Note: This discussion revolves solely around the drive unit used in the Kef HTS3001SE. For speaker (drive unit+egg enclosure) testing and data, click this link.
I recently ordered a set of the Kef HTS3001SE speaker. The reason wasn’t to have a great set of computer speakers. Rather, I was interested after reading Zaph’s test of the Q100 driver how much that performance was captured in the smaller 3001SE driver for potential DIY use. While a set of the HTS3001SE can be upwards of $400, if you consider you’re getting 4 drivers, that brings the cost down to about $100/piece with the benefit of a coaxial design. Still, that’s not exactly cheap and the results need to be viewed to determine relative worth. The Q100 driver is noted as a 6 inch driver, whereas the HTS3001SE is a bit smaller. My measurements (discussed below) show it to be closer in size to a typical 5.25 inch driver. Zaph’s data on the Q100 driver shows a whole lot of potential and, as far as coaxial/coincident drivers go, it’s the best I’ve seen. The Q100 features the Z-surround which my guess was to keep the surround out of the wave of the tweeter as much as possible yielding a smooth high frequency response both on and off-axis. If you look at coaxials from other manufacturers data you’ll find disparaging HF response; especially on-axis. Like I said, I was really curious how well the “little brother” – the HTS3001SE – would perform. Unfortunately there wasn’t any data on this and I knew I’d have to measure it myself if I wanted to know. So, I ordered a set to test and they showed up today.
Started off with the full Egg…
Luckily, dissecting them was extremely easy and not destructive. I dismantled the egg simply by removing the rubber facing which revealed two allen heads. Consequently enough, Kef includes an allen wrench for purpose of positioning the Egg on the swivel mount. I removed the two allen heads and the shell came apart with ease. This revealed the driver attached to the front half and the back half which contained the port (what feels like plastic), the crossover, and some acoustical stuffing. I disconnected the speaker leads from the driver. This left the driver attached to the front portion via 4 screws. I removed those screws, applied a small amount of force to ‘break’ the driver free of the mount and voila… raw driver.
As you can see above, the driver itself has a plastic frame and basket and (seemingly) a neodymium motor. The driver features a ribbed cone. Upon further investigation, the cone and the ribs seem to be one molded plastic piece rather than a paper/aluminum cone overlaid on a plastic ribbed frame. I would assume the rib design is implemented to prevent modal issues higher in frequency outside of the woofer driver’s passband and hopefully testing will show us how well that worked, though there is no way to compare what the results would be had the cone not been ribbed since I don’t have a version of this. We’ll just have to guess. Also, the driver uses the ‘tangerine’ waveguide found in some of the other Uni-Q drivers. When the cone is traveling forward and rearward with the voice coil, the tweeter assembly stays fixed. The cone itself is cut to allow the tweeter assembly to fit inside the driver, therefore there is no contact of the cone with the tweeter mechanism. This implies you would expect strict control of the suspension in order to keep from the cone touching the tweeter and, judging from the suspension measurements illustrated in the Woofer LSI section below, this is in fact the case.
Raw Driver Physical Measurements
First off, given this isn’t sold as an individual driver, I have taken my own measurements. These are rough measurements taken with my not-so-recently calibrated calipers, but should be good within +/-1mm.
|Outer Diameter Including Tabs||144.8||mm|
|Effective Piston Diameter**||102||mm|
|Effective Piston Diameter***||70.5||mm|
|Mounting Tab Thickness||0.65||mm|
|* Note this driver’s mounting diameter is extremely close to the outer diameter excluding tabs. Look at the picture of the driver and you’ll see the tabs are what allow you to have any bit to a baffle at all.|
|**Half surround to half surround; including space consumed by coincident tweeter.|
|***Half surround to half surround; NOT including space consumed by coincident tweeter.|
|****Flange thickness is not uniform. At the mounting location the flange is a bit thicker.|
As noted, the flange isn’t uniformly thick. The mounting tabs are a tad bit thicker. To make sure this point is understood, I have taken a few pictures to illustrate:
Another thing worth noting to anyone who may want to order these to dissect and use the drivers themselves is this: front mounting these may be more trouble than its worth. There’s a circle, but there’s not much meat here. The tabs provide mounting privileges. The wire lead blocks also stick out a decent bit so you’ll have to notch out your mounting ring. Which is probably why Kef rear mounts the driver in the egg. This may be the easiest way to go depending on your plans.
