No but the same goes for most audio equipment, and wouldn´t most in here not say that thay can hear differences, and thats why we spent time and effort searching for better sound. And no, I can´t provide any double blind test or statistic argument to proove this. And yes the hi-end industry is a funny place with many ill defined claims and high pricetags, but why the sudden need to be scientific now??
And BTW, there are other sciences than natural science, so an argument (though from wiki -the sudden true source of facts
) in one science does not necesarily translate into a fact in another science... So are we getting any wiser? -No, as long people create answers on other terms than the question they seem to answer, we really aren´t !!Eva, I have no doubt that you do understand the techniques well and even master them. The problem is that when you are asked about the physics and answer with techniques, you go around the question. No one knows everything, and thats the one to premises we all have to face, and if we don´t recognize it, we tend to live in falsh belief and WILL go wrong at some point.
There is no point in proving that the coil (hence filter (+ mechanical FS of driver)) in a loudspeaker is adequate to turn pulses from a PWM amp or any other class D into sound. Of course the inherent filter in the speaker will do that, but it doesn´t change the physics about what the filter actually does. No NFB knowledge or technique can explain that. Eva, not to disrespect you in any way, but I really do think you could benefit from ackowledging the physical explanation `behind´ what you are doing, and realize how imperfect everything is. These imperfections are the actual guidelines to create something truely better. Another thing is; one thing is to make something work, another is to understand it. Alot of things have been made to work, but with very little understanding of why it actually works. this is also a basic premise for any engineering and a lesson we have to remember in order to stay reflexive about our results and learn from them. No computer, signal, modulation, A/D or D/A conversion is perfect. Neglecting that fact, is like living in a parallel world. Maybe a world of computer utopi, but still parallel
regards,
Subwoofer is an euphemism. I managed to get +/-0.25dB load-independent bw to 14khz and -3dB@20khz from 120khz switching.
Juhleren: It's your highly discouraagin opinion against the fact of working amplifier prototypes (highly encouraging). I don't care about your opinion at all as long as measurements tell the opposite. I feel you are very good writer but you have not yet learned to think in the frequency domain. What's your relationship with electronics? There are at least a couple of class D companies in Denmark that are not having a good time due to wrong planning and global crisis (production costs going too high, product released early without meeting specs, etc...)
Juhleren: It's your highly discouraagin opinion against the fact of working amplifier prototypes (highly encouraging). I don't care about your opinion at all as long as measurements tell the opposite. I feel you are very good writer but you have not yet learned to think in the frequency domain. What's your relationship with electronics? There are at least a couple of class D companies in Denmark that are not having a good time due to wrong planning and global crisis (production costs going too high, product released early without meeting specs, etc...)
Last edited:
Eva,
measuring is technique
Interpretating the measurements is science (or should be...)
Practitioners dwell between these paradigms, and should reflect on how to make sense of either. How do you build reference? What are your measurements? Under which conditions? Upon which scientific models do you base your assesment of the data you constructed?
I have only the highest regards of practitioners as You, Eva, but sometimes the choice of theories to back a certain technique overrule any kind of reflection and critical behaviour. Thats sad IMHO. We are all here to learn, no?
My affiliation as practitioner is only 14 years of loudspeaker development and a master of science. Currently i´m doing a phd funded by my university, on a different topic, though the way reasoning is made between real world empirical work and scientific theories is also a fundamental part of my scientific work. This thread could actually make a terrific case
I´m not one of the class D danes, and do feel sad about the way business seems to be going, but i´m not affiliated in any way.
What is you affiliation with class D and audio design, Eva?
measuring is technique
Interpretating the measurements is science (or should be...)
Practitioners dwell between these paradigms, and should reflect on how to make sense of either. How do you build reference? What are your measurements? Under which conditions? Upon which scientific models do you base your assesment of the data you constructed?
I have only the highest regards of practitioners as You, Eva, but sometimes the choice of theories to back a certain technique overrule any kind of reflection and critical behaviour. Thats sad IMHO. We are all here to learn, no?
My affiliation as practitioner is only 14 years of loudspeaker development and a master of science. Currently i´m doing a phd funded by my university, on a different topic, though the way reasoning is made between real world empirical work and scientific theories is also a fundamental part of my scientific work. This thread could actually make a terrific case
I´m not one of the class D danes, and do feel sad about the way business seems to be going, but i´m not affiliated in any way.
What is you affiliation with class D and audio design, Eva?
