Well pay attention!
What I suspect is happening is that when one attenuates first the HF energy being lower drops out if there is no dither. Bill W. suggests correctly the LF will also acts as dither. My reference source points out that music is a frequency weighted signal, so it is possible to lose HF when no or little level from LF present and that there is sum and difference signals which are perceived as add HF but actually are similar to harmonic distortion.
BTY still up for Monday if I recall correctly?
Your mistake is here, the dither is added as TPDF noise at the 16 bit level then you truncate, the information is encoded in the density of the last couple of LSB's. Looking at one sample is meaningless. With dither the signal is below the noise floor, remember the quoted -96dB is rms 20-20k not sample by sample. Really Ed this is basic and easy to demonstrate.
Really so you can't do 6kHz by itself at -80dB or if you do 100Hz at 0dB plus 6kHz at -80dB you get different answers? Just asking, since this is easily proven wrong.
BTW please provide me with a plot of these sum and difference signals, but not from a bad A/D with DNL problems
What dither? You presume this was done and I don't. The issue brought up was the other musical energy would act as dither. I believe that does not meet the Triangular Probability Density Function effectiveness and results in artifacts.
(Again I assume that is what you meant by TPDF.)
(Can it be we actually are in agreement except for how the sample was prepared, which neither of us really knows.)
Ok so we differ on this. If I have any signal at -80 on a 16 bit file, reduce the level by 60 db and then reproduce it as the MSB 16 bits without dither, yes I think you lose the signal.
If you add a close to full scale signal at the same time as the -80 one then as Bill W. pointed out you will still have the lower level signal.
Right, anyone serious would use dither this way it is generally the default behavior on all software I have seen. Arguing assuming that it was done, wrong in this case BTW, seems a bit odd. Any professional mixing and preparing tracks knows this.
Who said attenuate. Your statement implied digitizing low level signals in the presence of high level ones gave different answers. I'm not talking Bill's process now. As I said a minute ago at this time if someone is working without understanding when and how to dither they are back circa 1982 and this discussion is a waste of time. The file mix can be done with no loss of HF content.
It may be we are in agreement, just not communicating!
1982? Maybe 1942! Certainly by 1962.
Statements of "fact" like this contribute to that. What else am I to think? Yes you can mix a -60dB and 0dB track and get it to work as long as you do not do it the wrong way.
Is there any hard rule that determines that you can't wind a speaker with say a 20 ohm voicecoil and have it be as good a speaker as if you used a 6 ohm voicecoil?
Also, what does the schumann resonance actually look like on a scope? Are there any FFTs or waveforms, maybe even recordings floating around?
Also, what does the schumann resonance actually look like on a scope? Are there any FFTs or waveforms, maybe even recordings floating around?
Keantoken, there is no reason that you can't change the impedance of the voicecoil but there are always trade-offs in what you are doing. You could add a much longer winding length with the same gauge wire and this would increase the impedance at the cost of an increased mass. You can also decrease the wire gauge and this will increase the impedance at the cost of power handling. So you see it is always a balancing act. There have been many older bass speakers in the Pro audio realm that had higher impedance such as 16 ohm 15" bass speakers and also compression drivers. Run a pair in parallel and you had an 8ohm impedance and this was often done in some of the very large bass enclosures with two 15" bass drivers in a ported enclosure. There are many factors such as gap length and whether you are going with an underhung or overhung voicecoil and you'd have to look at the BL factor to determine efficiency. Just a few of the things a speaker builder has to look at. Here is a website you can look at for some simple demos of the variables you have to deal with. This makes things look simple but I assure you that there is much more to it than what they are showing in those demos. FINEMotor - Speaker Design software - Professional Products - loudsoft
actually the differences are 2nd order from practical differences in fill factor, insulation thermal conductivity
the 1st order principle is the same motor geometry, voice coil geometry and conductor mass can be wound to "any" Z
the wire gage/turns/inductance/resistance/BL factor all change in coordination to give effectively an ideal transformer between the electrical terminals and essentially "the same" mechanical-acoustic response of the given motor design
but the 2nd order limits are noticeable, single or dual layer windings are the more common, winding to too high a Z does make the 2nd order considerations limiting with high layer counts, fine wire
this is clearly evident in dual coil speakers - the coils can be connected in parallel or series for "the same" acoustic parameters, only the electrical terminal Z is changed by the expected transformer ratio ( 1:4 )
the 1st order principle is the same motor geometry, voice coil geometry and conductor mass can be wound to "any" Z
the wire gage/turns/inductance/resistance/BL factor all change in coordination to give effectively an ideal transformer between the electrical terminals and essentially "the same" mechanical-acoustic response of the given motor design
but the 2nd order limits are noticeable, single or dual layer windings are the more common, winding to too high a Z does make the 2nd order considerations limiting with high layer counts, fine wire
this is clearly evident in dual coil speakers - the coils can be connected in parallel or series for "the same" acoustic parameters, only the electrical terminal Z is changed by the expected transformer ratio ( 1:4 )
Last edited:
AFAIK
I did learn my Linear Systems from the Mech E dept so multi-domain modeling doesn't make my eyes cross as quickly as some straight EE types
speaker manu white papers, Klippel tech notes make a lot of dynamic speaker info available freely
I have found similar relations useful in motion control - DC motors, actuators have similar principles
knowing operating principles, limitations can be helpful even just for specifying parts from OEMs
I did learn my Linear Systems from the Mech E dept so multi-domain modeling doesn't make my eyes cross as quickly as some straight EE types
speaker manu white papers, Klippel tech notes make a lot of dynamic speaker info available freely
I have found similar relations useful in motion control - DC motors, actuators have similar principles
knowing operating principles, limitations can be helpful even just for specifying parts from OEMs
It is cool but not the best you can do distortion wise. Translate the curves into an analog equivalent and gain 5-10dB. Also gets rid of the latency issues that make it more complicated to use DSP'd speakers in an AV setup.
