I've just taken to looking more closely at the Stereophile amplifier measurement pages, by JA, and the distortion graphs can be very illuminating in terms of predicting how the amps will perform subjectively.
There are 4 things I'm now looking at:
* THD at precisely 1W into 8 ohms: this varies from -70 to nearly -100dB for pretty decent to excellent, SS, units
* How precisely linear the distortion vs. power curve is: less premium units have a decent upward bulge midway along, crossover distortion issues. The hyped Halcro looks rather wonky once real power levels are hit; in comparison the Soulution unit is beautifully straight right to clipping
* The flatness of the THD vs. frequency curve: the "engineered" units are ruler flat; others gracefully start to climb in distortion heading towards 20kHz
* And finally, how THD varies with load, especially the 2 ohm setting
Combine those 4 factors, and it's quite straightforward to connect the subjective reactions of people exposed to the units with the objective measurements. So, for example, the Halcro unit was not liked by many people and to me it's very clear from those graphs why this was so ...
Frank
There are 4 things I'm now looking at:
* THD at precisely 1W into 8 ohms: this varies from -70 to nearly -100dB for pretty decent to excellent, SS, units
* How precisely linear the distortion vs. power curve is: less premium units have a decent upward bulge midway along, crossover distortion issues. The hyped Halcro looks rather wonky once real power levels are hit; in comparison the Soulution unit is beautifully straight right to clipping
* The flatness of the THD vs. frequency curve: the "engineered" units are ruler flat; others gracefully start to climb in distortion heading towards 20kHz
* And finally, how THD varies with load, especially the 2 ohm setting
Combine those 4 factors, and it's quite straightforward to connect the subjective reactions of people exposed to the units with the objective measurements. So, for example, the Halcro unit was not liked by many people and to me it's very clear from those graphs why this was so ...
Frank
Decent power levels, 20 to 80 watts. As an example of an "engineered" THD performance there's Bryston: Bryston 28B-SST monoblock power amplifier Measurements | Stereophile.com. And for superbly linear to clipping, Soulution: Soulution 710 power amplifier Measurements | Stereophile.com. Finally, an all-out assault on optimum performance, MBL: MBL Reference 9011 monoblock amplifier Measurements | Stereophile.com.
Krell is a bit of an also-ran in this crowd: Krell Full Power Balanced 350mc monoblock amplifier Measurements part 2 | Stereophile.com ...
Frank
Krell is a bit of an also-ran in this crowd: Krell Full Power Balanced 350mc monoblock amplifier Measurements part 2 | Stereophile.com ...
Frank
Indeed for optimum linearity, but people paralleling many devices may have a different agenda, like thermal stability, accuracy of sharing, tolerance to mismatch, etc
Very smart: changing a problem into a feature, I love it.
I personally couldn't rely on chance alone, Murphy would make that impossible, but a nice hand-picked spread could certainly help smooth out and straighten the transfer curve.
Here is class AC for comparison: it offers a no-compromise wide class A region, but without the penalty of an increased THD when that range is exceeded.
It strikes me that it can be likened to Edmond's idea of statistical spreading, but discretized down to two elements.
The same principle could be applied to a larger number of devices pairs of suitable Is, and fitted with corresponding series resistors.
This example uses schottky diodes to minimize the class A quiescent (even with this trick, it is still 0.7A, quite a "hot" class).
It strikes me that it can be likened to Edmond's idea of statistical spreading, but discretized down to two elements.
The same principle could be applied to a larger number of devices pairs of suitable Is, and fitted with corresponding series resistors.
This example uses schottky diodes to minimize the class A quiescent (even with this trick, it is still 0.7A, quite a "hot" class).
Attachments
Hi Guys
Self made a comment regarding keeping the idle current the same and having more devices in parallel inasmuch as the current variations through each device are reduced, thus reducing the distortion contributions individually and in total. There is no specific mention of increasing Re but that would be implied.
His conclusion is that crossover-related distortions are greatly reduced, especially when sustained-beta devices are used, making the method of simply adding output pairs one which potentially allows these specific distortions to be reduced to arbitrarily low levels. This seems to contradict Leech's assertion that THD will stay the same by adding pairs but keeping Iq constant.
