Go Back   Home > Forums > >
Home Forums Rules Articles diyAudio Store Blogs Gallery Wiki Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

TB W4-657SB Measures
TB W4-657SB Measures
Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 8th August 2007, 03:48 PM   #51
MarkMcK is offline MarkMcK  United States
diyAudio Member
 
Join Date: Aug 2003
Location: Ohio
Just a general warning . . . if easy is what you are looking for then you need to look in another thread. There are plenty of threads from a limited number of authors claiming significant improvements with ease. They slather the cones with all sorts of doping materials. One even suggested toothpaste.

All I can say is "BEWARE!"

As I have already stated, I am not going to be giving modification details. I do want to speak about measurement/analysis/graphing and measurement of the W4-657.

I am starting with measurements. There will be five images to illustrate just a couple aspects of measurement/analysis/graphing.

For me this is obvious and general knowledge, so forgive me if it is obvious and general for you too. First, all signal acquisition and analysis software suites have quirks. Not good or bad, just the way things are.

The first attachment is a Praxis display of an ideal dirac impulse response (IR). The data for the IR consists of just one positive integer preceded and followed by long strings of zero values. In the old days, we would graph this with bars. Since we were working with discrete data values we graphed them as such. Today, most of the software tries to imitate what would be seen on the screen of an analog oscilloscope displaying an analog signal.

For this example, the three data points graphed have time signatures of 3.7926667, 3.8135, and 3.8343333 msec. The only actual time signature with an associated data value is 3.8135. Regardless, as is proper, the graphing shows symmetrical rise and fall.

The second graph in the first attachment shows the frequency response and phase response of the ideal dirac. You can't actual see the frequency response because it is a perfect match of the phase response and is thus hidden.

Now, how do you get the phase response. Phase is an arbitrary concept. You might have seen a few SpeakerWorkshop posting with this weird looking increase frequency sawtooth pattern overlying the frequency response. This is caused by the arbitrary start of the phase transformation being set to the start of the data file and not the relevant data.

To get a more useful phase transformation you have to account for this difference in the beginning of the data set and the time position of the relevant data. For Praxis you input a delay value. This moves the start reference for the phase transformation to whatever data file time position you want. For the phase transformation presented her, that delay time was set to 3.8125 msec. This places the start for the phase transformation to the center of a data place.

SpeakerWorkshop uses a slightly different approach. SpeakerWorkshop sets the start place of the phase transformation through the data start and end of the data truncation option. Instead of placing the start point in the middle of a data time value, it references it to the beginning of a data time value.

Anyway, the next four posts from me will introduce some minor imperfections in the phase transformation set point and some minor imperfections to the IR to show how they impact the frequency and phase response.

Mark
Attached Images
File Type: gif one.gif (10.2 KB, 363 views)
  Reply With Quote
Old 8th August 2007, 05:59 PM   #52
soongsc is offline soongsc  Taiwan
diyAudio Member
 
soongsc's Avatar
 
Join Date: Mar 2005
Location: Taiwan
Look forward to seeing some interesting things, specifically if it's the TB W4-657SB.

I think there are many ways to work with cones, depending on the material of cones and the purpose of treatement. I'm probably the one that started with toothpaste on metal cones just because it is easy to find the critical points, look at the effects, easy to remove, and the sort to alternative treatment when production of some sort is desired. I actually have toothpaste still on some drivers laying around. Whether tweaking something is easy or not really depends of the final goal and the driver selected, one thing that we can be sure of is that one way normally does not fit all.

For whomever is interested in understanding Speaker Workshop, the source code is now open source. It's quite interesting to know that SW has made an attempt to remove time of flight.

Recent studies helped me realize how flat impedance makes a difference in sound, I difference between smooth to flat and wavily flat made quite audible difference. Equally interesting is that shorting rings create certain sonic signatures in the mid and high frequencies even though graphically it shows better high frequency response due to lowered inductance. Is this easy to understand? Not really. We just need to be aware what the limitations of each implementation are. Yes, the impulse response is a much better way of understanding some of the problems, I think this is an indication that different data makes different problems more understandable, and hopefully easier to solve.

However, how critical is the quality of the generated impulse?
below are impulses generated from two different cards. If we do a loop back test, both will give perfectly flat phase response overlapping the frequency response. Does this make a difference during measurement? Something to think about.
Attached Images
File Type: jpg 4impulses.jpg (75.4 KB, 348 views)
__________________
Hear the real thing!
  Reply With Quote
Old 10th August 2007, 01:05 PM   #53
MarkMcK is offline MarkMcK  United States
diyAudio Member
 
Join Date: Aug 2003
Location: Ohio
My second attachment shows the frequency and phase response when the Praxis delay is set one sample before and one sample after the data. Remember there is only one sample with data value. For this first run I am working with an ideal dirac.

If the delay is set to the sample before the ideal dirac, then the energy is lagging this arbitrarily determined point. And if we set the delay to the sample after, then the data is leading this second arbitrarily set point. Of course, this leading and lagging is frequency dependent and the transformation produces a graph showing the amount of lead or lag increasing with frequency.

