No big deal, but I am a bit confused. You checked the list from Madisound and noticed that you had forgotten to order the one resistor value. I thought you said you ordered those and sent them out, but I must have assumed that. Again, it's no big deal, I just thought that had happened. If I had realized that I would have ordered them with my parts express order.
Out of curiosity, I have always assumed that shunt parts had less of an impact on sound quality. Especially in LCR's, and I often used the cheapest parts I could. In your crossovers, you use relatively expensive solen capacitors for some of the parts (though I did get a yellow mylar cap for one too), is this just because it's what madisound sells, and is the cheapest parts they offer, or do better parts in the LCR make a small difference?
Out of curiosity, do you think an impulse response of the accelerometer readings would give you anything of value?
Also, my acoustic phase measurements look terrible, always have. Any recomendation on the best way to get clean phase measurements. It's been my assumption that the only solution was true anechoic. My assumption though is that a clean phase measurement will tell me if the waveguide/foam is working correctly and that I don't have too much paint build up or glue. What do you think, is that how you evaluate it?
One thing I am seeing consistently in my messy phase measurements is a sharp phase dip around 8K-10K. Does that make sense? It also seems to be relatively smooth with almost no phase angle change from around 900hz to that 8K point. Does that make sense? Would be a good thing if it does, but as I said, I don't trust my measurements. Is spatial averaging good for phase, or will it create more problems?
Out of curiosity, I have always assumed that shunt parts had less of an impact on sound quality. Especially in LCR's, and I often used the cheapest parts I could. In your crossovers, you use relatively expensive solen capacitors for some of the parts (though I did get a yellow mylar cap for one too), is this just because it's what madisound sells, and is the cheapest parts they offer, or do better parts in the LCR make a small difference?
Out of curiosity, do you think an impulse response of the accelerometer readings would give you anything of value?
Also, my acoustic phase measurements look terrible, always have. Any recomendation on the best way to get clean phase measurements. It's been my assumption that the only solution was true anechoic. My assumption though is that a clean phase measurement will tell me if the waveguide/foam is working correctly and that I don't have too much paint build up or glue. What do you think, is that how you evaluate it?
One thing I am seeing consistently in my messy phase measurements is a sharp phase dip around 8K-10K. Does that make sense? It also seems to be relatively smooth with almost no phase angle change from around 900hz to that 8K point. Does that make sense? Would be a good thing if it does, but as I said, I don't trust my measurements. Is spatial averaging good for phase, or will it create more problems?
I know that Dr. Geddes will be skeptical on this, but I highly recommend you bypass all Solen parts on crossover with 0.033uF K40Y-9 400V Soviet surplus from Ebay. They cost dirt cheap and it will be interesting at least. They make the Solen sing like a high dollar sweetie.
I'll pass on Salas comment
I know of no reason why series quality would be any different than shunt quality. I think the whole thing is a bunch of cr_p, but at the very least if you are going to worry about one you should also worry about the other. The electronics don't know that they are in the signal path when they go through a component, its all the same to them. If there is nonlinearity in the shunt path this will show up in the series path as well. This is the kind of thing that points out the hypocrasy of the component craze.
There is no doubt that the overall circuit of a crossover will have a greater or lessor dependency on certain components, but I don;t think that you can assume that shunt makes it less important. In my crossover the shunt LRC paths are all fairly narrow bandwidth so this would make their effects somewhat limited. But the shunt resistor is broad band. The most critical components would likely be the series cap to the HP and the series inductor to the woofer. The woofers series inductor is indeed a critical component and is also one of the most expensive in the whole parts list. I don't ever short change this component.
When you sent the parts list I was on line and posted right back to you that I had forgotten one resistor. I was going to go to Madisound and order it, but somehow that didn't happen. So do you want me to send you the 2.7 ohm or are you all set?
I don't look at phase, so I'm not much help. Averaging of any kind basically destroys the phase.
Without some tie of a vibration measurement to some meaning, its hard to say if an impulse resonse is useful or not.
I know of no reason why series quality would be any different than shunt quality. I think the whole thing is a bunch of cr_p, but at the very least if you are going to worry about one you should also worry about the other. The electronics don't know that they are in the signal path when they go through a component, its all the same to them. If there is nonlinearity in the shunt path this will show up in the series path as well. This is the kind of thing that points out the hypocrasy of the component craze.
