I think even less than true 16 bit resolution would be ok but maybe higher sample rate would be a good idea.
I'd like to stay away from audio products because as we all know all sorts of fishy filtering may be involved.
one of my concerns is that some of these guys are Chinese manufacturers and their products are not up to spec. the big guys seem to have way larger prices.
what would you do?
you're right in some ways.Should I? I might see it quit as opposite. You guys, like from Philips marketing department ones, has to convinced us, consumers, that Class D is superior not only for your profits but it is somewhat better for us, consumers, as well besides just been ECO-friendly. That is what marketing department is getting payed for actually right?
BTW I cannot see any issues with that since most consumers has not idea what frequency domain is about or why absence of FIR makes LPs sounds better than Philip's RedBooks. Or like mercury filled CFLs are well know to be ECO-friendly nowadays.
the way I see it class D has the advantage of being small and light. less (cheaper) case, less heat sink, less shipping cost.
gets up to operating temp faster that a large class A amp.
but the recent super expensive and physically large offerings?
but if it can be done small, light and right, why not? that's what we're talking about here.
the "big and hot" guys about class D:
http://media.avguide.com/BG_Audio_Electronics_2012.pdf
(from pg. 25)
Bob Carver, John Curl, Nelson Pass, Jeff Rowland etc.
my favorite part is the electron spin direction line. it never ceases to amaze me how some of these super-high-end guys cling to the "everything is audible" theory.
CES 2011 Audio Research interview (mentions class D starting from 7:23):
CES 2011 - Audio Research - YouTube
did you hear that PS Audio is planning to release a UCD-based product?
and that Jeff Rowland is rumored to give op on D?
The hilarious part ...
Is that if you take ANY system - analog, single ended, or PP, or hybrid, or tube entirely, or Class D, or Class H ... ANY system - and compare the digitized input to the sound-pressure waveform at say 12 feet from the speakers - binaural/stereo of course - with $1000 or better instrumentation reference microphones ... you find that what is "at your ear" is so, so, so different from the original signal as to make ones jaw drop.
This is the myth of all this tight control, all this Class Warfare (A, B, AB, C, D, G, H) crap. So much of the influence on the sound is the damned speakers, and all the nonlinear EMF feedback and counter-action taken by the amplifier's designers to "correct" things.
Give me a really clean and well laid out class D, 100 watts-or-better per channel, some well made, but not ridiculously over-spec'd speakers, with a soft 3-way crossover ... good phase response ... and I'm happy, very much so. Give me me a nice simple, non-feedback open loop Class A, 30 watts or better ... tube output and well designed, but not over-the-top signal path ... and I'm also happy. Definitely need different speakers though - different amplifier dynamics! yet, I'd be happy. Same goes for a nice class AB/PP or even a well designed transistor-output modern class B! Get the speakers right for each, and you'll be happy.
i've OWNED all of these ... in various incarnations. I've recommended and installed these for others ... in a wide variety of incarnations. The bottom line is ... can't choose an amplifier topology without making a speaker matching "thru listening" session or two.
That's why they have high-end audiophile stores.
GoatGuy (and no: the at-listener waveform NEVER looks much like the source!)
Is that if you take ANY system - analog, single ended, or PP, or hybrid, or tube entirely, or Class D, or Class H ... ANY system - and compare the digitized input to the sound-pressure waveform at say 12 feet from the speakers - binaural/stereo of course - with $1000 or better instrumentation reference microphones ... you find that what is "at your ear" is so, so, so different from the original signal as to make ones jaw drop.
This is the myth of all this tight control, all this Class Warfare (A, B, AB, C, D, G, H) crap. So much of the influence on the sound is the damned speakers, and all the nonlinear EMF feedback and counter-action taken by the amplifier's designers to "correct" things.
Give me a really clean and well laid out class D, 100 watts-or-better per channel, some well made, but not ridiculously over-spec'd speakers, with a soft 3-way crossover ... good phase response ... and I'm happy, very much so. Give me me a nice simple, non-feedback open loop Class A, 30 watts or better ... tube output and well designed, but not over-the-top signal path ... and I'm also happy. Definitely need different speakers though - different amplifier dynamics! yet, I'd be happy. Same goes for a nice class AB/PP or even a well designed transistor-output modern class B! Get the speakers right for each, and you'll be happy.
i've OWNED all of these ... in various incarnations. I've recommended and installed these for others ... in a wide variety of incarnations. The bottom line is ... can't choose an amplifier topology without making a speaker matching "thru listening" session or two.