Now, the preceding post presents data that belongs in a different conversation. First, those are not graphs of diracs. Instead they are transformations of an mls signal. That transformation and the anti aliasing filter produces very high frequency artifacts. As long as everything has been designed properly in the software, when you further transform the transformed IR to a frequency and phase response the artifacts are higher in frequency than the highest frequency of the analysis. Thus even the odd looking IR responses shown can produce flat frequency responses. If you extend the frequency response graph out to the sample rate (assuming the software you are using will allow for this) you will see the very steep roll-off of the anti aliasing filter. That is what causes the ringing after the IR. Since these are IR transformations, this ringing is also mirrored before the before the IR.

Just because a claim or statement can be said or written does not mean that it has any significance or even meaning. Knowledge builds on knowledge. The more you know about something the better questions you can formulate. The more you know the less likely you are to be confused by meaningless complications. The more you know the less likely you are to confuse symptom and cause.

I suppose a funny example of this is Forrest Gump's mother in the movie "Forrest Gump." One of her two favorite sayings was, "Stupid is as stupid does." One reason that it is funny is that it is empty of meaning. Thus the person saying it is shown to be just as lacking.

So, let us try to the best of our abilities to avoid as much stupidity as we can. Impossible to achieve perfection, but it is good to minimize whenever we can.

Mark
Attached Images
File Type: gif two.gif (15.6 KB, 333 views)
  Reply With Quote
Old 10th August 2007, 03:23 PM   #54
soongsc is offline soongsc  Taiwan
diyAudio Member
 
soongsc's Avatar
 
Join Date: Mar 2005
Location: Taiwan
A few things that we need to know about the Dirac ideal impulse are:
Infinitely short
Infinite power
Unit energy

Due to this limitation, this cannot happen in reality. We can however use the not-so-ideal Dirac as shown if we are aware of a few things.

1. Output smoothing filters effects of the signal generator and other limited bandwidth equipment before the signal reaches the DUT.

2. The anti-aliasing filter effects.

3. Input and output sample rate effects.

To minimize the effects of these non-ideal conditions, we cannot use the mathematically generated Dirac as the reference, but must use whatever reaches the DUT as the reference. Assuming that in the process of reading back the signal that reaches the DUT and the output from the DUT there is not difference between the anti-aliasing filters, and no latency between the read data, we can then assume that the difference between the signals is caused by the DUT, and thus derive what kind response to the DUT would have occurred if driven by the mathimatically generated Dirac. Note that the assumptions may not be realistic depending on desired accuracy of the measured data.

As shown in the previous post, this is one of the difficulties most face when trying to measure minimum phase, especially when there is a significant time lag involved when reading the DUT response. The higher the sample rate we use, the more ideal Dirac we can get, and the less phase change due to one sample difference when creating a frequency/phase response.
__________________
Hear the real thing!
  Reply With Quote
Old 12th August 2007, 03:55 PM   #55
MarkMcK is offline MarkMcK  United States
diyAudio Member
 
Join Date: Aug 2003
Location: Ohio
My posts are barely on thread subject and then only because I have measured the 657SB and will be posting more measurements of the 657SB. Soongsc keeps wanting to take this way off topic. I have noticed that this is a pattern he often follows. I guess he just needs to say things and have people respond to him. I will pass on the opportunity from here on.

This attachment is where measurements are a little quirky. The importance of this and my making these preliminary posts is to share some of the problems of measurement. I hope that others looking at posted data will be cognizant of the problems and be more critical. The data fakers always leave behind clues. It is possible to tell who is faking and by how much they are faking their "data." There is no need for me to serve as the data police and catch who is doing dirty deeds. With a little sophistication and knowledge, you can do it to.

The dirac pulse in this example has been low pass filtered. It does, however, still have identical rise and fall times. The data has been redistributed to represent this bandwidth limitation while keeping the energy level the same. The ideal pulse was one significant data point with a value of 1. This second sample has three significant data points with values of .2, .6, and .2. The three significant data points still sum to the number 1.

Despite the symmetrical data, the software graphs the resulting IR as having asymmetrical slopes. Also, by using the same delay from attachment one, while the high frequency response is now limited, we still can show perfect phase. It does this because of the symmetrical data. When measuring transducers, you will never find one with perfectly symmetrical rise and fall times.
Attached Images
File Type: gif three.gif (10.4 KB, 298 views)
  Reply With Quote
Old 13th August 2007, 05:13 AM   #56
soongsc is offline soongsc  Taiwan
diyAudio Member
 
soongsc's Avatar
 
Join Date: Mar 2005
Location: Taiwan
Sound like it's time for me to listen and see what Mark presents. Very interesting subject on how data can be faked. Hope we all learn a thing or two from this. I still might pop in with questions if something looks mysterious to me.
__________________
Hear the real thing!
  Reply With Quote
Old 13th August 2007, 02:51 PM   #57
MarkMcK is offline MarkMcK  United States
diyAudio Member
 
Join Date: Aug 2003
Location: Ohio
With attachment four, the impulse response is made asymmetrical. First significant data point has a value of .2 and the second significant data point has a value of .8. Again, the sum is one.