Final pictures, featuring the Aura Sound NS3 (left), the Kef HTS3001SE driver (center), and Scan Speak 12MU (right) for comparison purposes. As you can see, the Kef 3001 driver is a bit larger. I measured the Scan’s effective diameter to be about 84mm while the Kef’s is approximately 102mm. Again, ‘effective diameter’ is half-surround to half-surround across the driver. So, really, this particular Kef driver is much closer in size to a 5.25 inch driver.
Driver & Speaker Testing
Now, on to the testing. This will be broken up in to 3 parts:
- Tweeter Only (Raw Driver: no passive used)
- Woofer Only (Raw Driver: no passive used)
- Coaxial set playing with Kef’s passive crossover in use
Part I: Tweeter Testing
Tweeter Thiele-Small and Impedance Results
|Re||2.91||Ohm||electrical voice coil resistance at DC|
|Krm||0.0299||Ohm||WRIGHT inductance model|
|Erm||0.22||WRIGHT inductance model|
|Kxm||0.0001||Ohm||WRIGHT inductance model|
|Exm||0.86||WRIGHT inductance model|
|Cmes||147||µF||electrical capacitance representing moving mass|
|Lces||0.06||mH||electrical inductance representing driver compliance|
|Res||1.2||Ohm||resistance due to mechanical losses|
|fs||1757.4||Hz||driver resonance frequency|
|Qtp||1.41||total Q-factor considering all losses|
|Qms||1.945||mechanical Q-factor of driver in free air considering Rms only|
|Qes||4.738||electrical Q-factor of driver in free air considering Re only|
|Qts||1.379||total Q-factor considering Re and Rms only|
Tweeter Frequency Response
Measurements taken in the neafield and farfield, then merged together at approximately 1200hz, representing a quasi-anechoic response at 2.83v/1m. On-axis and off-axis measurements are represented at 0, 30, and 60 degrees.
Tweeter Harmonic Distortion
Measurement taken in the nearfield at SPL equal to 96dB/1m.
Obtained from the same nearfield conditions as HD above is used. In this case, however, I have set a floor equal to that of the lowest in-band response level minus 40dB. Which means, at approximately 6250, the response of the tweeter was lowest at 107dB. I set the floor for 67dB. This chops off any data that might not be needed. It is common convention to do do this as the thought is ringing audibility is only -40dB within 3 milliseconds. Personally, that’s about the extent of my knowledge here but to those who find this useful, I have provided it.
Part II: Woofer Testing
Woofer Thiele-Small and Impedance Results
|Re||3.21||Ohm||electrical voice coil resistance at DC|
|K||0.002||LEACH inductance model|
|n||0.77||LEACH inductance model|
|Cmes||478||µF||electrical capacitance representing moving mass|
|Lces||6.98||mH||electrical inductance representing driver compliance|
|Res||18.79||Ohm||resistance due to mechanical losses|
|fs||87.1||Hz||driver resonance frequency|
|fm||61.8||Hz||resonance frequency of driver with additional mass|
|(using add. mass)|
|Mms||11.986||g||mechanical mass of driver diaphragm assembly including air load and voice coil|
|Mmd (Sd)||11.655||g||mechanical mass of voice coil and diaphragm without air load|
|Rms||1.334||kg/s||mechanical resistance of total-driver losses|
|Cms||0.278||mm/N||mechanical compliance of driver suspension|
|Kms||3.59||N/mm||mechanical stiffness of driver suspension|
|Bl||5.006||N/A||force factor (Bl product)|
|Qtp||0.735||total Q-factor considering all losses|
|Qms||4.919||mechanical Q-factor of driver in free air considering Rms only|
|Qes||0.841||electrical Q-factor of driver in free air considering Re only|
|Qts||0.718||total Q-factor considering Re and Rms only|
|Vas||0.7692||l||equivalent air volume of suspension|
|n0||0.058||%||reference efficiency (2 pi-radiation using Re)|
|Lm||79.84||dB||characteristic sound pressure level (SPL at 1m for 1W @ Re)|
|Lnom||80.8||dB||nominal sensitivity (SPL at 1m for 1W @ Zn)|
Woofer Large Signal Analysis (LSI) Results
|Displacement Limits||thresholds can be changed in Processing property page|
|X Bl @ Bl min=82%||2.7||mm||Displacement limit due to force factor variation|