That's why they have high-end audiophile stores.
GoatGuy (and no: the at-listener waveform NEVER looks much like the source!)
P-P,
If you are concerned with 'fishy' filters, you can always buy it and try it with test waveforms. Most major distributors will take returns if you don't like it. I think you can select filterless inputs and choose 24-bit 96 kHz PCM for a huge raw data file.
If I were to do it and 8-bits (2 mV sensitivity) is enough, one of the newer O-scopes from Tek are great. 500 Msamples/sec 4 channels, color screen $700. Done.
TBS1000 Digital Storage Oscilloscope | Tektronix
I actually have much nicer $10k O-scopes at work but they are overkill (2 Gsamples/sec).
If you are concerned with 'fishy' filters, you can always buy it and try it with test waveforms. Most major distributors will take returns if you don't like it. I think you can select filterless inputs and choose 24-bit 96 kHz PCM for a huge raw data file.
If I were to do it and 8-bits (2 mV sensitivity) is enough, one of the newer O-scopes from Tek are great. 500 Msamples/sec 4 channels, color screen $700. Done.
TBS1000 Digital Storage Oscilloscope | Tektronix
I actually have much nicer $10k O-scopes at work but they are overkill (2 Gsamples/sec).
I think Jeff Rowland says in the interview above that there's no thing as the sound of a bias class.
I belong to the camp that believes that excessive bandwidt isn't needed in terms of accuracy. I.e. 40 khz is definitely sufficient for an amp, which can be easily achieved even with a class-d amp running at 250 kHz carrier frequency.
Most sources that we use nowadays are restricted to approx 20 kHz so there simply isn't anything one could gain in using a shortwave transmitter as audio amp.
It is clear that a wide bandwidth amp will contribute less additional time-domain distortion to an given signal than one with a smaller bandwidth. But keep in mind that the time domain distortion of speakers and source (i.e. storage method) is already greater than that of the low-bandwidth amp.
The time-domain distortion of an amp can be corrected by using methods that are known/used for decades in the telecomms industry already, but which are seldom used in the audio business.
Regards
Charles
Most sources that we use nowadays are restricted to approx 20 kHz so there simply isn't anything one could gain in using a shortwave transmitter as audio amp.
It is clear that a wide bandwidth amp will contribute less additional time-domain distortion to an given signal than one with a smaller bandwidth. But keep in mind that the time domain distortion of speakers and source (i.e. storage method) is already greater than that of the low-bandwidth amp.
The time-domain distortion of an amp can be corrected by using methods that are known/used for decades in the telecomms industry already, but which are seldom used in the audio business.
Regards
Charles
The trick is called phase-equalisation.
By the use of 2-nd order Allpass sections one can achieve flat group-delay response at the expense of slightly increased total group-delay. So if you achieve flat group-delay up to 20 kHz with your bandwith-limited amplifier and feed it a signal that is bandwidth limited to 20 kHz then there is simply nothing that an amp with excessively wide bandwith could do better.
Phase equalisation could even take the time-domain distortion of speakers into consideration.
Regards
Charles
By the use of 2-nd order Allpass sections one can achieve flat group-delay response at the expense of slightly increased total group-delay. So if you achieve flat group-delay up to 20 kHz with your bandwith-limited amplifier and feed it a signal that is bandwidth limited to 20 kHz then there is simply nothing that an amp with excessively wide bandwith could do better.
Phase equalisation could even take the time-domain distortion of speakers into consideration.
Regards
Charles
It's not a sound card but can act as a 2-input, 2-output x 24-bit @ 96 kHz, 128x oversampling audio interface while plugged into USB. By no filters, I meant that the feature for the recorder to 'simulate' the response curve of certain famous studio microphones can be defeated. Not sure what else you expect from a $270 recorder - I think this is quite amazing performance.
But if you are truly looking for a scientific measurement, you should use an O-scope or vibration analyzer.
Here you can see a worked out exaple of a phase equaliser. It isn't for a class-d-amp though but for a 2nd order LR crossover but the principle is always the same.
