Here I'm presenting some of my own measurements of the KEF Q900. All measurements are in-room, made with a cheapo Dayton EMM6 and ARTA software, with the laptop soundcard serving as the I/O. The signal is sine sweep, output sampled at 44.1 kHz.
Pics:
1. Impulse response at 1m, on axis.
2. On axis gated measurement at 1m. Clearly the tilt in the HF is just too much.
3. On axis dual-gated at 1m. Dual gated means the HF is gated at 5ms (200 Hz) and below 200 Hz, the gate is at 200 ms. ARTA then joins the two responses. This is smoothed at 1/3rd octave, just to see the overall trend. Bass and midbass looks about right to me.
4. Off axis, dual-gated at 1m. This is about 30 deg off axis. It comes back to near flat, but the balance is still bright.
5. Steady state response at the listening position. Perceptively, a gently downward sloping HF in this measurement sounds flat to me.
I bought these thinking they could serve as a long term reference, as a sanity check for my own creations. But they simply don't sound right to me.
Questions:
1. Is there something wrong with my measurement setup?
2. Why is the treble tilted up so much?
3. Could this be a defective pair?
4. Is this what audiophiles are listening to?
Pics:
1. Impulse response at 1m, on axis.
2. On axis gated measurement at 1m. Clearly the tilt in the HF is just too much.
3. On axis dual-gated at 1m. Dual gated means the HF is gated at 5ms (200 Hz) and below 200 Hz, the gate is at 200 ms. ARTA then joins the two responses. This is smoothed at 1/3rd octave, just to see the overall trend. Bass and midbass looks about right to me.
4. Off axis, dual-gated at 1m. This is about 30 deg off axis. It comes back to near flat, but the balance is still bright.
5. Steady state response at the listening position. Perceptively, a gently downward sloping HF in this measurement sounds flat to me.
I bought these thinking they could serve as a long term reference, as a sanity check for my own creations. But they simply don't sound right to me.
Questions:
1. Is there something wrong with my measurement setup?
2. Why is the treble tilted up so much?
3. Could this be a defective pair?
4. Is this what audiophiles are listening to?
Attachments
I wonder if the distance from the lower woofer to the top one is causing a cancellation tendency around 1k. You may have to average a range above or below the tweeter axis (might get lucky at a single point below tweeter axis and a little further back... and to examine this I'd use RTA with noise and sweep the mic up/down in front of the speaker).
You'll have to do this for each horizontal measurement to the level of your wanted resolution. I like the trend with some of the later plots, hard to tell just how much is the room but I wonder about distance from the corner.
The trend either side of 1k, I'd agree is not what I normally go for, rather more flat until after 1k then roll off a little.
BTW, the tweeter (waveguided) would seem hotter near on-axis than the power response might suggest (ie. high D.I.)
You'll have to do this for each horizontal measurement to the level of your wanted resolution. I like the trend with some of the later plots, hard to tell just how much is the room but I wonder about distance from the corner.
The trend either side of 1k, I'd agree is not what I normally go for, rather more flat until after 1k then roll off a little.
BTW, the tweeter (waveguided) would seem hotter near on-axis than the power response might suggest (ie. high D.I.)
Last edited:
The room is kind of large, but the speakers are off to the left side along the longer wall that goes into the kitchen.
I see what you mean about the DI and power response of the tweeter. I would just shelve it down a few db to achieve my desired power response. If only they had provided a treble tone control knob!
I actually prefer the steady state response of the last graph below 1k. Subjectively, it sounds flat or even a tad bass heavy (but I prefer it). But above 3k or so, I want to see a drooping response.
I see what you mean about the DI and power response of the tweeter. I would just shelve it down a few db to achieve my desired power response. If only they had provided a treble tone control knob!
I actually prefer the steady state response of the last graph below 1k. Subjectively, it sounds flat or even a tad bass heavy (but I prefer it). But above 3k or so, I want to see a drooping response.
Look here for other measurement
KEF Q900 loudspeaker Measurements | Stereophile.com
Same trend in the low, not that much of an upwards rise.
vac
KEF Q900 loudspeaker Measurements | Stereophile.com
Same trend in the low, not that much of an upwards rise.
vac
Yes, I've seen that and the HT Mag measurements. In fact, I bought them based on those measurements. But you have to read very carefully what they are showing. They show the spatially averaged response, which comes closer to flat (still a small rise). They never show the pure on axis response. Neither do they show the power response in the room. Both these measurements are shown sometimes for other speakers. Look at the Quad 2805 measurements, for example.