|X C @ C min=75%||3.6||mm||Displacement limit due to compliance variation|
|X L @ Z max=10 %||>3.9||mm||Displacement limit due to inductance variation|
|X d @ d2=10%||12.7||mm||Displacement limit due to IM distortion (Doppler)|
|Asymmetry (IEC 62458)|
|Ak||-3||%||Stiffness asymmetry Ak(Xpeak)|
|Xsym||0.34||mm||Symmetry point of Bl(x) at maximal excursion|
Woofer Frequency Response
Measurements taken in the neafield and farfield, then merged together at approximately 500hz, representing a quasi-anechoic response at 2.83v/1m. On-axis and off-axis measurements are represented at 0, 30, and 60 degrees.
Woofer Harmonic Distortion
Measurement taken in the nearfield at SPL equal to 96dB/1m.
Obtained from the same nearfield conditions as HD above is used. In this case, however, I have set a floor equal to that of the lowest in-band+some response level minus 40dB. Which means, at approximately 10khz, the response of the woofer was lowest at 100dB. I set the floor for 70dB. This chops off any data that might not be needed. It is common convention to do do this as the thought is ringing audibility is only -40dB within 3 milliseconds. Personally, that’s about the extent of my knowledge here but to those who find this useful, I have provided it.
Finally, here is the result of the Tweeter’s on-axis response overlaid on the same graph with the Woofer’s on-axis response. This is raw driver data; no passive crossover used.
Part III: Coaxial Speaker Testing Using Kef Passive Crossover
Up next is testing the HTS3001SE speaker on the baffle using the supplied passive crossover.
This test was done to simulate the equivalent level of 96dB/1m. This usually involves testing at 112dB in the nearfield. Since there’s a dip in response in the 1-2khz area, likely due to the passive using baffle step compensation here, thus making the lower passband of the woofer have more output, I used the average output from 100hz-10khz and increased the output until I achieved approximately 112dB here, though I wound up closer to 115dB. You’ll see what I mean below. Just look at the dashed line labeled Fundamental Mean. That line represents the mean output level which is approximately 115dB.
Coaxial Driver CSD
Taken in the nearfield with a floor set 40dB below the lowest dip in response below 10khz, which occurred at about 110dB.
Individual Driver Frequency Response with Respective Crossover
This is the result of testing each driver with it’s respective passive crossover implemented.
This may be a bit scattered, so forgive me.
Raw driver response:
- The woofer is comparable to 5.xx” drivers in cone area.
- The measured Fs is 87hz, with a Qts of 0.718.
- The linear xmax, as measured by Klippel’s LSI, came in at 2.7mm; limited by Bl. This is the xmax at 10% THD. Suspension derived linear xmax clocked in at 3.6mm.
- As noted in the opening discussion about build quality, the cone is designed to move around the tweeter assembly and there’s really not much wiggle room here. This means the suspension should not have any tilt or asymmetry while in motion. Otherwise the cone would crash against the tweeter housing. And it doesn’t. The most obvious way to tell this is by throwing some power at the driver and listening for any contact. Then, if you want to really see just what the suspension is doing, you can do so by analyzing the suspension performance measured by Klippel’s Large Signal Identification (LSI) module provided in the woofer results section. As shown in the Kms Symmetry Range graph in this section, suspension symmetry is nearly ruler flat. The delta from the driver at rest (0mm) to peak measured excursion (~4.4mm) is about 0.23mm which is roughly the thickness of a folded sheet of paper. This is textbook symmetry for a suspension system. So, by this standard, the driver’s design and build quality are spot on. Below is a video of the woofer’s movement to show how the tweeter assembly stays static while the woofer cone moves around it.