There is no doubt that the overall circuit of a crossover will have a greater or lessor dependency on certain components, but I don;t think that you can assume that shunt makes it less important. In my crossover the shunt LRC paths are all fairly narrow bandwidth so this would make their effects somewhat limited. But the shunt resistor is broad band. The most critical components would likely be the series cap to the HP and the series inductor to the woofer. The woofers series inductor is indeed a critical component and is also one of the most expensive in the whole parts list. I don't ever short change this component.
When you sent the parts list I was on line and posted right back to you that I had forgotten one resistor. I was going to go to Madisound and order it, but somehow that didn't happen. So do you want me to send you the 2.7 ohm or are you all set?
I don't look at phase, so I'm not much help. Averaging of any kind basically destroys the phase.
Without some tie of a vibration measurement to some meaning, its hard to say if an impulse resonse is useful or not.
If you don't mind sending me the resistors, I do not have them. If you don't send them I will have to order them myself.
Thanks for the comment on parts, it sounds like my thinking wasn't so divergent from yours. Maybe I was mistaken in believing that the lcr's were of very little issue, but then again, I had mentioned thinking that the Series cap and Series inductor in your crossover were the most critical. You had mentioned that matching the dcr of the series woofer inductor was important. What effect other than spl does it make? I mean, it's a fraction of an ohm, a small fraction at that, so what else does it impact? For instance, if I change from the part thats in there right now, and go from .27 ohms to a new part that is .15 ohms, is the .12 ohm difference really going to throw off the response any? On the other hand, my experience has been that such a small difference in resistance does seem to cause the inductor to heat up less, which I have to imagine is a good thing. I think matching the dcr of that metal cored inductor with an air cored would require a 10 gauge inductor and would be quite huge, but I still may change that part out at a later date.
As to that series resistor, given it's critical nature (relatively speaking), would a MOX not be a better part. I know the MOX resistors aren't the right wattage, but if you combined 4 of them you would have the right wattage, and you would just need a value 4 times the needed one. Even two would probably be ok, or three. Only issue would be matching, they would need to be fairly closely matched to ensure one isn't different enough to become that path of least resistance.
Thanks for the comment on parts, it sounds like my thinking wasn't so divergent from yours. Maybe I was mistaken in believing that the lcr's were of very little issue, but then again, I had mentioned thinking that the Series cap and Series inductor in your crossover were the most critical. You had mentioned that matching the dcr of the series woofer inductor was important. What effect other than spl does it make? I mean, it's a fraction of an ohm, a small fraction at that, so what else does it impact? For instance, if I change from the part thats in there right now, and go from .27 ohms to a new part that is .15 ohms, is the .12 ohm difference really going to throw off the response any? On the other hand, my experience has been that such a small difference in resistance does seem to cause the inductor to heat up less, which I have to imagine is a good thing. I think matching the dcr of that metal cored inductor with an air cored would require a 10 gauge inductor and would be quite huge, but I still may change that part out at a later date.
As to that series resistor, given it's critical nature (relatively speaking), would a MOX not be a better part. I know the MOX resistors aren't the right wattage, but if you combined 4 of them you would have the right wattage, and you would just need a value 4 times the needed one. Even two would probably be ok, or three. Only issue would be matching, they would need to be fairly closely matched to ensure one isn't different enough to become that path of least resistance.
but get I use mills a lot they are good parts, no worries. I like mox, and it's a more recent discovery for me to use MOX instead of wirewound. We shall see, as I actually have some Mills in the correct value for the series resistors. Problem is power handling. They also need to be matched, and those require buying at least a 1/3 as many more for proper matching.
Matt
I'm not going to get into the "parts" discussion.
No a few tenths of an ohm is not going to matter. But I don't see how one iron core inductor would be significantly different than another. Air core would be a problem to get this high a value at that low a DCR. Probably $50-100. And I doubt that heat is a major factor.
I'm not going to get into the "parts" discussion.
No a few tenths of an ohm is not going to matter. But I don't see how one iron core inductor would be significantly different than another. Air core would be a problem to get this high a value at that low a DCR. Probably $50-100. And I doubt that heat is a major factor.
