john curl said:
Oh?
I am listening to music, which even in some strident post-modern forms still exhibit tunes (well, mostly). Most of this are written in 5-bar music style, i.e. where each note has a frequency, in other words lasts for some several (dozen?) cycles. I see continuous waves as contrasted to percussion and take the points argued here regarding the same; but what percentage of any music programme does that constitute? I am also told by analysts that the "sound" of percussion can vary quite substantially depending on your seat vis-a-vis the precussion instrument. To what degree are small differences in percussion characteristics (I mean that of the instrument as heard by the time it reaches one) audible?
I mentioned earlier that I stood on the sideline, having come onto this (most interesting) thread late. I still have to find an audible reason for avoiding a small output series inductor. In my own case I simulated as well as measured as well as I am able to, with such an inductor of up to 6uH. I included tests mentioned by Graham here or on another thread, regarding transient anomalies that can exist in nfb applications not necessarily shown by steady state signals. While I could do without an inductor, what folks would hang on to amplifier outputs is outside my control.
No, my amplifier is not stable with all sorts of capacity hung on, neither do I expect it to be. It is clear from the loop gain/nfb/phase characteristics, which are there for other reasons, that it will not be so unconditionally. I have found improvement in stability spread by the use of an inductor, and no reason to not use it. My loudspeaker simulation (lo-hi, cross-over at 2,2 KHz) shows an inductive impedance at 20 KHz of some 80uH. Another 6 odd uH will have an audible influence under domestic conditions how?
I am not into a "look Ma, no coil!" competition, and while respecting the arguments put forward here, my stance must be as stated. But I will keep on reading .... there are knowledgable guys here. If not regarding those offending series coils, one learns other things.
Regards
Hi Graham,
I have posted before having received/simultaneous to your last post. It would appear that I missed your previous comments regarding the effect of the error voltage giving a perceived lower frequency. Again, I tried to catch up here at some speed. Before "re-inventing the wheel", would you mind telling about where I should go back to - date, number of post, whatever?
Thanks.
I have posted before having received/simultaneous to your last post. It would appear that I missed your previous comments regarding the effect of the error voltage giving a perceived lower frequency. Again, I tried to catch up here at some speed. Before "re-inventing the wheel", would you mind telling about where I should go back to - date, number of post, whatever?
Thanks.
Philosophically, I tend towards the Peter Aczel POV on what you can and can't hear. The straightforward analysis of what a 2-7 uH inductor does says you shouldn't hear it. I've spent the afternoon making, measuring, and installing new inductors on an ancient amp I wanted to fool with. The disturbing thing is that my output network has a large effect on the distortion distribution- in places where it should have no effect at all. No evidence of oscillation/instability is present, though maybe there's some local stage thing going on- have to look harder. IMO, that almost has to be it, unless there's some kind of magnetic interference from the coil. I also put various coils on the vector impedance meter and confirmed what I mentioned earlier. The common small diameter coils people wind don't follow the formulas very well, and you simply have to measure them, preferably at a MHz or so. Coils I wound years ago thinking they were in the couple uH range, were about 0.7 uH. IMO, simulation only deals with part of the problem here- there are a lot of strays to think about, and some seemingly minor construction details that play a large roll. Hey, at least I'm having fun! 🙂
To make some sense of what I previously stated, what is a continuous sine wave? I know that you think that you know, BUT you forget that a REAL continuous sine wave started at the beginning of the universe and will continue until the end of the universe. It cannot in fact, turn on and off. Guess what music does, it continually turns sine waves on and off, therefore there are no continuous sine waves in our world experience. However, sine waves used for simulation or even testing are close to REAL continuous sine waves, but they don't closely resemble music, because sine waves making music are continually turning on and off. THINK!
Now what is the result of this?
Now what is the result of this?
Most interesting point, John.
That was why I tried to steer away from yet another term in audio not well defined. But I would have thought that there is clarity/realism on this (well, mostly). In practical (engineering) terms there could be several aspects. One is to define a continuous wave as one consisting of enough cycles to cause any change as a result of its presence/commencement to reach a new steady state. In contrast a transient causes a "temporary" (but now I must define that!) change that is not complete before the cessation of the transient.