This type of data envelope is commonly seen in transducers. This impulse response is the pattern for when you have a bandwidth limited rising response. The magnitude of the error with actual transducers is usually much worse than shown here, but these two simple imperfect impulse responses of attachments three and four show the sensitivity to error.

We use impulse response transformations to increase our understanding of what we see in the impulse response. Mainly we use a frequency response transformation. Phase response is used less often and is more arbitrary in its presentation and interpretation. I find phase response of little value when evaluating a minimum phase system such as a full range transducer. Still, it is very useful for providing clues for when the data is problematic (whether because of a sad lack of knowledge on the part of the tester or intentional deception).
Attached Images
File Type: gif four.gif (10.5 KB, 264 views)
  Reply With Quote
Old 14th August 2007, 12:45 PM   #58
MarkMcK is offline MarkMcK  United States
diyAudio Member
 
Join Date: Aug 2003
Location: Ohio
Attachment five shows the one before and one after sample delay offsets for leading and lagging phase. As stated in the reply of attachment two, this is a phenomena of the extremes. Here I have shown high frequency examples. The same applies to the low frequencies as well. No transducer is capable of infinite frequency response and will always show high and low frequency roll-off.

When examining an impulse response file with a phase transformation, look for the unexpected at the extremes. This will always show when excessive equalization has been used.

Also look to the bandpass. As the leading and lagging graphs show, phase will still be zero away from the stop band. When a transducer is claimed to have flat response and yet you cannot get the midband (bandpass) phase close to zero regardless of how you adjust your delay, this is a clue that something unnatural has been done to the data.

This assumes that you have access to the impulse response file. I doubt that I will have many people willing to send their impulse response files anymore. Since my interest was not in if the files were altered, but rather how, I am not concerned with ending my access to questionable data.

Still, most who alter their data are not sophisticated in doing so. Most are much better at changing a frequency response or phase response than they are at changing and impulse response. If you want to question a particular claim, just ask the author to publish the impulse response graph.

The impulse response graph is always more revealing of accuracy. It is also why few people publish it.

As I make available measurements of my last work on the W4-657SB I will try to clarify some of the points I have been trying to make here with these five attachment posts.

The 657SB is an interesting transducer as much because of its problems as because of its potential. It is a nice test case to examine what may be a myth about metal's superiority as a diaphragm material. Maybe it is superior, but I would like to understand why the typical claims for metal are not in evidence.
Attached Images
File Type: gif five.gif (15.8 KB, 217 views)
  Reply With Quote
Old 15th August 2007, 12:53 AM   #59
soongsc is offline soongsc  Taiwan
diyAudio Member
 
soongsc's Avatar
 
Join Date: Mar 2005
Location: Taiwan
Quote:
Originally posted by MarkMcK

...I doubt that I will have many people willing to send their impulse response files anymore. Since my interest was not in if the files were altered, but rather how, I am not concerned with ending my access to questionable data.
...
I think anyone would not mind sending such data as long as there can be some meaning full discussion concerning the data. At least I know I would if so. Not all tools are perfect. If discussion can reveal shortcommings of various tools, it only helps us all that are using them.

Up to know, the data showing phase shifts due to window starting location are typical for shifts of one sample, meaning that unless the impulse itself is altered, one can make shifts limited to the sample rate of the sound card, and thus the amount of phase shift. I do not see the limit of the previous data, but if those data showing the phase leads and lags are one sample increments when adjusting the window start location, then probably a sample rate of 48KHz is used.

Isn't faking data illegal?
__________________
Hear the real thing!
  Reply With Quote
Old 15th August 2007, 02:36 AM   #60
DDF is offline DDF  Canada
diyAudio Member
 
DDF's Avatar
 
Join Date: Sep 2004
Location: Ottawa
Quote:
Originally posted by MarkMcK
I doubt that I will have many people willing to send their impulse response files anymore. Since my interest was not in if the files were altered, but rather how, I am not concerned with ending my access to questionable data..
Since this is in this thread, I have to ask: are you insinuating I or anyone else in this thread faked the data, or that they were taken incorrectly? The generalities don't serve this discussion, please be specific.

Cheers,
Dave
  Reply With Quote

Reply


TB W4-657SB MeasuresHide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
3875 sounds better, 3886 measures better? sendler Chip Amps 9 24th February 2009 06:30 AM
3-way around the TangBand W4-657SB: what drivers? tcpip Multi-Way 4 1st June 2006 07:07 AM
W4-657sb.... nerd of nerds Multi-Way 0 1st November 2005 01:36 PM
Charles Hansen measures audio interconnects? Steve Eddy Everything Else 1 21st June 2004 06:03 AM
pot feels broken but measures ok? travis Parts 3 12th May 2003 06:46 PM


New To Site? Need Help?

All times are GMT. The time now is 09:11 AM.


Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 15.00%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Copyright ©1999-2018 diyAudio
Wiki