Location: Seaford UK
Posts: 1,861
I'm Dennis.
What you have done seems to me to be quite an achievement without any formal training, I have had none in computers or computing, and my Win 10 PC is such an awkward and malfunctional thing that I am dissuaded from pursuing computer measurement, but I'm sure that it would be a major aid. If you saw the pages of tables of calculation I did, to get a value for a series L for the mid/woofer, you would probably laugh. But the Xover seems inaudible, can't hear a peak or dip.
My approach is to 'nudge' the curves up to each other with small changes in Xover values, but mine was only two way, and three is much harder to do.
If you measured distortion, you must have some pretty sophisticated equipment for that.
Despite years of training my ears, I think it is very difficult to actually correlate measurements with what one hears. We used at the Beeb to train ourselves to recognise a frequency with spot checks, ie., use a sig genny and choose a frequency, and ask a 'blind' subject to say what that frequency was. After a while we got close, and we also used to see what the minimum level change that we could detect was, the 'wisdom' at that time was 1dB at 1KHz, but I proved I could detect 1/4dB. Of course my ears now at 71 are showing deterioration - partially caused by years of listening to loud music no doubt.
The ear/brain is a very sensitive device, which is why we all hear differences and agonise over them, but when it comes to assessing what is going on in a speaker by ear, even knowing the appropriate band to look at when trying to correct errors is hard. A few years ago in a discussion with Bill Woodman of ATC, I asserted that most cymbal energy was from tweeters, to which he replied that it was actually in the mid range.
I don't know what you mean by 'model' the impedance curves, just seeing the graphs would be of major help.
I agree about the units, KEF were the first to really use computer measurement to control consistency.
There are those Martin whoh think that there should be a gradual slope downwards towards the top end
Location: Glasgow
Posts: 109
I'm Richard.
That's a good point, Martin, but I was thinking more about the balance I have between the upper bass frequencies, mids, and high frequencies, which seems to have very little downward slope.
In recent times folk have been recommending a -10dB differential to provide a natural tonal balance for domestic listening, i.e. a consistent roll-off as frequency increases so that the highest frequencies are -10dB down from the lowest frequencies.
I can see at least one significant limitation to this: where you set the start and end points for such a target slope. e.g. A system that only has a bandwidth of 50Hz to 20kHz will sound duller with a -10dB target slope than one that has a bandwidth of 20Hz to 20kHz. My 50Hz to 20kHz bandwidth has a slope of -5dB, but the expanded 20Hz to 20kHz bandwidth has a slope of -10dB. Also, I have a broadish dip in my response between 80Hz-300Hz which could be argued to contribute to an unnatural leanness in the upper bass region, though I personally don't mind such a response (in fact in lieu of a linear response I find a dip in this area much preferable to an elevation!).
Older research by Bruel & Kjaer suggests a slope between 20Hz and 20kHz of -6dB to be optimal (see Fig.5 on page 4).
I reckon a slope that's between -6dB and -10dB from bass to treble without excessive scribbling in between should provide a natural tonal balance, but at the end of the day it also comes down to what the individual listener prefers.
I should note that my measured response was with the grilles off. Putting the grilles on the speakers knocks an average of -2dB off everything above 4kHz, which tips the presentation into sounding too dull/veiled IMO. The grille also affects certain frequencies more than others (it has a metal grid structure which I suspect causes considerable diffraction), another reason why I listen with the grilles off. It's a shame as I think these speakers look better with the grilles on, and they also help reduce my anxiety over dust getting stuck to the tacky surface of the MD500 dome, which would be a PITA to remove!
Last edited by ToTo Man; 02-10-2019 at 17:02.
Main System: Mac Mini > Schiit Yggdrasil 2 OG > Yamaha A-S3000 > Celestion Ditton 66
Headphone System: Mac Mini > Schiit Yggdrasil 2 OG > Sparkos Aries / Schiit Mjolnir 2 > Sennheiser HD600
I agree one size is never going to fit all. There's personal preference in terms of how toppy or bassy you want it.
Current Lash Up:
TEAC VRDS 701T > Sony TAE1000ESD > Krell KSA50S > JM Labs Focal Electra 926.
Location: Glasgow
Posts: 109
I'm Richard.
I can't help you with Windows I'm afraid, Dennis, having been a lifelong Mac user. The last time I used a PC, Windows XP was still the most widely used OS!
Quite the opposite actually. All I use is a miniDSP UMIK-1 USB microphone and RoomEQ Wizard software. The software displays a plethora of graphical data from a single frequency response sweep. I don't know how accurate/reliable the distortion data it produces is, but it certainly helped me to weed out mid and high frequency drivers that had abnormal resonances.
The human ear is indeed an incredibly sensitive device and is sensitive to changes that are often barely detectable in measurements, but I also agree that it is very easy to confuse/doubt oneself as to what it is that has changed, which is why measurements can be helpful in providing direction and/or validation.
That's probably because I've used incorrect terminology! I mean measuring the impedance curves, using a system like DATS. Apparently, due to the ageing process and hand-made nature of the Celestion drive units there can be significant intersample impedance curve variations between otherwise seemingly identical units, variations that only manifest as significant frequency response differences when the drivers are actually connected to the crossover (when measured in isolation and without a crossover they often have near identical frequency responses, which can be misleading).
Main System: Mac Mini > Schiit Yggdrasil 2 OG > Yamaha A-S3000 > Celestion Ditton 66
Headphone System: Mac Mini > Schiit Yggdrasil 2 OG > Sparkos Aries / Schiit Mjolnir 2 > Sennheiser HD600
Location: Glasgow
Posts: 109
I'm Richard.
Main System: Mac Mini > Schiit Yggdrasil 2 OG > Yamaha A-S3000 > Celestion Ditton 66
Headphone System: Mac Mini > Schiit Yggdrasil 2 OG > Sparkos Aries / Schiit Mjolnir 2 > Sennheiser HD600
Location: Seaford UK
Posts: 1,861
I'm Dennis.
The downward slope idea is not an established and universal criterion. It is touted by some designers, but many speakers are still designed to be, and quoted as, flat, mine are 22Hz to 50kHZ +/- 3dB, so must be flat. (with wiggles)
I was a Mac user from'91 onwards, SE30, (which died in '05) but as the price increased and I became poorer I adopted the wretched PC school, going through 4 PCs and '95, 98, XP, Win7 and now 10, I think XP was the most stable and sensible.
I've just bought a second hand Mac running Ojave. Lovely, but I am not yet very familiar, and I tend to 'save it for best'.
Yes in those early days driver irregularities were often offset with tweeked Xovers.
Most speakers are designed to be flat anechoically, so that once you put them in a typical room you get boundary gain and HF gets soaked up so you get a tilted response downwards from bottom to top. If they measure flat in a typical room they'll probably sound a bit toppy. I do wonder about the people who find modern speakers too bright and how sparsely furnished their rooms are - wood floors, hardly any furniture, nothing on the walls and so on. Might be the reason.
Current Lash Up:
TEAC VRDS 701T > Sony TAE1000ESD > Krell KSA50S > JM Labs Focal Electra 926.