversing this region results in Phase error which is dependent (amongst other factors) on the ratio of the OPS Real output impedance and the load impedance. This is a classic example where global FB is not a cure for all ills.
Being a phase error - individual frequency components within a complex musical signal are affected to a greater or lesser extent the PM having a greater effect on higher frequency components. As this Phase modulation effect is "non linear" across the audio BW phase coherence is lost during this critical part of the musical signal (Sound stage and "resolution" information is contained within the Low level signal region where output stages typically transverse rapid impedance variations).
Pure ClassA, Non Switching Class B designs (and other Super Class A derivates) do not suffer this effect - its also why its preferable to bias ClassA/B MOSFET designs (which do not suffer from GM doubling) heavily into ClassA operation in an effort to move this PM / impedance transition beyond the critical low level information region.
Over the past 1.5 year I've worked very heavily in developing the topology of our little MIMP amplifier which uses a Bootstrapped ClassA design - the Spice simulations clearly show the effects of back EMF and PM "amplified" by poor real output impedance (and the rapid impedance variations during the crossover region). Not only do these impedance variations open the Gate to back EMF reaching the input stage, it also allows the speaker cables conducted RF unheeded access to the input devices which if bipolar junctions are only to happy to demodulate this RF energy with wild abundance
Speaking for myself and my own design ethos, inherently low OPS impedance is critical on many levels Global feedback should be considered a friend whos helping hand should only be called upon in rare times of need
Pure ClassA, Non Switching Class B designs (and other Super Class A derivates) do not suffer this effect - its also why its preferable to bias ClassA/B MOSFET designs (which do not suffer from GM doubling) heavily into ClassA operation in an effort to move this PM / impedance transition beyond the critical low level information region.
Ah, that'll be the design flaw of the Musical Fidelity A-100 then. The fan-cooling had sucked household dust inside the hot bits, thereby progressively insulating them until they blew up from self-combustion. Clearly, there was no thermal switching ...
John. whilst you raise some valid issues (and a few tenuous ones), it's gone rather off topic to what you first said and to my reply! My point was merely that a high impedance OPS can have just as low an output z once the NFB loop is closed....
Does it really need that much bias to avoid the issues you mention? I don't know much about amp design, but I'd have imagined so long as you are far enough away from the zero-crossing point for distortion to remain low, the NF must be working enough to keep output Z low too?
I think what John is referring to is If the output device has not turned on, then the forward gain is zero and its a high Z output. That's why non switching, error correction and class A are preferred.
I think what John is referring to is If the output device has not turned on, then the forward gain is zero and its a high Z output. That's why non switching, error correction and class A are preferred.
I stand corrected. My A-100 died a natural heat-related death within its first few years. The repair carried out by a tech made it sound noticeably inferior, so I moved it on.Actually they did have a thermal switch. In spite of this fail-safe which would operate at about 90C, their normal operating temperature combined with standard 85C rated caps tended to eventually kill them off.... the A1 is just the same. I regularly carry out rebuilds on these models to bring them back to better than new electrical condition and the problems I find are usually heat related!