That as far as electronics are concerned. Musically I would stick with my definition as being enough cycles of the same frequency to cause the ear to hear a definable tone or pitch, as e.g. "that is Middle-C" or such. But this may not correspond to the electronics-valid definition if one splits hairs. Yet we are mainly dealing with what is audible, or at least capable of creating something that is audible.
Any way (and the above is not above reproach), I would contend that a good amplifier design should not be upset by transients, i.e. by really any stimulus that it is supposed to handle. This brings me to another point allied to this whole business, if not part of the central theme. No amplifier is really capable of perfectly handling whatever is thrown at it, e.g. somewhere there will be phase shift, causing a signal in that region to cause undesirable artifacts. Point then: What are such signals doing there? This is hopefully in some super-sonic region - why are there not filters to keep such an area clean? It is a little like showing a high overshoot on a <1uS risetime square wave input - that should not matter (within good design practices, of course). Throw a square wave at it with the rise time limited to say >5 uS and if that is OK, see to it that nothing else can come along.
I hope that this is proof of my thoughts, John. If so, thanks for the stimulus.
Regards.
That was why I tried to steer away from yet another term in audio not well defined. But I would have thought that there is clarity/realism on this (well, mostly). In practical (engineering) terms there could be several aspects. One is to define a continuous wave as one consisting of enough cycles to cause any change as a result of its presence/commencement to reach a new steady state. In contrast a transient causes a "temporary" (but now I must define that!) change that is not complete before the cessation of the transient.
That as far as electronics are concerned. Musically I would stick with my definition as being enough cycles of the same frequency to cause the ear to hear a definable tone or pitch, as e.g. "that is Middle-C" or such. But this may not correspond to the electronics-valid definition if one splits hairs. Yet we are mainly dealing with what is audible, or at least capable of creating something that is audible.
Any way (and the above is not above reproach), I would contend that a good amplifier design should not be upset by transients, i.e. by really any stimulus that it is supposed to handle. This brings me to another point allied to this whole business, if not part of the central theme. No amplifier is really capable of perfectly handling whatever is thrown at it, e.g. somewhere there will be phase shift, causing a signal in that region to cause undesirable artifacts. Point then: What are such signals doing there? This is hopefully in some super-sonic region - why are there not filters to keep such an area clean? It is a little like showing a high overshoot on a <1uS risetime square wave input - that should not matter (within good design practices, of course). Throw a square wave at it with the rise time limited to say >5 uS and if that is OK, see to it that nothing else can come along.
I hope that this is proof of my thoughts, John. If so, thanks for the stimulus.
Regards.
Hi John,
What a change it is having someone else trying to explain this. (Someone to share the hits I have had to fend off on my own up to now.)
Hi Johan,
Most theory in the books relates to steady/isolated sines.
Go back to fundamental thinking about what happens when the energy a plectrum stores in a string is SUDDENLY released, or a triangle is SUDDENLY hammered.
The voltage developed by a series output choke from any suddenly starting current is simultaneously superimposed upon all voltage waveforms no matter how many cycles are involved at other frequencies.
Even if the input to a power amplifier is filtered - as with CD - the amplifier will be incapable of error correcting to a high degree of accuracy if it does not have a much higher closed loop bandwidth.
When the amplifier open loop bandwidth (within that closed loop after stabilisation) is less than or is phase shifted with respect to the signal input bandwidth, the NFB will end up overdriving small signal stages, leading to loss of crossover control in class-AB, due to correction running out of phase with both input voltage and output current.
Cheers ............ Graham.
What a change it is having someone else trying to explain this. (Someone to share the hits I have had to fend off on my own up to now.)
Hi Johan,
Most theory in the books relates to steady/isolated sines.
Go back to fundamental thinking about what happens when the energy a plectrum stores in a string is SUDDENLY released, or a triangle is SUDDENLY hammered.
The voltage developed by a series output choke from any suddenly starting current is simultaneously superimposed upon all voltage waveforms no matter how many cycles are involved at other frequencies.