CAUTION: You may want to turn down your volume.
- The woofer exhibits a sharp drop in response just above 2.2khz, as evidenced in all the response charts. Thinking this may be a function of the hole in the center of the woofer cone (to pass beyond the tweeter) and stem from a phase anomaly. My first inclination was to investigate via phase comparison of the driver tested at two different levels (90dB/1m vs 96dB/1m). I thought that if I drove the woofer further in travel it might show a more distinct alteration in the measured response in this area. However, I found no obvious evidence with this method. All phase and magnitude plots between the two drive levels lined up almost perfectly. Though, it is clear to see the minimal phase takes a notable dip as shown below.
- Frequency Response:
- The average SPL at 2.83v/1m is 83dB from 100hz to 2khz. Above 2khz, the response diverges pretty rapidly and falls off pretty hard until 4khz which is roughly where the -3dB point is from 0 to 30 degrees. Overall, if you can determine exactly what the cause of disturbance here is and/or you have the DSP to adjust the response above 2khz (assuming this is possible) you might be able to eek out more usable bandwidth.
- 2nd order dominates here and rides 1% distortion from about 200hz to 1.5khz. Above this, there’s significant increase in 2nd order then alternating with 3rd order starting at 2.7khz. Though, 3rd order is staying around 0.25% from about 150hz to 2khz (with a small blip at 250hz rising to just under 1%).Even order distortion follows the same profile as the FR above 2khz which could be telling… I just don’t know what it’s telling.
- While there the inductance over excursion plot doesn’t show a symmetrical Le(x) curve, the overall inductance delta represented via Le(x) from -4.4mm to +4.4mm is less than 0.08mH, which indicates possible use of inductance mitigation.
- The tangerine waveguide, as it’s called, looks pretty interesting. Frankly, I’ve not found any white papers or technical documents on the design so I’ll hold my subjective and likely wrong thoughts on this. But, at the least, it’s an interesting means of controlling directivity and, compared with other coincident/coaxial offerings on the market, it seems to do a pretty good job.
- Frequency Response:
- There definitely appears to be a horn loading effect. Off Axis response seems to look pretty well behaved, though the on-axis response shows potential issues caused by a null at 8khz. I did some listening and didn’t notice this to stand out but it may require attention should you be interested in using this driver for your own design. This may be an inherent trait of this type of design as I notice it occurs in the Seas Coaxial offerings as well. Could just be coincidence, though.
- Harmonic Distortion:
- As with the woofer, the HD results of the tweeter is dominated almost solely by 2nd order distortion components. 3rd and subsequent order distortion is almost non-existent below 2khz.
Combined driver response with passive crossover:
Remember, I am measuring a coaxial driver with the supplied crossover intended to be used in an enclosure on a raw baffle.
- It looks very much like there’s baffle step compensation here, evidenced by the boosted response below 1khz. On the test baffle, the boosted low end shows itself. There’s hearty low end here. In the speaker egg, this shouldn’t be as much of concern.
- Of course, the top end has a boost as well which mean a trough in response between 1khz to 2khz. I thought maybe I had one of the drivers out of phase but testing the polarity the other way yielded results that showed me that obviously wasn’t the issue. Furthermore, looking at the individual driver results with supplied passive crossover yield very similar results for the woofer driver. Someone with speaker building experience may be better able to explain this. This could be intentional design to have this dip in response. I have yet to have a chance to measure the speaker itself but hopefully I’ll get to soon.
That’s it. I’ve tried to cover what I can, but given the breatdh of this test, I’m bound to have missed noting something or possibly mislabeled a graph. If you notice anything like that, let me know and I’ll correct it.
If you have any specific questions or you have feedback on the performance of this driver, feel free to post to my page or check discussion here.
If you like what you see here and what to help me out, there’s a Paypal Contribute button at the bottom of each page. Every little bit does help.