On Troels' Discovery 3WC-10, I'm unclear how the bass driver measurement shown can be real as it doesn't show any baffle step loss or expected diffraction?
10" 26W 4534 datasheet - nominal sensitivity 90.5db, SD 350cm^2
Troels' 26W measurement on baffle 320mm wide x 495mm high with no crossover. 100hz = approx 91db
I may not have done this right but I traced the SPL from the datasheet and used the diffraction and calculator tool for the baffle in vc and I get this. 100hz more like 86db.
He says his measurement is merged with nearfield response at 200hz, but can someone explain what's what?
The same issue arises with presentation of the 6" midrange measurements, too.
Wouldn't the system SPL tuning be completely off doing this? He claims system spl of 89-90db.
10" 26W 4534 datasheet - nominal sensitivity 90.5db, SD 350cm^2
Troels' 26W measurement on baffle 320mm wide x 495mm high with no crossover. 100hz = approx 91db
I may not have done this right but I traced the SPL from the datasheet and used the diffraction and calculator tool for the baffle in vc and I get this. 100hz more like 86db.
He says his measurement is merged with nearfield response at 200hz, but can someone explain what's what?
The same issue arises with presentation of the 6" midrange measurements, too.
Wouldn't the system SPL tuning be completely off doing this? He claims system spl of 89-90db.
Last edited:
The guy does not reveal everything. Looking at his drawings, Vb is about 47-50L and according to the impedance chart is tuned at 23-25Hz. That woofer would require about 90L for standard SBB4 tuning but he chose a higher sensitivity over bass extension.
TG (48L @ 25Hz)
Standard SBB4 (90L @ 25Hz)
TG (48L @ 25Hz)
Standard SBB4 (90L @ 25Hz)
Your data looks sensible enough to me, certainly in terms of the shape and the 6db approx. of diffraction loss Vituixcad has modelled.
If you splice in the low frequency response of a nearfield measurement of the woofer cone that will or should at least flatten off the region below 200 - 300Hz .
I haven't actually done this myself but in VituixCad tutorials there is a nice tool for splicing around 350Hz or so.
One small problem right now is if you don't have the driver yet and the baffle/cabinet assembled you cannot do this final bit to confirm to yourself how accurate it is.
I think you have a small discrepancy between your data and what was posted at the website. You are probably within an dB or so so maybe simulation or measurement discrepancies. Without input from Mr Gravesen to explain the actual data posted, difficult to say.
Typically Setup, calibration measurement, and distances used for measurement could all come into play here. Not forgetting room gain and the final placement of the speaker is going to change the actual LF response depending on your environment.
If you splice in the low frequency response of a nearfield measurement of the woofer cone that will or should at least flatten off the region below 200 - 300Hz .
I haven't actually done this myself but in VituixCad tutorials there is a nice tool for splicing around 350Hz or so.
One small problem right now is if you don't have the driver yet and the baffle/cabinet assembled you cannot do this final bit to confirm to yourself how accurate it is.
I think you have a small discrepancy between your data and what was posted at the website. You are probably within an dB or so so maybe simulation or measurement discrepancies. Without input from Mr Gravesen to explain the actual data posted, difficult to say.
Typically Setup, calibration measurement, and distances used for measurement could all come into play here. Not forgetting room gain and the final placement of the speaker is going to change the actual LF response depending on your environment.
I also noticed this new design from Troels recently, and I've also been puzzled by several aspects of it. If that woofer measurement is in the actual cabinet, which has a baffle measuring 32 cm by 70cm, it definitely ought to show baffle step loss. I have a pair of these Scanspeak 26W4534 woofers in a cabinet with almost the same baffle dimensions, and my measurements show the expected loss. (And merging a near-field LF measurement with a quasi-anechoic measurement shouldn't hide the baffle loss, surely?)
Also, I can't understand how Troels' design can have a system sensitivity of 90 dB, if there's baffle step loss. I reckon I get about 85-86 dB for an 8 ohm watt out my system, though that's a guestimate, not a calibrated measurement.
I'm also a bit puzzled by his box - I use my woofers in 55 litres sealed. That models well and my measured response is a good match for what I modelled. Yet he's got his in a box that can't be more than 40-45 litres, and he's got a vent. That really looks odd, when I model it. So I'm pretty perplexed all round. Troels has built a lot more speakers than me, though. (I guess that shadowplay62 is right, above - the vent in Troels' box isn't going to extend the bass much, but it will add output between 50 and 80 Hz. Not 4 or 5 dB though.)