I have no double blind listening tests to support a sound difference, so this is purely a measurement thing. I have a few different inductors of roughly the same value but built different ways. Aircore, CFAC, iron core, and toroid iron core. For the most part they measure similarly, where the big differences seem to come is the rise in impedance with frequency. Aircore and toroid iron core both have much less rise in impedance with frequency than do the normal steel core types. I will show you the data if you want, but I don't know if its valid. I just reran some of it to see, as I couldn't remember, and noticed that the iron core you sent me has over an ohm of resistance by 500hz, where as the CFAC I have is still less than an ohm. In other words, while the dcr of the iron core is lower, the resistance through much of the operating range seems to be lower in the cfac. Could that make a difference? What about the capacitance of the inductors and resistors, inductance of the resistors? I measured the 10 ohm resistors and they have an inductance of .015. The mills is .0098 by comparison. I tried modeling the effect and noted a very slight rise in the high frequencies, but it was on the order of tenths of a decibel, and was very high above 5k for sure. None the less, when I get into the effect of parts, this is the stuff I worry about. But as you said, there has been little or no double blind listening tests to show an actual impact on perceived sound, and I would guess it's very low, a poor value.
The inductor data interests me as this could be significant since thats a change from about .25 to 1 ohm or almost four times. I'll check this on the units that I have when I get a chance. The thing is that its not that this is a bad thing, but I don't account for it in my crossover design and if its that significant then I'll need to.
The inductance on the series resistor should lower the response at the upper end, which would be a good thing. I'd like some series inductance to drop the response above 10 kHz to protect the tweeter. A clipped amp sends lot of energy up into this range and will burn out a tweeter quite easily. Some series inductance would help this a lot.
The inductance on the series resistor should lower the response at the upper end, which would be a good thing. I'd like some series inductance to drop the response above 10 kHz to protect the tweeter. A clipped amp sends lot of energy up into this range and will burn out a tweeter quite easily. Some series inductance would help this a lot.
The series inductance comment makes more sense than what I thought I saw. I wasn't sure if very small trace amounts of inductance somehow had a magical reverse effect. Well you could use a normal vishay wirewound power resistor, which has more significant amounts of series inductance. I think I have some of these to test, but they are probably in the 100's of ohms.
As for the inductor data, let me know what format you want it. In the mean time, I will give you a graph. I need to send the data to my other computer though. Look at the phase change of the two inductors as well, let me know if you think that is important. They seem quite different. The two I will send, since they are the only two I have on hand to show you for values in your crossover, are the iron core you send, and a 16 gauge copper foil air core. The one I have more different types for is a 2.2mh type. The biggest problem with comparing the 16 gauge cfac is that its resistance is so much higher throughout. A more fair comparison would be a 12 gauge cfac and 14 or 16 gauge toroid type.
at 500hz the resistance is 9.7 ohms, 1khz it's 20 ohms. The CFAC and iron core show about a half ohm difference (in favor of the iron core) but by the time you get into higher frequencies over 100hz, the resistance becomes so high the half an ohm no longer seems significant.
phase response of the inductors is interesting. Probably unimportant, but both show a 90 degree shift from 0-90 degrees from DC to 20khz. However, the iron core shows greater phase change than the cfac, followed by a crossing point where the cfac levels off and the iron core begins to reverse again.
Can you accept impedance data in zma or text file format? Let me know and I will send it to you.
As for the inductor data, let me know what format you want it. In the mean time, I will give you a graph. I need to send the data to my other computer though. Look at the phase change of the two inductors as well, let me know if you think that is important. They seem quite different. The two I will send, since they are the only two I have on hand to show you for values in your crossover, are the iron core you send, and a 16 gauge copper foil air core. The one I have more different types for is a 2.2mh type. The biggest problem with comparing the 16 gauge cfac is that its resistance is so much higher throughout. A more fair comparison would be a 12 gauge cfac and 14 or 16 gauge toroid type.
at 500hz the resistance is 9.7 ohms, 1khz it's 20 ohms. The CFAC and iron core show about a half ohm difference (in favor of the iron core) but by the time you get into higher frequencies over 100hz, the resistance becomes so high the half an ohm no longer seems significant.
phase response of the inductors is interesting. Probably unimportant, but both show a 90 degree shift from 0-90 degrees from DC to 20khz. However, the iron core shows greater phase change than the cfac, followed by a crossing point where the cfac levels off and the iron core begins to reverse again.
Can you accept impedance data in zma or text file format? Let me know and I will send it to you.
gedlee said:
I am giving info in good will so to help out, I have nothing to prove or to sell. I am no scientist or vendor. Those capacitors are dirt cheap, anybody can bypass their Solens and report.
Cheers, keep up the good work.
Actually some series resistance can extend the bass -3 dB point significantly.