Even if the input to a power amplifier is filtered - as with CD - the amplifier will be incapable of error correcting to a high degree of accuracy if it does not have a much higher closed loop bandwidth.
When the amplifier open loop bandwidth (within that closed loop after stabilisation) is less than or is phase shifted with respect to the signal input bandwidth, the NFB will end up overdriving small signal stages, leading to loss of crossover control in class-AB, due to correction running out of phase with both input voltage and output current.
Cheers ............ Graham.
Circuits that are indistinguishable from a simple perfect gain stage using constant amplitude sine wave testing :
- A Compressor
- A circuit that mutes the first period of any signal, whatever it is
- A noise gate
- MP3 compression
- etc
Basically uif you scratch your head you can find zillions of things that cannot be characterised by ye old sinewave test, or any constant amplitude test... I do believe we'll have to think about new tests someday 😀
- A Compressor
- A circuit that mutes the first period of any signal, whatever it is
- A noise gate
- MP3 compression
- etc
Basically uif you scratch your head you can find zillions of things that cannot be characterised by ye old sinewave test, or any constant amplitude test... I do believe we'll have to think about new tests someday 😀
I understand, Graham. Like you, I listen first, then 'rationalize' later, when confronted with a new idea.
Since when have sine wave tests been the only game in town? I think my GR tone burst generator was made back in the '40s or '50s to address a need they were well aware of. One cycle anyone? Part of a cycle? Obviously we have square waves, but IMO you need control over rise and fall time to do serious testing. At work we just got a nice Agilent arbitrary function generator- that would seem an essential item for anyone building amps commercially and worried about this stuff. A bit too many $ for home, but I suspect you can program a sound card to do the same thing. The examples of things that don't lend themselves to sine wave testing are, I hope, far different than linear amplifiers. Linear amplifiers should in fact behave predictably and according to the conventionally understood equivalency between time and frequency domain. Ain't no magic here. As for my unstable amp, when I pulled out the old output network and rebuilt it, I added some extra wire length on the output zobel and didn't return it to the same ground point. As usual, the builder was at fault, and regardless of what's printed on the schematic, the amplifier will dutifully behave according to the actual values in the box, strays, finite ground resistances and all. But unless that stuff gets entered into the simulation...
I built my first tone burst generator from an EICO electronic switch back in 1959. Most people here do not have a tone burst generator and believe in simulations which are continuous ac signal based instead. That is the problem, and IF you did not know what to look for, you would miss what Graham is talking about.
Actually, nearly anyone using a soundcard for measurement has tone burst capability. Looking at the results on my scope, it seems to work pretty darn well.
Hi John,
Thanks for that.
This is a point I have repeated. Using sines, by the time you have enough to observe a measurable and readable (unchanging/stable) amp/phase response, the error has long passed and been missed.
With single cycle excitation both start-up and run-down error can be observed, but within music and with a series output choke these responses become part of the continuing waveform, and are thus immeasurable in isolation unless compared with a fixed group delayed version of input (old fashioned fundamental nulling), or by differencing with an identical second channel driving a nominal value, purely resistive load.
Cheers ......... Graham.
Thanks for that.
This is a point I have repeated. Using sines, by the time you have enough to observe a measurable and readable (unchanging/stable) amp/phase response, the error has long passed and been missed.
With single cycle excitation both start-up and run-down error can be observed, but within music and with a series output choke these responses become part of the continuing waveform, and are thus immeasurable in isolation unless compared with a fixed group delayed version of input (old fashioned fundamental nulling), or by differencing with an identical second channel driving a nominal value, purely resistive load.
Cheers ......... Graham.
john curl said:
John,
Yes, you are right. Music must by necessity consist of sinewaves that come and go, although with traditional string and wood instruments you most often see sines that start small, grow in amplitude and then decay. But in principle they have indeed to start from zero amplitude.
Now, if such a sine would start from zero in one instant, to a certain amplitude the next instant, that would be an infinite fast step. This would then contain an infinite spectrum of all possible frequencies. This is the signal Graham uses in his work.