Also, I can't understand how Troels' design can have a system sensitivity of 90 dB, if there's baffle step loss. I reckon I get about 85-86 dB for an 8 ohm watt out my system, though that's a guestimate, not a calibrated measurement.
I'm also a bit puzzled by his box - I use my woofers in 55 litres sealed. That models well and my measured response is a good match for what I modelled. Yet he's got his in a box that can't be more than 40-45 litres, and he's got a vent. That really looks odd, when I model it. So I'm pretty perplexed all round. Troels has built a lot more speakers than me, though. (I guess that shadowplay62 is right, above - the vent in Troels' box isn't going to extend the bass much, but it will add output between 50 and 80 Hz. Not 4 or 5 dB though.)
Last edited:
So, I’m not the only one that’s noticed some of his measurements seem questionable.
On Troels’ NS-1000 crossover upgrade page, his before and after response were clearly not taken at the same measurement location. Adding to that, the T&S parameters published there do not remotely match the modelled response for the stated enclosure volume.
The revised crossover response is not real, at least not at the listener distance and height. His page shows this ruler-flat response after dropping in the expensive crossover, and that is not possible without multiple notch filters (the only one is at 2.5-2.6kHz), or more than likely, very creative mic placement.
That’s not the worst of it, either. The whole premise of the crossover was it fixed a phase misalignment between the woofer and mid, and a dip at 4.5kHz. Now, that speaker was designed to be listened to half-way up the height of the baffle. It’s right in the manual. At mid-height, the tweeter and mid are time aligned. That exact dip which his “before” measurements show is also found in the manual when measuring the speaker too high along the front baffle.
I was able to recreate his blunder by measuring the speaker at 50cm distance, tweeter height. That is about 20 degrees off the proper vertical axis. Imagine designing a $700 crossover using the incorrect listening angle.
On Troels’ NS-1000 crossover upgrade page, his before and after response were clearly not taken at the same measurement location. Adding to that, the T&S parameters published there do not remotely match the modelled response for the stated enclosure volume.
The revised crossover response is not real, at least not at the listener distance and height. His page shows this ruler-flat response after dropping in the expensive crossover, and that is not possible without multiple notch filters (the only one is at 2.5-2.6kHz), or more than likely, very creative mic placement.
That’s not the worst of it, either. The whole premise of the crossover was it fixed a phase misalignment between the woofer and mid, and a dip at 4.5kHz. Now, that speaker was designed to be listened to half-way up the height of the baffle. It’s right in the manual. At mid-height, the tweeter and mid are time aligned. That exact dip which his “before” measurements show is also found in the manual when measuring the speaker too high along the front baffle.
I was able to recreate his blunder by measuring the speaker at 50cm distance, tweeter height. That is about 20 degrees off the proper vertical axis. Imagine designing a $700 crossover using the incorrect listening angle.
Creating 3 to 5 dB boost with enclosure or high Qts driver is well known tricks.
Using mechanical properties to help with full space losses is nothing new. And known
Studio Monitors are sitting on stands to make tweeter the listening position.
designing a crossover with the listening position at actual real world listening position.
If speakers are low to the ground, your way above the tweeter.
Did you design your crossover at the baffle? then in real world way above the tweeter or assume speakers are magical
Do you know the average height of a human standing ? Do you know average height human sitting.
Do you know average seating height of furniture. Did you design your speaker for a stand, TV stand
End table. Where will it actually be. And where will your ear actually be. sitting or standing?
Has anyone ever bothered to turn on the room in sim?
Did you sim your speaker position in your room to the wall distance.
Did you sim the floor bounce, did you raise or lower the floor.
to the height of your stand.
Did you model the extra Q from series resistance.
Can you add Q with impedance compensation.
Using mechanical properties to help with full space losses is nothing new. And known
Studio Monitors are sitting on stands to make tweeter the listening position.
designing a crossover with the listening position at actual real world listening position.
If speakers are low to the ground, your way above the tweeter.
Did you design your crossover at the baffle? then in real world way above the tweeter or assume speakers are magical
Do you know the average height of a human standing ? Do you know average height human sitting.
