Mark K's Speaker Pages
...when you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind; it may be the beginning of knowledge, but you have scarcely in your thoughts advanced to the state of Science...Lord Kelvin
RS28a nonferrofluid tweeter evaluation
Drivers purchased 11/08, evaluation completed 12/08
The RS28a tweeter lost it's ferrofluid recently. It's unclear exactly when. You can open your driver and look without too much difficulty. If you're interested, search the PE forum for threads on the tweeter changes, and the official PE response. Chris B sent me four new nonferrofluid RS28a tweeters to look at so we could assess the low frequency handling. And I also promised him I would redesign my RS28A if he would send me the four tweeters to look at.
The RS28a, though a technically excellent design, had been hampered by some QC issues early on. I was curious to see if the newest versions were consistent and what kind of performance could be expected.
Without further ado, let's get to the measurements.
Below is a set of frequency response measurements of all four drivers on my standard baffle at one meter.
The corresponding impedance curves
Well, that's enough linear distortion for me...on to the nonlinear distortion. There are a whole bunch of graphs to follow. What you will see is the graphs grouped by frequency, 1.3k, 1.5k, 2k, 2.5k, 3k, and 5k. Now, only unit A has all of these. As you'll see, the consistency is so close, I started skipping frequencies for the other drivers. While this may make the data a little less robust, given the overall results, I'll convince you at the end why this was probably justified.
Anyway, let's start out with the graphs at 1.3k.
OK. Before you get lost in the barrage of graphs to follow, take a minute to absorb the graphs above. The FR curves are statistically identical. The impedance curves are very close, and the nonlinear graphs are also statistically remarkably close.
What's so interesting is that this is very much unlike the behavior of previous units I've looked at. Look at the variability of some of the prior units I've tested here.
Now you might note there is more second order distortion. Noted. And I'll throw out some comparison grapsh with the Seas 27TBFC/G, Peerless HDS, SS6600. But for now note the overall low level of higher order products and consistency compared with the prior units. These four units are statistically indistinguishable.
OK, how about a look at 1500 Hz?
Again, very consistent at 1.5k. Let's move on to 2k.
Again, persistence of somewhat more than average second order distortion, but otherwise very clean patterns.
For 3k and 5k I just looked at unit A.
Again, fairly clean.
The higher second order products bothers me a bit, as all my tests done recently in an unheated garage show more second order products for a wide variety of drivers. Can cold temps be to blame for some increased suspension asymmetry. Well, let's find out. We'll warm up one RS28a unit, and keep the other in the garage. Also, I've got the Seas 27TBFC/G and the SS6600 cooling down in the garage. While they are cooling, let me post the following graphs.
NOTE: I'm breaking my own rule with the following graphs so take them with a grain of salt. That is, the SS6600, HDS, and Seas are comparisons from a year or two ago under different test conditions. This is technically a bit sloppy, but nonetheless useful. However, it takes a little time for the SS6600 and Seas to cool down, and one of the RS28's to warm up. While we're waiting, let's take a look at the comparison graphs.
OK, here's a busy one. The SS6600 is in the black. At 2k, it's clearly superior to the RS and Seas. The RS28 comes out just marginally worse.
OK, how about a comparison with the HDS at 2k? It looks to be a statistical dead heat.
OK, now with the Seas at 1.5k.
Marginally worse second and third order. Let's flip the curves. Now the RS28A is in black. Any extra red is extra Seas distortion. Very close.
Now remember, these are comparisons from SS6600, HDS, and Seas data done months ago in warmer weather. Are these really comparable? Let's see.
After an overnight cooldown in the garage for the SS6600 and Seas, as well as a warmup for an RS28a, let's get to retesting.
But first, a picture to sooth your retina which has, so far, been seared by too many graphs. You can see the design is pretty much the same. Robust, copper, and the felt looks to be placed with a high degree of accuracy. There is a little enlarged cutaway in the lower right hand corner. Now if you look close at the cutaway, the spacers sure look like rectangles. Right? As I recall, there used to be blobs of silicone. A precut spacer with adhesive would make the alignment better. A small modification, but it looks like it works better. DLR's the resident expert here, so I'll defer to him.
After an overnight chill, let's look at a couple of graphs.
It's hard to pick this up on the graph below, but the RS28A that was kept at 68 degrees seemed to test a bit better than the 40 degree one. However, unit A was the one kept at 68 degrees was marginally better anyway. So, maybe a slight effect, but not quite what I was expecting.
OK. Well, let's try something different. Don't laugh, but I got out the hair dryer and heated up both the drivers and the mic. I did this a couple times. Sure enough, heating the driver and mic up to just over 100-110 degrees (that's pretty much summertime testing temperatures in my garage) gave more of an effect. I could now get the RS28A second order products to about -33.5dB. Below in red is a cold mic and driver vs the heated up units in black. (-30 vs -33.4dB)
So what does this tell us?
There is a small effect on the absolute values, though it's not dramatic.
Second, trying to compare multiple graphs done under different conditions is suspect, or at least has to be interpreted cautiously.
Third, my practice of always using a known driver and retesting it under the exact same conditions/same sitting seems justified.
So, what about the RS28a? Well, one more graph-let's just heat everything up...
Well look at that. The RS28a is still marginally behind the Seas 27TBFC/G on second order, but, and it's an important but, the RS28a is an overall cleaner driver higher up. This is no easy feat. And really, given the motor construction, the way it should be. The RS28a has a much more advanced motor. It's the equivalent to a VTEC dual overhead cam vs Seas' OHV set up. The problem with the RS28a is that the execution has been poor previously, leading to some really mediocre performing units. Just look at some of my older tests.
This current batch of four RS28a's is back up to a very high standard and the consistency is outstanding. If the folks at PE can maintain this high level of consistency then this becomes one of the better performing tweeters, regardless of price. Still, it remains to be seen, based on past performance, that they can all test this good. I'll purchase a pair in 6 months and retest them. If they do well, I may swap out my Seas 27TBFC/G for a pair of RS28a's in my dipole, since, well, I'm sure I still won't be finished...