Hi,
Independent of the preparation of Juhleren, I totally agree with him. his philosophy is correct.
only by solving problems can increase knowledge, no looking the other chips that perform for you.
the class D is very simple in concept but damn complicated if you want to know what defects on certain signals, they generate other defects on the audio or behavior.
eg. I spent a long time to understand and take action to obtain a completely clean signal near the rail Vcc (near clip) with only 700mV of loss.
Today, I realize that this feature does not exist in any amp.
perhaps not worth anything, there is too much ignorance on the market but I Including cement my knowledge, this has value.
I have set of measures of this (photo) for Eva, I do not remember the thread
Independent of the preparation of Juhleren, I totally agree with him. his philosophy is correct.
only by solving problems can increase knowledge, no looking the other chips that perform for you.
the class D is very simple in concept but damn complicated if you want to know what defects on certain signals, they generate other defects on the audio or behavior.
eg. I spent a long time to understand and take action to obtain a completely clean signal near the rail Vcc (near clip) with only 700mV of loss.
Today, I realize that this feature does not exist in any amp.
perhaps not worth anything, there is too much ignorance on the market but I Including cement my knowledge, this has value.
I have set of measures of this (photo) for Eva, I do not remember the thread
Do any of you ever listen to what you have designed? It's like our former engineer who was of the belief that digital is digital peroid, and that it should all sound the same. But I would always say that is true until one converts to analog. That's where the difference comes in. Same with the Class D designs, they do ampilify the signal. But don't sound very good in doing so. Build your best Class D design and then do a comparision to a good Class A design.
Hi,
I personally have used an advanced instrumentation (use in MDI) But even my ear (I think that is trained to hear jazz and classsic), and comparation with Usher amp I think I have succeeded in what I wanted.
even if it is not for me the final decision. But customers
I agree with you.
measures will help to eliminate defects which can listen to it, eg, if high IMD, you listen very poor electric piano or snare etc.. or low resolution, cimbals and brush, are confusion of "SSSS " eheh.
Just to note, I had assumed that one of the bigger problems with slowing down the carrier too much is frequency dependent change (dv/dt actually) in comparator response, which is an issue in both clocked and self oscillating circuits (anywhere you use a comparator). If you can make the carrier fast enough that the input difference goes blazing through the transition region whether the signal input is tiny, full-scale, 20Hz or 20k, then you're really getting somewhere. Again, this is just a supposition. Maybe some better endowed engineering type brains could calculate out the distortion characteristic and practical magnitude that this situation would cause. That would be very interesting but is beyond me.
Yes, with most comparators response time is dependant on input overdrive. But that wouldn't be a problem in our case since it is quite a simple task to keep the slew-rate at the comparator's input the same for a lower switching frequency: You will just have to increase its level.
When we are at it: Yes - working with decent levels at the comparator input is always a good idea.
Regards
Charles
When we are at it: Yes - working with decent levels at the comparator input is always a good idea.
Regards
Charles
LT1016 has been there for 20 years now. Datasheet gives some insight into high speed circuit design.
cds.linear.com/docs/Datasheet/1016fc.pdf
And TL3016 has been there for 14 years.
<10ns response time with 5mv overdrive.
Some people here is speaking with far more authority than their actual level of electronics know-how would allow for.
cds.linear.com/docs/Datasheet/1016fc.pdf
And TL3016 has been there for 14 years.
<10ns response time with 5mv overdrive.
Some people here is speaking with far more authority than their actual level of electronics know-how would allow for.
What's the difference in TL3016 output delay from zero crossing to peak where there is an input sine of 5V, 15kHz and a 5V 45kHz triangle?
That part bottoms out at 80dB common mode rejection too, so running the inputs full scale could get you some measurable problems due to that.
One thing I know for sure, perfect never enters the picture.
That part bottoms out at 80dB common mode rejection too, so running the inputs full scale could get you some measurable problems due to that.
One thing I know for sure, perfect never enters the picture.
Last edited:
Orders of magnitude smaller than the error from the triangle wave modulation process itself, or the timing error from a real world output stage
Particularly if some HF hysteresis is added (just 10s of nanoseconds time constant).Note that a 100ns error at 45khz switching produces as much distortion as a 10ns error at 450khz. This is the main reason for using a frequency as low as allowed by the modulation scheme, which defines the lower limit.