Do you know average seating height of furniture. Did you design your speaker for a stand, TV stand
End table. Where will it actually be. And where will your ear actually be. sitting or standing?
Has anyone ever bothered to turn on the room in sim?
Did you sim your speaker position in your room to the wall distance.
Did you sim the floor bounce, did you raise or lower the floor.
to the height of your stand.
Did you model the extra Q from series resistance.
Can you add Q with impedance compensation.
As far as I know Troels merges the higher frequency farfield response (which shows the baffle diffraction and some baffle step loss) to the low-end nearfield response of the woofer, so the total baffle step loss in the merged response (which we see above) is just an estimate.
Moreover as said above, smaller than theoretically optimal vented box yields to higher voltage sensitivity but more modest low-end extension.
Moreover as said above, smaller than theoretically optimal vented box yields to higher voltage sensitivity but more modest low-end extension.
Last edited:
Keyword: 'nearfield'. I could write paragraphs to avoid stepping on toes and not using the word 'fake' -oops! Too late. But a nearfield bass measurement is influenced by the box, which boosts the bass, so it's just not real*. It's probably fine anyway, since a large-ish box will probably be sitting on the floor and some reflecting walls in a real listening situation anyway.
*not a real representation of what you'd get at the listening position, which is almost certainly going to be a lot further away.
*not a real representation of what you'd get at the listening position, which is almost certainly going to be a lot further away.
Well, the 26W4534 isn't really a high Qts driver - published figure is 0.36, and mine measured 0.38. A 40 litre enclosure gives sealed system Q of around 0.7, so no boost there. Here's a Basta model of it in a 40L vented box with Troels' baffle size, vent tuned to 23 Hz (black trace) and my 55L sealed box (red trace). The vented box gives about 1.5dB extra at 50 Hz.
You can probably get another 1 or 2 dB from the crossover bandpass gain - so OK, you could get system sensitivity up to 89-90 dB, I guess. I suppose my question then would be about what the sound quality of this kind of bass alignment might be like. Maybe I'll try it some time.
On the subject of merging near-field LF measurements, something I really like about Vituixcad is that it has the ability to apply a modelled baffle step to near-field measurements, prior to splicing them with quasi-anechoic curves. So you can get a decent approximation of a full-range quasi-anechoic measurement.
Last edited:
then adds nearfield to the response and it makes full space losses look better.
so it is questionable.
probably involves politics of some kit buyers
that dont even understand full space
so many would critique the system as being weak
in the bass department
when it is just pretty normal for people that understand.
So the effect is opposite for people who understand full space.
then look at measurements and think...hmmm
to good to be true.
Seems like the shown measurements should have more peaks in the low end
visible in the datasheet.
So could be a gracious amount of smoothing in the merged low end as well
far as making a box small intentionally to raise Q
Depends. Can sound " bloomy" with some recordings
or actually be ok and make the bass seem more apparent.
this one likely ok, cause yes Qts is not high.
Bloomy examples more towards a high Qts driver that needs a big box.
but the designer refused to accept the larger box.
I dont like pushing Q more than 3 dB. And wont do it with high tuning
like some do.
So his box size, but low tuning is respectable and likely realistic
from experience
so it is questionable.
probably involves politics of some kit buyers
that dont even understand full space
so many would critique the system as being weak
in the bass department
when it is just pretty normal for people that understand.
So the effect is opposite for people who understand full space.
then look at measurements and think...hmmm
to good to be true.
Seems like the shown measurements should have more peaks in the low end
visible in the datasheet.
So could be a gracious amount of smoothing in the merged low end as well
far as making a box small intentionally to raise Q
Depends. Can sound " bloomy" with some recordings
or actually be ok and make the bass seem more apparent.
this one likely ok, cause yes Qts is not high.
Bloomy examples more towards a high Qts driver that needs a big box.
but the designer refused to accept the larger box.
I dont like pushing Q more than 3 dB. And wont do it with high tuning
like some do.
So his box size, but low tuning is respectable and likely realistic
from experience
Last edited:
Clearly I don't have any measurements and it's not easy to guess all the details (i.e. effect of stuffing, position of the L-shaped port) and also we do not know how he is performing his measurements.