The cyan trace is the original group-delay, the peaking curves are the group-delays of each 2nd order allpass section alone. The royal blue one is the sum of all of them showing a reasonably flat response up to higher frequencies. If the bandwith of the signal is lower than the bandwith of the correction then there is no time-domain distortion in the output signal anymore. There is however an increase in total group-delay.
When we look at the figures involved for a typical class-d amp this would be like pushing the start button of the CD player some fractions of a millisecond later.
regards
Charles
The cyan trace is the original group-delay, the peaking curves are the group-delays of each 2nd order allpass section alone. The royal blue one is the sum of all of them showing a reasonably flat response up to higher frequencies. If the bandwith of the signal is lower than the bandwith of the correction then there is no time-domain distortion in the output signal anymore. There is however an increase in total group-delay.
When we look at the figures involved for a typical class-d amp this would be like pushing the start button of the CD player some fractions of a millisecond later.
regards
Charles
Attachments
phase accurate,
If it was only so easy to use a simple filter to correct the phase angle of a speaker you would have a big hit on your hand! But the phase angle with crossover attached is often a complex value, with multiple phase angle changes if it is not a single full range speaker. The moment we bring the actual acoustical center into the equation and the complex interaction of all the different crossover topologies into the picture you are generally screwed. Physical time alignment of device helps a bit but then that also would require a coaxial design which again I know is still not correct in any implementation that I know of, there is still a physical though much smaller error in the alignment of the acoustic center. It is very hard to design a physical solution and this is just with let's say two devices and now add a third and put that on a distributed baffle and it is getting very complex indeed. The speaker amplifier interface is a mess and that is where the magic is going to come from to finally get where we are trying to get. I agree that it doesn't matter what amplifier topology you are using, each and every topology has its own unique properties to overcome.
If it was only so easy to use a simple filter to correct the phase angle of a speaker you would have a big hit on your hand! But the phase angle with crossover attached is often a complex value, with multiple phase angle changes if it is not a single full range speaker. The moment we bring the actual acoustical center into the equation and the complex interaction of all the different crossover topologies into the picture you are generally screwed. Physical time alignment of device helps a bit but then that also would require a coaxial design which again I know is still not correct in any implementation that I know of, there is still a physical though much smaller error in the alignment of the acoustic center. It is very hard to design a physical solution and this is just with let's say two devices and now add a third and put that on a distributed baffle and it is getting very complex indeed. The speaker amplifier interface is a mess and that is where the magic is going to come from to finally get where we are trying to get. I agree that it doesn't matter what amplifier topology you are using, each and every topology has its own unique properties to overcome.
so let me see if I get it. what shows in AP2's plot is actually group delay. which is pretty much phase. and an amp with a bit more BW than 20kHz (which even most mundane current class D amps achieve) guarantees low phase shift at 20kHz.
and if it isn't low enough you equalize for phase.
am I correct?
Since the main topic in this thread isn't speakers I don't get into this deeply here. The diagram shown is for a Linkwitz Riley crossover of 2nd order. Not for a whole speaker. But it can be done for complete speakers and it IS being done.
The whole point was about the time-domain distortion of a class-d amp with its main culprit the output filter. This can be corrected easily with a phase equaliser. If you are able to measure the group-delay distortion caused by a speaker's lowpass behaviour accurately then there is nothing stopping you from taking this into consideration as well. You will end up with a system that would make hyperfast amplifiers obsolete.
Regards
Charles
Edit: Although this thread is a bout class-d amps resist to link to the step response and FR diagram of a coaxial two-way speaker that was corrected this way. Although there is still need for some fine-tuning you can see an already superior step response compared to most other speakers:
http://www.d-amp.org/include.php?path=forum/showthread.php&threadid=1192&entries=40
The whole point was about the time-domain distortion of a class-d amp with its main culprit the output filter. This can be corrected easily with a phase equaliser. If you are able to measure the group-delay distortion caused by a speaker's lowpass behaviour accurately then there is nothing stopping you from taking this into consideration as well. You will end up with a system that would make hyperfast amplifiers obsolete.
Regards
Charles
Edit: Although this thread is a bout class-d amps resist to link to the step response and FR diagram of a coaxial two-way speaker that was corrected this way. Although there is still need for some fine-tuning you can see an already superior step response compared to most other speakers:
http://www.d-amp.org/include.php?path=forum/showthread.php&threadid=1192&entries=40
Last edited:
I think his claim is that it affects the sound of transients, veils them etc.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.