The B&C woofers are low Qe and give overdamped response in sealed boxes.
The B&C woofers are low Qe and give overdamped response in sealed boxes.
Well maybe the 16 gauge inductor is worth trying. It's not enough to lower the -3 much, I just checked. Based purely on winisd models, as my own measurements have room effects, -3db point goes from 120 hz to 105hz by going from a series resistance of .23 to .75. I mean, that is lower, but it's not like I'm going to hear a much fuller sounding speaker. Additionally adding resistance to around 2-3 ohms makes a more significant difference, bringing the F3 to around 75hz, but you also will lose 2db's in efficiency, and would need a 2 ohm resistor capable of dissipating quite a lot of power. I'd imagine you would want a 50-100 watt wirewound, and it probably should be externally mounted, like Wilson Audio does, for proper cooling.
noah katz said:
That depends on the volume of course. Its not the damping that concerns me but the effect at the crossover. The woofer needs to be as efficient as possible to match the waveguide. Any series resistance is a loss of sensitivity throughout the whole bandwidth.
I just watched the movie "Sunshine" and have to say, these speakers have really transformed my theater, and that's with the crossover being a bit off for the moment and very little setup time and room tuning. I spent the better part of 3 hours last night working on the bass and finally got something that was smooth and even (relatively speaking). While the movie is a bit of a cheesy sci-fi movie, I kept reading mixed reviews, with a lot of acclaim for the sound track. I like most of Danny Boyles films, and the band Underworld, who does the soundtrack, really does a pretty nice job. I figured the film would offer a good mix of explosions and loud noises as well as the potential for a cool sound track.
The ending credits has a song by Underworld that I'm confident, at the levels I'm watching, would have sounded a bit compressed and screachy. Additionally the sound stage was most impressive. Throughout the film I found myself a bit more immersed, and the track didn't sound like it was coming from below the screen, as usual. While it still didn't flow fully throughout the screen (they aren't going to be magic), the stage did seem to be more naturally placed across the screen. With the song, the singer seemed to float in the right area of the screen and wasn't attached to the speakers as usual. Additionally, I noted a very little of the typical compression I get during loud sequences.
On a related note, I used to keep trying to add more and more bass. During explosions I heard distortion that I had always assumed was coming from my subwoofers. My current setup has the output capability to rival most any typical home theater subwoofer, and it's become clear to me that it was never the bass. That distortion was coming from the main speakers. Even with 100hz highpass filters on the mains I'm seeing that they simply could not keep up. Keep in mind too that, while I've talked a lot about high spl ability, I am not listening at particularly loud levels. I would say that the general speech in the film is around 65-70db's placing it roughly 3db's or so below thx reference levels.
Dr. Geddes, I now know why you keep blowing surrounds.
The ending credits has a song by Underworld that I'm confident, at the levels I'm watching, would have sounded a bit compressed and screachy. Additionally the sound stage was most impressive. Throughout the film I found myself a bit more immersed, and the track didn't sound like it was coming from below the screen, as usual. While it still didn't flow fully throughout the screen (they aren't going to be magic), the stage did seem to be more naturally placed across the screen. With the song, the singer seemed to float in the right area of the screen and wasn't attached to the speakers as usual. Additionally, I noted a very little of the typical compression I get during loud sequences.
On a related note, I used to keep trying to add more and more bass. During explosions I heard distortion that I had always assumed was coming from my subwoofers. My current setup has the output capability to rival most any typical home theater subwoofer, and it's become clear to me that it was never the bass. That distortion was coming from the main speakers. Even with 100hz highpass filters on the mains I'm seeing that they simply could not keep up. Keep in mind too that, while I've talked a lot about high spl ability, I am not listening at particularly loud levels. I would say that the general speech in the film is around 65-70db's placing it roughly 3db's or so below thx reference levels.
Dr. Geddes, I now know why you keep blowing surrounds.
gedlee said:
But you do account for this in the crossover right? When you designed the crossover, you tweaked it until you got the measurements you want. All with that inductor in the circuit. Is this not right?
I'll let Dr. Geddes respond to this comment most directly, but looking at my frequency response graphs, it should be clear that the crossover is pretty well optimized. The few problems I have are my own doing, namely I changed a resistor value used in the slopes contour tuning, as I didn't have the right values. The correct values will be here today, and I can post new measurements indicating how flat the response is.