There are at least two reasons that such a step will not occur in the real world:
- in the case of real instruments, the vibration of the string or air or whatever cannot start infinitely fast as these are mechanical systrems;
- All musical signals eventually come through signal processing equipment that limits the bandwidth like ADC and DAC anti-aliasing filters. Therefore, that infinite spectrum from the infinite fast starting edge will be reduced to a spectrum within the audio band.
So, while an inifitely fast edge can be expected to bring on all kinds of trouble in an audio amp that was never designed to handle that, it is misleading to base your judgement of an amp on how it does handle this lightspeed stuff. The amp will, in real world, be asked to handle audio-band (or a bit above) signal frequencies. So, if you want to see how your amp handles suddenly starting sines from real-world sources, run the sines through an audio-band low-pass or a 50kHz low-pass if you want a belt and suspenders too.
I have invited Graham several times to do that, so far fruitlessly. This is a pity, because he may well have an interesting thing which is less convincing because the test signal where this all is based on, is a non-audio signal.
Jan Didden
Hi Janneman,
there are in most power amps an effective low pass filter, all be it ~200kHz.
Our starting and stopping "music" signals will be filtered by this and the amp must handle them.
Why insert an extra 50kHz low pass filter?
Graham,
if the instant start, fast rise time signal were passed through the "standard" low pass filter attached to most amps, would the error be masked by the filtering? Would the error become unmeasurable?
Presumably your amps have some form of RFI filter. I assume YOU can still hear the improvement in sound quality that your proposed/actual modifications bring about after passing through that "standard" filter.
there are in most power amps an effective low pass filter, all be it ~200kHz.
Our starting and stopping "music" signals will be filtered by this and the amp must handle them.
Why insert an extra 50kHz low pass filter?
Graham,
if the instant start, fast rise time signal were passed through the "standard" low pass filter attached to most amps, would the error be masked by the filtering? Would the error become unmeasurable?
Presumably your amps have some form of RFI filter. I assume YOU can still hear the improvement in sound quality that your proposed/actual modifications bring about after passing through that "standard" filter.
AndrewT said:
Andrew,
I'm not married to a 50kHz filter, it was just an example. 200kHz I find quite high though; the further you open the window the more junk flies in!
The point is that whatever signal you use to test this or anything should realistically be bandlimited as any real world signal would be. Whether it would need to be 20kHz, 40kHz, 60kHz or whatever low-pass is to some extent arbitrary. My preference would be to make it as low as possible to keep any hf noise, junk etc out, but that's just that, my preference.
Note that CD or DVD audio is either filtered just above the audio band, or, when they use oversampling, much higher because the oversampling moves the digital artifacts higher up the band, but in both cases there is no signal above the audioband.
So, even if you would have in theory a music signal with very fast starting rise time, by the time it gets in the amp input jack it is pretty much reduced to the audio band.
Jan Didden
But,
50kHz is just 3.2uS and the usual range is 0.3uS to 1.5uS fitted to the front end. 200kHz is right in there at 0.8uS.
Some DVD and DVD-A are going out to 50kHz and beyond, but whether there is real signal content in there is debateable. It will however allow for a decrease in the rise/fall times of start/stop signals.
Most/many amplifier manufacturers are now advertising that they have stretched the upper end response to accomodate the high sample rate digital sources.
50kHz is just 3.2uS and the usual range is 0.3uS to 1.5uS fitted to the front end. 200kHz is right in there at 0.8uS.
Some DVD and DVD-A are going out to 50kHz and beyond, but whether there is real signal content in there is debateable. It will however allow for a decrease in the rise/fall times of start/stop signals.
Most/many amplifier manufacturers are now advertising that they have stretched the upper end response to accomodate the high sample rate digital sources.
AndrewT said:
That doesn't seem too smart. Who wants to include sample rate artifacts in his listening experiences? 😉
Higher sampling rate means that the sampling artifacts are 'folded' to a band much higher than audio. So that allows you to use a filter that starts just above audio but is rather shallow and has less phase shift while still cutting out most of the sampling noise that is farther out.
Higher sampling freq does NOT produce higher freq music!
Jan Didden
It's amazing how some people with alleged golden ears are complete deaf to irrefutable argument.
Cheers,
Cheers,
- Status
- Not open for further replies.