Thanks for that vacuphile. It seems I might have been looking at some of this the wrong way based on the sites I looked at. Must go to sleep now though. Later 😉
Look at fig 7 and what JA says about it here:
JBL Synthesis 1400 Array BG loudspeaker Measurements | Stereophile.com
He warns against a potentially hot treble because of the flat power response.
JBL Synthesis 1400 Array BG loudspeaker Measurements | Stereophile.com
He warns against a potentially hot treble because of the flat power response.
Crossover
Here's some crossover curves. The knob on the back allows decoupling the tweeter from the woofer array. The measurement setup is the same as before. The woofer response is dual-gated.
This pretty much agrees with JA's measurements, except he shows spatially averaged responses even for the crossover graph.
He also notes that the slopes are shallow. The tweeter looks ok to me (apart from the rising response), but the woofer certainly could use a sharper roll-off, and maybe a notch as well.
Here's some crossover curves. The knob on the back allows decoupling the tweeter from the woofer array. The measurement setup is the same as before. The woofer response is dual-gated.
This pretty much agrees with JA's measurements, except he shows spatially averaged responses even for the crossover graph.
He also notes that the slopes are shallow. The tweeter looks ok to me (apart from the rising response), but the woofer certainly could use a sharper roll-off, and maybe a notch as well.
Attachments
Last edited:
Thanks for the data! Though I'm curious to see what the listening position response would look like with the speakers toed in 45deg.
One thing to consider is that the Dayton calibration may not be that great above 10kHz or so. See here for a comparison of calibrations for my EMM-6 (factory, Cross Spectrum Labs on-axis, CSL grazing).
Also, are you doing your listening-position measurements at grazing incidence, or with the mike pointed straight ahead? IMO, grazing gives a picture that more closely resembles what I think I hear.
I wouldn't worry about the steepness of the rolloff, but IMO a notch filter for the midwoofer's breakup is not optional. That's the one area where I think KEF dropped the ball. I don't hear it as brightness, but rather fatigue over a decently-long listen (say, DSCH 7 at live-like levels.)
One thing to consider about these things is that most of the budget seems to have gone into the drive units. The crossover is at least a notch filter away from excellent, and the cabinet quality/finishing is indifferent. But nothing else near the price has drive-units that sophisticated, so for someone who can take measurements and model crossovers IMO they can't be beat for the price.
Thanks guys!
I tried them with EQ: a simple shelving filter of -6 db at 1 kHz, with a Q of 1.0. This flattens the on axis response. So much better! They have become very listenable. Still require some toe-in for good imaging. Currently toed in to cross about a foot or two in front of the listening position.
I would have hoped that KEF had figured this out better, or at least provided some kind of control on the treble setting. Some of you are suggesting listening extreme off axis, but the brightness persists even such a setup, at least to my ears.
I'm still pondering whether to send them back or not. Guess I have some more listening to do.
I tried them with EQ: a simple shelving filter of -6 db at 1 kHz, with a Q of 1.0. This flattens the on axis response. So much better! They have become very listenable. Still require some toe-in for good imaging. Currently toed in to cross about a foot or two in front of the listening position.
I would have hoped that KEF had figured this out better, or at least provided some kind of control on the treble setting. Some of you are suggesting listening extreme off axis, but the brightness persists even such a setup, at least to my ears.
I'm still pondering whether to send them back or not. Guess I have some more listening to do.
Oh, I did it using MiniDSP. Hardly a permanent solution. A better solution would be modding the crossover. A resistor in parallel with a coil can achieve this.
Looking at Jeff Bagby's spreadsheet, it looks like a 4.5 ohm resistor parallel to a 0.5mh coil will be required.
I'm not yet sure if I'm send these back, so I don't want to open them yet.
Looking at Jeff Bagby's spreadsheet, it looks like a 4.5 ohm resistor parallel to a 0.5mh coil will be required.
I'm not yet sure if I'm send these back, so I don't want to open them yet.