Others more knowledgable than I can talk more eloquently about this, but keep in mind that a part of my comment about the inductor was a bit of a "no duh!" An Inductor is, by its nature, a reactive part, whose true circuit is not just inductance, and not just static resistance and inductance either. It reacts with changing frequency. An inductor needs to have a reactive impedance, otherwise it wouldn't filter. The issue I was discussing is more dealing with how good a job the inductor does. An ideal inductor would have zero resistance through it's pass band and then transition into a text book perfect 6db per octave rising impedance. However, inductors aren't perfect, they have dc resistance, which ends up being the starting point of the resistance, the baseline, and it just increases from there. Further analyzing my own data is showing very little difference between closely matched inductors, with almost identical transfer functions and almost identical impedance. It turns out that some of my claims initially (I supplied Dr. Geddes with correct data) was due to my own incorrect measurement, and the two inductors being different from each other by about .15mh. I also found a 16 gauge aircore inductor which I unwound to 3mh, and measured it. The impedance looks basically the same. The shape is roughly the same, with the difference being that whatever the dcr is will remain roughly the difference between the inductors throughout. Once you get well into the khz range you begin seeing further changes, including the impedance lines crossing each other, suggesting differing slopes, but it's slight, and well outside the pass band.
My tests are all very low power, so the one thing that really would be interesting would be to see at what point you start seeing the core saturation of the iron core for both frequency and power level. It appears that the only advantage, after correcting for value differences as best I could, was that the CFAC had less phase change within it's passband area than did the others, and had a slightly steeper impedance rise (Very slight), that wasn't really noticeable until beyond 10khz. The DCR on the CFAC I have is .75 as compared with roughly .25-.3 ohms for the steel cored unit. That half an own difference remains throughout, but again, by 500hz the resistance has risen greatly.
I wonder what impact the inductor has on the bass alignment given that it's impedance rises so much with frequency. The numbers I gave were wrong earlier, by 500 hz it was 9.7 ohms, not 1 ohm. The CFAC at the same frequency was 10.1 or something like that.
Also keep in mind that I was a relatively early adopter of the speaker kit, as is the case with anything, the first generation has the most kinks to be worked out. I'm sure, as is the case with even the largest speaker manufacturers, Dr. Geddes will have more and more time listening and testing the Abbey's and will be able to improve the crossover appropriately.
Others more knowledgable than I can talk more eloquently about this, but keep in mind that a part of my comment about the inductor was a bit of a "no duh!" An Inductor is, by its nature, a reactive part, whose true circuit is not just inductance, and not just static resistance and inductance either. It reacts with changing frequency. An inductor needs to have a reactive impedance, otherwise it wouldn't filter. The issue I was discussing is more dealing with how good a job the inductor does. An ideal inductor would have zero resistance through it's pass band and then transition into a text book perfect 6db per octave rising impedance. However, inductors aren't perfect, they have dc resistance, which ends up being the starting point of the resistance, the baseline, and it just increases from there. Further analyzing my own data is showing very little difference between closely matched inductors, with almost identical transfer functions and almost identical impedance. It turns out that some of my claims initially (I supplied Dr. Geddes with correct data) was due to my own incorrect measurement, and the two inductors being different from each other by about .15mh. I also found a 16 gauge aircore inductor which I unwound to 3mh, and measured it. The impedance looks basically the same. The shape is roughly the same, with the difference being that whatever the dcr is will remain roughly the difference between the inductors throughout. Once you get well into the khz range you begin seeing further changes, including the impedance lines crossing each other, suggesting differing slopes, but it's slight, and well outside the pass band.
My tests are all very low power, so the one thing that really would be interesting would be to see at what point you start seeing the core saturation of the iron core for both frequency and power level. It appears that the only advantage, after correcting for value differences as best I could, was that the CFAC had less phase change within it's passband area than did the others, and had a slightly steeper impedance rise (Very slight), that wasn't really noticeable until beyond 10khz. The DCR on the CFAC I have is .75 as compared with roughly .25-.3 ohms for the steel cored unit. That half an own difference remains throughout, but again, by 500hz the resistance has risen greatly.
I wonder what impact the inductor has on the bass alignment given that it's impedance rises so much with frequency. The numbers I gave were wrong earlier, by 500 hz it was 9.7 ohms, not 1 ohm. The CFAC at the same frequency was 10.1 or something like that.
