Site Links
Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Categories
Who's Online (5)
- Analogkid455 12:54PM
- rjj45 12:54PM
- Silver1omo 12:54PM
- traw 12:54PM
- Turn2 12:54PM
Please review the site Rules, Terms of Service, and Privacy Policy at your convenience. Rules, TOS, Privacy
Get familiar with the reaction system: Introducing the Reaction System
Distortion vs Flat Frequency Response
I recognize that this is one of those topics that isn't conducive to easy answers, but I'd appreciate any insights nonetheless. I'm in the (glacially slow) process of designing my next build. For the mids I'm considering the SB15CAC and the Satori MR13 or MR16 (and several others). The SB features exceptionally flat response through the pass band (a few hundred to 2k Hz or so, in this case) and very good distortion. The MRs feature state of the art distortion with less impressive response.
So, how do you weigh response vs distortion? Part of what has gotten me thinking about this is the posts in the distortion thread that point out that the distortion measurements have to be made at a very high volume to get the distortion out of the noise floor. If that's the case, does it really matter that much?
Oh, and ignore price.
So, how do you weigh response vs distortion? Part of what has gotten me thinking about this is the posts in the distortion thread that point out that the distortion measurements have to be made at a very high volume to get the distortion out of the noise floor. If that's the case, does it really matter that much?
Oh, and ignore price.
Comments
My take is you need both. Lower distortion sounds better, so it’s best to use each driver in their low distortion range. That means I’d select a mid that had a frequency response that could be crossed above where the tweeter distortion starts to rise. It’s more "matching" to create a low distortion speaker than selecting just a low distortion or flat FR driver.
The reason we want to test at a relatively high SPL is two fold, first, to ensure that the harmonic information is well within the measurement capabilities, remember we are talking <-40dB or <1% of the fundamental in many cases. The second is to actually put some stress on the driver, get more than 0.1mm of excursion to make sure that the motor and suspension system is being excersiced. Ideally, distortion sweeps over several SPL is ideal to see how the driver copes with a dynamic music signal.
If we look at the chart below, we see that a generally low background noise is around 30dB. That means that if we want to capture a harmonic that is 40dB below the fundamental, the fundamental must be greater than 70dB. Greater than 70dB means that our lowest discernable distortion will be 1%. What about 0.1%, or 60dB. Well that will require a greater than 90dB output to discern above the 30dB of background noise.
Another thing to point out is that most harmonic measuring systems are just peak detection, which really brings out the worst case in the noise floor. You could probably drop the noise floor a good 6-12dB in the measurement by simply taking an average of multiple measurements.
To your original question of weighing frequency response vs distortion, you don't really want to favour one over the other, every speaker is a result of compromises. A non-linear frequency response on its own doesn't mean much to me, because it's more about how easily that response can be shaped to what I have in mind. Same goes for distortion products, analyze each harmonic over frequency and decide what works best for what you have in mind. If it were my choice of a driver with poor frequency response but excellent distortion products, or a driver with great frequency response but mediocre distortion products, I'd take the one with better distortion products. You can EQ the crap out of a bad FR and make it better, but you cant remove distortion generated by the driver.
I would argue that additional capabilities for a great mid are transparency/detail as well as strong dynamic capabilities.
Anecdotally, I've never heard a driver with naturally flat response sound better or worse than a driver without once it is integrated into a loudspeaker. On the other hand, low distortion drivers integrated into loudspeakers always sound better to me that mid to high distortion drivers. This assumes in all cases the xo is competently implemented.
Dan
I am going to use my pair of SB15CAC60-8 as mids in a big 3 way. Before I go to all the work of building those big enclosures I want to hear what the mids sound like. So yesterday I started repurposing a pair of 12 L enclosures. I will be machining the new baffles today. If they sound good as a 2 way I will have something to demo at InDIYana too
As others have said, you can and will shape the FR with the xo, but you can't change a driver's distortion.
Other than that, I also land where you and dcibel do with regards to driver selection.
Consider the contrast... a crossover that is simulated using factory traced graphs but the designer screws up and forgets to account for baffle step loss or diffraction. Or the factory graphs are taken at different distances or drive levels. That would be an incompetently implemented design, right?
Lower distortion sounds better at normal listening levels. I found that when listening to horn loaded and non-horn loaded drivers.
IMHO equalizing sharp little dips is a waste of time and money. You can't hear the sharp dips (peaks yes, dips no). More parts mean more chance of part tolerance build up creating differences between the speakers.
One reason we test louder than we normally listen is to find the bottlenecks. Truth be told, I generally listen around 75db on average - but that means on certain music I can have peaks close to 90db. 90db sustained will definitely show the uglies in a driver - but those sharp transients take a lot longer to figure out what is wrong. Distortion testing at higher volumes helps us understand a drivers limitations, which leads to making more informed decisions.
Another aspect of distortion testing that I often do, but rarely pontificate on, is doing multiple loud sweeps for several continuous minutes. If a driver has poorly designed voice coil cooling, you can see distortion creep up. In really poorly designed drivers, the second or third sweep will start showing thermal related distortion.
In both cases - the single loud sweep and the multiple loud sweeps - you discover some of the drivers character. A driver can use copper in the motor and shorting rings and neodymium magnets but absent a place for that heat to go (and even at only several watts, heat is generated - touch the tip of a 5W soldering iron some time), performance will suffer.
The flat response thing is very important, as well. But there is a distinction to be made between "smooth" and "flat", as I have discovered over the years a slight spectral tiltdown from bass to treble is pleasing to my ears. As long as that tilt is devoid of significant ripples, anyways.
Flat for the sake of flat, however, is not an end goal - there are few designers who would argue that, but I will point out a few reasons why it should not be an end goal in and of itself.
1. It is often achieved by using a high Q filter on the tweeter to fill in the dip typically created on a flat baffle. This looks good in PCD, but it is essentially the same circuit used to make a bass boost on a subwoofer plate amp. Not wise, and it is probably the number one error I see in new guys who have discovered PCD and FRD/ZMA stuff.
2. It ignores off-axis (relevant to point 1 as well) in that those on-axis dips due to diffraction quite often fill out by moving off-axis even slightly. If you have ever wondered why some speakers with +/-1.5db on-axis can sound a little shrill or even fatiguing - it is most likely due to the designer not taking off-axis into consideration.
3. 4th LR lends itself to abuse via both of the above flaws since it is so easy to arrive at a goal that is of arguably minor importance - the reverse null. Long story short, going 3rd electrical on the tweeter is often what brings the null down, but it also fills in that dip due to the baffle.
More later - meeting time.