Yes but his example shown in figure 7 of your linked article suggests that the right room/placement should give you a reasonable acoustic result to work with.
Using DSP on the whole signal will EQ whilst retaining the original speaker character. Modding the crossover may change the speaker character, hopefully for the better but will want measurements to ensure that phase isn't thrown out with these filters.
Definitely too bright. Much too bright, based on your measurements.
Unlikely. 1/2" capsule microphones will typically have an upwards slope from about 8Khz up to 15Khz of up to 3-4dB if they're not calibrated, but what you're measuring is an upwards slope all the way from 500Hz. Even a cheap ECM8000 un-calibrated is accurate from 50Hz up to about 6Khz. So I'm sure the overall upwards trend through the midrange and treble really is there, especially if it sounds too bright to you.
Good question. Doubtful that its a fault, (why would both measure the same if so ? I assume you measured both...) I suspect the answer is simply that they were designed with misguided notions of what the target response should be - somebody thought that jacking up the top end to give a more flat power response was a good idea, but it isn't.
Just because its a good brand like KEF doesn't mean that they don't make the odd clunker with poor design choices. 😛
Possibly quite a few. Another one that seems to often have tilted up treble is B&W. I've spent quite a lot of time (years ago now) listening to the B&W Nautilus 802 (original version not diamond) and while I liked some aspects of their performance, I was not happy with what I perceived to be a tilted up upper end treble which sounded a bit brittle and over bright to me.
Years later I read the stereophile review, and sure enough there is the tilted up treble in the measurements...
I too have noticed that some of the stereophile reviews have only a 30 degree spatially averaged response and not an on axis response, and I think its disingenuous of them to say the least.
The cynic in me would have to say that they've done that to make speakers with a rising on axis response such as the Q900 look better on paper, if you don't read the fine print you won't even realise, and technically they're not lying.
They're just providing measurements that don't mean much and don't correlate to what we hear. (How does a 30 degree averaged window correspond to what we hear at high frequencies when our head is in one location at a time 😕 )
Quite possibly the Q900 was designed with the same philosophy, that a 30 degree averaged window is more meaningful than an on axis measurement, unfortunately that idea is misguided IMHO.
A 30 degree averaged window is useful to help sort and separate driver resonances from diffraction effects (diffraction effects will even out, resonances will add together) but for determining tonal balance its wrong, wrong, wrong. 🙄
Flat on axis response through the treble and falling power response is what sounds balanced, the only real question is how much should the power response fall towards the top end so that it doesn't become over-bright in highly reflective rooms...
Last edited:
Thank you, Simon! I pretty much agree with everything you've said. Stereophile's measurements were quite misleading in this review at least.
Flat power sounds bright, no matter how what axis you're listening to. Why KEF designed it this way? Could be like you say that they were going for listening window average, or it could be that this is the sound the market prefers or has gotten used to.
Flat power sounds bright, no matter how what axis you're listening to. Why KEF designed it this way? Could be like you say that they were going for listening window average, or it could be that this is the sound the market prefers or has gotten used to.
Allen, I cannot agree that room furnishings are going to change the balance at 1 kHz. They might have some impact above 5 kHz, but even then, the tilt in this case is too much.
DSP is a good solution, but hardly a permanent one. I don't want to have to plug the DSP into the signal chain everytime I listen to them. And you're right. Putting in crossover components will not be easy. It will likely be a totally new crossover.
That looks about what I heard in a store. I ultimately preferred some cheaper JBLs that had a more laid back sound. Audition SPL was quite high. Either woofer breakup or tweeter brightness made me dislike this speaker.
Looks like an easy modification to really good speakers. I agree with Pallas that the driver are probably very good.
Looks like an easy modification to really good speakers. I agree with Pallas that the driver are probably very good.
It's not just the overall upwards tilt that would make it sound bright and uncomfortable, that 6dB peak between 5-6Khz in the on axis measurement (image 2) would sound harsh even if the overall trend was balanced.
It's unclear whether that's a tweeter resonance, diffraction from the tweeter, or whether it's cone breakup from the woofer showing through due to lack of a notch and steep enough slope, or perhaps a combination of both woofer and tweeter responses, but I know from past experience that that type of response problem wouldn't sound good at higher SPL...
Last edited:
- Status
- Not open for further replies.