Also keep in mind that I was a relatively early adopter of the speaker kit, as is the case with anything, the first generation has the most kinks to be worked out. I'm sure, as is the case with even the largest speaker manufacturers, Dr. Geddes will have more and more time listening and testing the Abbey's and will be able to improve the crossover appropriately.
tomcat9 said:
Well yes, this is true and not true. I rely a lot on the computer models of the crossover so when this model is inaccurate I can be lead astray. But I do check the final crossover and may make changes to it if it doesn't work as planned. But these changes may not be the ideal since I can't try out different approachs as I can with the computer model. Having an accurate model is a real necessity.
I have to say, I don't think people realize just how much a speakers crossover is designed via modeling software these days. I've toured some of the biggest names in hifi, and I've seen the R&D departments, while they claim they do extensive listening tests and measurements to optimize the crossover, what they really mean is, they take a whole lot of measurements of the raw drivers, load that into software, and model 100's of crossover options. Once they narrow it down to one or two basic designs, real world measurements and listening tests are done, but not extensive ones, not that I saw. Listening tests are too unreliable to over use here, and typically models and measurements correlate high enough that it's a waste of research money to keep measuring every little change and option.
The other thing I found in my tours was that often these engineers were stuck trying to deal with a lot of compromises. Price level didn't matter, even the most expensive speakers had a compromise or two they were dealing with. Better on-axis or power response, more efficiency or smoother response, better phase or better impedance, etc etc. They also were often working on the updated crossover they would release in V2 of the model before V1 hit the shelves. I really don't think, in talking with the engineers, that this was just a money making scam. It really seemed like it was a necessity of marketing. Your always able to tweak and modify things endlessly, but at some point you have you put something on the shelves.
One thing I was told, and I recall this from my engineering internships as well, is that new designs are almost always the worst because most of the focus is on the big picture. Often little problems aren't found until it's too late, as they were mostly focused on getting all the big details right, some little details get through. It's a lot easier to fix small problems in an already good design than it is to design from scratch, and not create new little problems.
I also feel, from my own experience, that using listening of a product to modify its design requires extensive long term listening. I really need to spend a lot of time with something before I'm comfortable enough with it's overall character to begin making changes, and being certain they are for the better.
Dr. Geddes may disagree with me here, but based on my own training and experience in analyzing human behaviors and perceptions, it's my opinion that certain things can be easily detected via quick A/B succession, while other things can not. I don't know that small differences made by slight design changes can or even need to be easily picked up in an ABX experiment. In my own research, it's not uncommon for the changes that take place as a result of our experimental condition to be undetectable in immediate stimulus response conditions. Sometimes people are effected by a stimulus in a way that takes a while to develop the visible response. In my work with parenting style, education, etc. it's not uncommon to see no effect for years, development tracks or life tracks not actually changing until the kids are in highschool or even adults.
The other thing I found in my tours was that often these engineers were stuck trying to deal with a lot of compromises. Price level didn't matter, even the most expensive speakers had a compromise or two they were dealing with. Better on-axis or power response, more efficiency or smoother response, better phase or better impedance, etc etc. They also were often working on the updated crossover they would release in V2 of the model before V1 hit the shelves. I really don't think, in talking with the engineers, that this was just a money making scam. It really seemed like it was a necessity of marketing. Your always able to tweak and modify things endlessly, but at some point you have you put something on the shelves.
One thing I was told, and I recall this from my engineering internships as well, is that new designs are almost always the worst because most of the focus is on the big picture. Often little problems aren't found until it's too late, as they were mostly focused on getting all the big details right, some little details get through. It's a lot easier to fix small problems in an already good design than it is to design from scratch, and not create new little problems.
I also feel, from my own experience, that using listening of a product to modify its design requires extensive long term listening. I really need to spend a lot of time with something before I'm comfortable enough with it's overall character to begin making changes, and being certain they are for the better.
Dr. Geddes may disagree with me here, but based on my own training and experience in analyzing human behaviors and perceptions, it's my opinion that certain things can be easily detected via quick A/B succession, while other things can not. I don't know that small differences made by slight design changes can or even need to be easily picked up in an ABX experiment. In my own research, it's not uncommon for the changes that take place as a result of our experimental condition to be undetectable in immediate stimulus response conditions. Sometimes people are effected by a stimulus in a way that takes a while to develop the visible response. In my work with parenting style, education, etc. it's not uncommon to see no effect for years, development tracks or life tracks not actually changing until the kids are in highschool or even adults.