Also note that in phase-shift self-oscillating schemes the slope of the differential signal that appears at comparator inputs is pi/2 the slope found on a triangle wave scheme idling at the same frequency (50% better). Furthermore, the slope does not start to drop until extreme duty cycles are reached (approx. 10%-90%, 70% power) despite the drop in oscillation frequency, so actually comparator and power stage accuracy is better employed than in constant frequency schemes.
Last edited:
I just punched in a test circuit to LTspice with an LT1016.
Using a 20kHz sine (exacerbating a little), the difference in carrier rising edge induced output transitions when the sine was falling versus sine rising was almost 10nS. At 45kHz carrier that's 1 in 2200 parts.
With a 500kHz carrier the difference in delay was about .05nS, or 1 in 40000 parts.
About 13dB difference in pulse width distortion?
Using a 20kHz sine (exacerbating a little), the difference in carrier rising edge induced output transitions when the sine was falling versus sine rising was almost 10nS. At 45kHz carrier that's 1 in 2200 parts.
With a 500kHz carrier the difference in delay was about .05nS, or 1 in 40000 parts.
About 13dB difference in pulse width distortion?
Hi Jeff, I think you overstate/exaggerate a little..(?) your point is taken.
I've not listened to Bel Canto or Audio Zone, from what I've read, they are not the best. I have listened a bit to nuforce. Nuforce was alright. I can't remember anything specific, but I do know I would not have bought it.
I did put together a a UCD400 (HG w/HrX regs) and dual mono linear supplies. I think the results are great. Certainly not night and day! (I doubt anyone would suspect it was class D only by listening)
I've not listened to AVI, but I'd say UCD beats most A or A/B amps I've listened to or compared. I will admit that a top class, class A amp may still have an edge on sound quality. I also think class D is has recently become competitive, and is closing this gap. It might be a while before everyone gets a chance to compare again, and lose the old impressions of class-D. (so many mediocre ICEpower amps don't help any)
I'm certainly not against Class A or A/B, though I do like how cool my UCD's run, vs my old A/B power amp.
Maybe I'll build or borrow a good class-A and compare again myself.
Add some HF positive feedback to make the 45khz transition as fast as the 500khz one
Also, are you sure that the simulator is accurate and is not quantizing everything to 10ns intervals?
PSpice has "step ceiling" and "print step" settings to control that.
Last edited:
Time to come back to topic intention ...
I have not listened to a Pass F5 but I know that Nelson Pass earned a great reputation for years in the audio lover´s world. I know from the past that a good Class A can deliver a fascinating performance. In my case it was a musical fidelity A1. Since that I have respect from Class A but they are not "trendy" regarding nowadays greenpower and ecological mentalities.
I think that progresses in material research and nanotechnologies made the way open to develope better and better Class D implementations. Every principle (class A, A/B or D) can be done in a bad way but a well matched system with a good Tripath amp was my favourite up to now. I am also curious for new systems like UCD and will give them a chance in future.
A good Class A/B I found very good was a Creek A52 poweramp.
For the future I think Class D will dominate. Class A/B is (maybe was) good for the economy - much material, much profits, less sound I never could understand the sense of 30 kilogramms of amp material. I believe that the bigger audio companies offer the bad qualitity goods only to sell the better designed versions for the xxx time of the price. I don´t need to listen to a Pass F5 to come to the result that it will blow away the 1000 $ amps of the bigger companies. My life and diyaudioforum experiences tell me that ...
I have not listened to a Pass F5 but I know that Nelson Pass earned a great reputation for years in the audio lover´s world. I know from the past that a good Class A can deliver a fascinating performance. In my case it was a musical fidelity A1. Since that I have respect from Class A but they are not "trendy" regarding nowadays greenpower and ecological mentalities.
I think that progresses in material research and nanotechnologies made the way open to develope better and better Class D implementations. Every principle (class A, A/B or D) can be done in a bad way but a well matched system with a good Tripath amp was my favourite up to now. I am also curious for new systems like UCD and will give them a chance in future.
A good Class A/B I found very good was a Creek A52 poweramp.
For the future I think Class D will dominate. Class A/B is (maybe was) good for the economy - much material, much profits, less sound
I never could understand the sense of 30 kilogramms of amp material. I believe that the bigger audio companies offer the bad qualitity goods only to sell the better designed versions for the xxx time of the price. I don´t need to listen to a Pass F5 to come to the result that it will blow away the 1000 $ amps of the bigger companies. My life and diyaudioforum experiences tell me that ...