And for once, skin effect actually means something. Since copper coated aluminium is becoming more popular for voice coils I'm sure the math is more complex.
For commercial products we often tweak the driver impedance to maximize the power we can get from the amp chips. With current tools its not difficult to optimize everything. Unfortunately those tools are a bit out of reach for the hobbyist, not to mention getting a driver with a different voice coil to see if it works.
I just had a badly placed alligator clip arc through the silkscreen and short 45V out of 34mF of capacitance. This happened while I was turning up the variac, so it was apparently voltage breakdown through the silkscreen. It blew a few teeth off of the alligator clip along with about 1mm radius of power plane. The amp appears to be fine.
Sooooooooo... Do you think the silkscreen melted so the alligator clip contacted the plane, or did the arc through the silkscreen short out most of the energy without any metal to metal contact?
Also why did the foil just blow off in a radius rather than just fusing or melting? I think the sparks I saw were foil chunks, although it could be the foil just vaporized and it was the alligator clip that sparked. The radius blown out of the foil is not ragged at all, pretty well-defined and unburnt fiberglass showing through.
Sooooooooo... Do you think the silkscreen melted so the alligator clip contacted the plane, or did the arc through the silkscreen short out most of the energy without any metal to metal contact?
Also why did the foil just blow off in a radius rather than just fusing or melting? I think the sparks I saw were foil chunks, although it could be the foil just vaporized and it was the alligator clip that sparked. The radius blown out of the foil is not ragged at all, pretty well-defined and unburnt fiberglass showing through.
Last edited:
Maybe for a simple LR4+ level adjust, but if the crossover is complex and you need 10 or more gain stages in the X-over to implement the functions it would get challenging to beat the digital approach.
As any AV processor has delay management I really don't see where there is a problem? Can you explain please.
Bill,
Just my experience x 2.
When I design a speaker, I first set xover and compensations in miniDSP and subsquently translate the slopes into an analog version. Well designed drivers in well designed enclosures don't need 10 or more gain stages. My experience is with LR4 followed by the odd shelving or notch filter, and the analog version always measures better. But it may be just my awesome designing skills which threw me off track here.
As to the advantages of not having latency in your speakers in an AV setup: it does not have to be a problem, but as I mentioned correctly, it makes matters more complicated. It is not a trivial exercise to time align your speakers with your screen. It goes without saying that for live monitoring, latency is a killer.
Just my experience x 2.
When I design a speaker, I first set xover and compensations in miniDSP and subsquently translate the slopes into an analog version. Well designed drivers in well designed enclosures don't need 10 or more gain stages. My experience is with LR4 followed by the odd shelving or notch filter, and the analog version always measures better. But it may be just my awesome designing skills which threw me off track here.
As to the advantages of not having latency in your speakers in an AV setup: it does not have to be a problem, but as I mentioned correctly, it makes matters more complicated. It is not a trivial exercise to time align your speakers with your screen. It goes without saying that for live monitoring, latency is a killer.
Fairy nuff. I was thinking more about LX521 levels of complexity, where the trade off is harder. of course some may argue that SL was thinking like an EE when he designed it so there are ways of implementing it with fewer opamps. Example. he equalises flat, then adds in the curve he wants. I am on the fence over it.
Ah yes. I was thinking about front to rear speaker sync. Checked the nanoAVR and that doesn't have picture delay. Not sure how real a problem it is. I do know I am remarkably insensitive to sync issues at a low level. I used to have a set top box that would drift and I wouldn't notice until it was over .25 seconds. But I don't put much focus into watching TV compared to listening to music.
(aside: watching films in cinemas does annoy me as I do focus on the problems)
- Status
- Not open for further replies.