Of course, the point of class-i is to allow a similar reduction of crossover artifacts without resorting to infinite expansion of the output stage and its heat sink.
Elvee, other than as a design exercise, why build 50W if you only need 1W or so?
Have fun
Kevin O'Connor
Self made a comment regarding keeping the idle current the same and having more devices in parallel inasmuch as the current variations through each device are reduced, thus reducing the distortion contributions individually and in total. There is no specific mention of increasing Re but that would be implied.
His conclusion is that crossover-related distortions are greatly reduced, especially when sustained-beta devices are used, making the method of simply adding output pairs one which potentially allows these specific distortions to be reduced to arbitrarily low levels. This seems to contradict Leech's assertion that THD will stay the same by adding pairs but keeping Iq constant.
Of course, the point of class-i is to allow a similar reduction of crossover artifacts without resorting to infinite expansion of the output stage and its heat sink.
Elvee, other than as a design exercise, why build 50W if you only need 1W or so?
Have fun
Kevin O'Connor
Well, if you have to pay 20dB more for it then it's not. Having said that, what you really buy with the 50W unit is a much more decent power supply, that's the real gain. The fact that it does 50W is virtually purely irrelevant ...
Frank
Hi Guys
As I detailed in another thread, I scoped the signals while metering and playing music. Tried various music types and never found a peak higher than 4W.
It is commercially typical for lower-powered amps to have less well designed supplies and care put into the overall product. But... for yourself, and as a DIY project, you hopefully won't make the same compromises. The supply in my amp is far from deficient in any way. It has the headroom I need for my listening situation.
The output stage is "class-i" inasmuch as it has no crossover distortion artifacts and it is the idle current that is directly monitored. It is not true class-i as I do not use the class-i "engine" - maybe the definition of class-i should be stated clearly somewhere?
I used parallel output devices of the modern type with sustained-beta and high-Ft - large packages to dissipate the heat easily and total overkill for the power output. This was influenced by years of experience of my own, observing what other designers were doing and why, and combining the best ideas that worked together. THD is below the resolution of my Radford LDO and analyser, so <0.002% at any audio frequency and at any power level from a few tens of milliwatts to the full clean output of the amp.
As far as I can verify, Self's assertion above is correct (post 71) - he thought it was a new observation and it may be - as were those of many other respected designers. People still debate details and the 'why' of certain phenomena and that will continue; new ideas are often overlooked by the masses.
Have fun
Kevin O'Connor
As I detailed in another thread, I scoped the signals while metering and playing music. Tried various music types and never found a peak higher than 4W.
It is commercially typical for lower-powered amps to have less well designed supplies and care put into the overall product. But... for yourself, and as a DIY project, you hopefully won't make the same compromises. The supply in my amp is far from deficient in any way. It has the headroom I need for my listening situation.
The output stage is "class-i" inasmuch as it has no crossover distortion artifacts and it is the idle current that is directly monitored. It is not true class-i as I do not use the class-i "engine" - maybe the definition of class-i should be stated clearly somewhere?
I used parallel output devices of the modern type with sustained-beta and high-Ft - large packages to dissipate the heat easily and total overkill for the power output. This was influenced by years of experience of my own, observing what other designers were doing and why, and combining the best ideas that worked together. THD is below the resolution of my Radford LDO and analyser, so <0.002% at any audio frequency and at any power level from a few tens of milliwatts to the full clean output of the amp.
As far as I can verify, Self's assertion above is correct (post 71) - he thought it was a new observation and it may be - as were those of many other respected designers. People still debate details and the 'why' of certain phenomena and that will continue; new ideas are often overlooked by the masses.
Have fun
Kevin O'Connor
Well, we have no "masses" here. A lot of folks who march to different drummers...
Perhaps you should start a thread on your system, with some diagrams, schematics and pix especially for the speakers... (serious suggestion)
I don't know of too many speaker systems that produce significant bass on less than 4 watts peak. So that's of interest.
(implies something like a Goto bass compression driver + horn...)
At 4 watts, there are so many easy ways to power it, and class A operation is a non issue.