Last edited:
To eclectic2k:
- Standard low-bias class-AB is made by joining two different transfer functions (one for each half of the output stage) that intersect just at zero-crossing, the most critical point for low and medium level signal amplification (and the source of crossover distortion). These transfer functions are never matched (in most amps no attempt at all is made) resulting in all kinds of distortion to start with, before (very high) negative feedback is applied to attenuate harmonics.
- High-bias class-AB has a wide overlapping region between the two transfer functions creating a third class-A region around zero-crossing. This effectively gets rid of open loop distortion at zero-crossing, but at the expense of creating two high-distortion crossover zones placed symmetrically at some distance from zero volts. The distortion comes from the so called "gm doubling" effect: Open loop gain is 6dB higher in the middle class A region than in the other two class B regions, and the transition is abrupt.
- Class A has a continuous transfer function, it's only advantage, which does not give any crossover distortion. However, there is usually substantial asymmetric distortion at medium to high signal levels before negative feedback is applied. This happens due to the unavoidable non-linearity of gain devices as V and I are varied over a wide range (no matter if they are tubes or transistors). Clipping is seldom symmetrical.
- Class D has three operating zones and two crossover points between them placed symmetrically, like high bias class AB, but there is no gm-doubling effect at all. It features the same monotonic linearity as Class A in the middle zone, which is centered around zero crossing. The width of this middle zone can be extended by increasing inductor ripple current. For example, in my bigger amps (4kw/2ohm/4kg) this zone allows for 250W/8ohm of class-A-grade power, thanks to >16App inductor ripple. The other two zones are symmetrical and quite linear too, since it's timing precision what matters, not the linearity of the transistors themselves. However, the amount of negative feedback that can be applied without trouble is not as high as in class AB, but there is much less non-linearity to compensate too.
So in essence, high power class-D keeps the single and only good thing about class A, while providing tons of extra power.
To TheDealer: You are missing a very important fact. During the years Nelson Pass gained all his prestige the majority of the amplifiers on the market were pure junk, while the stuff he was doing was not junk at all, it was good, it was years ahead of his time. However, now we are at year 2011 and he is no longer ahead of his time. He and most of his customers are quite old now, and I don't see many chances of continuity in new generations.
- Standard low-bias class-AB is made by joining two different transfer functions (one for each half of the output stage) that intersect just at zero-crossing, the most critical point for low and medium level signal amplification (and the source of crossover distortion). These transfer functions are never matched (in most amps no attempt at all is made) resulting in all kinds of distortion to start with, before (very high) negative feedback is applied to attenuate harmonics.
- High-bias class-AB has a wide overlapping region between the two transfer functions creating a third class-A region around zero-crossing. This effectively gets rid of open loop distortion at zero-crossing, but at the expense of creating two high-distortion crossover zones placed symmetrically at some distance from zero volts. The distortion comes from the so called "gm doubling" effect: Open loop gain is 6dB higher in the middle class A region than in the other two class B regions, and the transition is abrupt.
- Class A has a continuous transfer function, it's only advantage, which does not give any crossover distortion. However, there is usually substantial asymmetric distortion at medium to high signal levels before negative feedback is applied. This happens due to the unavoidable non-linearity of gain devices as V and I are varied over a wide range (no matter if they are tubes or transistors). Clipping is seldom symmetrical.
- Class D has three operating zones and two crossover points between them placed symmetrically, like high bias class AB, but there is no gm-doubling effect at all. It features the same monotonic linearity as Class A in the middle zone, which is centered around zero crossing. The width of this middle zone can be extended by increasing inductor ripple current. For example, in my bigger amps (4kw/2ohm/4kg) this zone allows for 250W/8ohm of class-A-grade power, thanks to >16App inductor ripple. The other two zones are symmetrical and quite linear too, since it's timing precision what matters, not the linearity of the transistors themselves. However, the amount of negative feedback that can be applied without trouble is not as high as in class AB, but there is much less non-linearity to compensate too.
So in essence, high power class-D keeps the single and only good thing about class A, while providing tons of extra power.
To TheDealer: You are missing a very important fact. During the years Nelson Pass gained all his prestige the majority of the amplifiers on the market were pure junk, while the stuff he was doing was not junk at all, it was good, it was years ahead of his time. However, now we are at year 2011 and he is no longer ahead of his time. He and most of his customers are quite old now, and I don't see many chances of continuity in new generations.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.