But if as you say your peaks are 4 watts or lower, then is your average listening level 20dB down from there? Or perhaps your typical listening material is not super dynamic, uncompressed stuff? I ask, because I find 20dB peaks easy to find on material that is either self recorded or else from various other sources...
of course most commercial recordings try to keep the average levels up and the peaks limited...
Of course at the present time one can build a chip amp or two that will do 4 watts easy and have ultra-super low distortion figures...
at 4 watts max, I am curious why you don't just have a straight up class A amp, and why you want to or need to monitor the idle current at all??
_-_-bear
Perhaps you should start a thread on your system, with some diagrams, schematics and pix especially for the speakers... (serious suggestion)
I don't know of too many speaker systems that produce significant bass on less than 4 watts peak. So that's of interest.
(implies something like a Goto bass compression driver + horn...)
At 4 watts, there are so many easy ways to power it, and class A operation is a non issue.
But if as you say your peaks are 4 watts or lower, then is your average listening level 20dB down from there? Or perhaps your typical listening material is not super dynamic, uncompressed stuff? I ask, because I find 20dB peaks easy to find on material that is either self recorded or else from various other sources...
of course most commercial recordings try to keep the average levels up and the peaks limited...
Of course at the present time one can build a chip amp or two that will do 4 watts easy and have ultra-super low distortion figures...
at 4 watts max, I am curious why you don't just have a straight up class A amp, and why you want to or need to monitor the idle current at all??
_-_-bear
Hi Guys
Bear, I described the system in a thread where someone was talking about ways to measure the peak power being used with music, so see that thread. The sats are Yamaha NS10 Studios, and the subs are double labrynths called Behemoths - very efficient and go to 13Hz. Both are about 90db/1W@1m. 90db of sound in my room is intolerable. There are no neighbours to worry about, just our own hearing and desire to enjoy music for many more decades rather than days.
Just setting idle current "high" and leaving it uncontrolled usually results in very poor performing circuits. The JLH10 is a prime example of this. It is also difficult to address thermal aspects if you have no idea what is going on in the circuit. Besides, direct bias control provides tight control over the output stage which I believe helps reduce distortion.
Why do I have to repeat that I tried various music types? Most recorded music is quite compressed, especially anything approaching 'popular'. Even classical music has less than 25db dynamic range on average if it is to be truly listenable. If you want to reproduce 120dB in your own room you are welcome to it, but I implore you not to as you won't be hearing the music for very long at those levels.
If idle current control was not important, there would not be a class-i thread.
Have fun
Kevin O'Connor
Bear, I described the system in a thread where someone was talking about ways to measure the peak power being used with music, so see that thread. The sats are Yamaha NS10 Studios, and the subs are double labrynths called Behemoths - very efficient and go to 13Hz. Both are about 90db/1W@1m. 90db of sound in my room is intolerable. There are no neighbours to worry about, just our own hearing and desire to enjoy music for many more decades rather than days.
Just setting idle current "high" and leaving it uncontrolled usually results in very poor performing circuits. The JLH10 is a prime example of this. It is also difficult to address thermal aspects if you have no idea what is going on in the circuit. Besides, direct bias control provides tight control over the output stage which I believe helps reduce distortion.
Why do I have to repeat that I tried various music types? Most recorded music is quite compressed, especially anything approaching 'popular'. Even classical music has less than 25db dynamic range on average if it is to be truly listenable. If you want to reproduce 120dB in your own room you are welcome to it, but I implore you not to as you won't be hearing the music for very long at those levels.
If idle current control was not important, there would not be a class-i thread.
Have fun
Kevin O'Connor
then you didn't look thoroughly enough at a sufficiently wide range of music.
If your average level were 1W (2.83Vac=4Vpk=8Vpp) and your peak is 4W (5.6Vpk=11.2Vpp) then that is only +3dB average to peak.
White noise has a higher range (average : peak) than that !
Clear your browser cache of temporary internet files. Its a fairly common glitch.
I also asked myself 'what happened with this thread', though for quite a different reason.
I've never read a thread which is so polluted with irrelevant nonsense. Self said this Leech said that etc. They have never written anything about class-i.
So can we return to the original topic, that is, CLASS-I, please!
Cheers,
E.
PS: Damir, YGM.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.