I just noticed the difference of volume between the standard tuning and I agree with @WhiteDragon that he designed the box to be more appealing in the low end. All in all he has to sell stuff and I understand that, but I hope that clients interested in buying his designs obtain much more info (off axis / power response, x-over frequencies, etc.) than those published on his website.
Indeed, there's less information about this design than he's usually published in the past. It's asking a lot of potential buyers to cough up for all those high priced crossover parts, based on so little information. The cheapest crossover and accessory kit is 700 euros, not including the drivers. (That's actually more than the cost of the drivers.)
I don't understand why he doesn't have an option for a crossover kit using 'standard' parts. These are Scanspeak's most economical drivers, after all.
I don't understand why he doesn't have an option for a crossover kit using 'standard' parts. These are Scanspeak's most economical drivers, after all.
Well exactly there is some very good drivers out there that perform rather well.
and cost isnt that bad for drivers and performance very good.
problem is the old nickel n dimes really get fished out on enclosure wood.
and some, well many. really toss a lot of money at 400 watt coils
because series resistance is lower. for a system that might see 30 to 80 watts.
nothing wrong with MDF prototype boxes. actually easier to machine/paint/finish.
and buy cheaper crossover parts to hear the system.
then decide if you wanna build a plywood box or buy huge coils
and metal poly caps.
overall he is a very experienced designer and his woodworking skills
are rather darn good. wood and crossover component arguments
get beat to death. So over time as a builder you end up with a
collection of cheap and expensive coils/caps. and then real world
plays out.
Its the same politics. You start offering lower cost crossovers.
People second guess your designs. Or expect to see mystical
magical components. And might question a good designer.
Because of component choices
and cost isnt that bad for drivers and performance very good.
problem is the old nickel n dimes really get fished out on enclosure wood.
and some, well many. really toss a lot of money at 400 watt coils
because series resistance is lower. for a system that might see 30 to 80 watts.
nothing wrong with MDF prototype boxes. actually easier to machine/paint/finish.
and buy cheaper crossover parts to hear the system.
then decide if you wanna build a plywood box or buy huge coils
and metal poly caps.
overall he is a very experienced designer and his woodworking skills
are rather darn good. wood and crossover component arguments
get beat to death. So over time as a builder you end up with a
collection of cheap and expensive coils/caps. and then real world
plays out.
Its the same politics. You start offering lower cost crossovers.
People second guess your designs. Or expect to see mystical
magical components. And might question a good designer.
Because of component choices
Probably because approx 99% of the users wouldn't buy the boutique version and Jantzen Audio needs sales of these pricey parts. 🙂
Not all designers are doing baffle diffraction with near field measurements. And some are doing a hard splice instead of a blend. And this is the same as many reviewers or publications that show measurements. eg. Stereophile.
So, yes, it could be more correct, shall we say?
The nice thing is that Troels is designing for beginner level woodworkers and passive circuit board assemblers, providing lots of nice pictures of how to put it all together. And so a few simple measurements provides some idea of how it may sound. And he has many designs of all sizes for people to try out.
So, yes, it could be more correct, shall we say?
The nice thing is that Troels is designing for beginner level woodworkers and passive circuit board assemblers, providing lots of nice pictures of how to put it all together. And so a few simple measurements provides some idea of how it may sound. And he has many designs of all sizes for people to try out.
Last edited:
Yes, his site is a hugely useful resource. His older crossover designs are freely available, too, which is great.
A very interesting discussion.
Just like to add that I think the Scan-Speak Disco series are good drivers and I hope they keep supporting the DIY community.
Lets face it most of the component choices in commercial speakers are quite often several levels below what MR G uses in his kits. Also over at GR research they have a great time taking measurements without too much data describing the measurement setup and the smoothing used, and again sell some big and presumably more expensive parts.
I see it as similar to a bike or car customising shop, at the end of the day they have to make a living, and its up to the buyer to decide.
Back to the measurements depending on where the mic was and how close the woofer was to the floor or corner at the time of the measurement, and if the mic was on the bass axis or at the Tweeter/Mid axis I imagine you would have a very different response. Additionally as mentioned the actual Xover itself may give a bit of a rise in the bass.
Maybe some body could investigate this if they wanted to drive VituixCad and enter driver data. diffraction individual driver spacings and do it that way. I think it should be possible to reverse engineer his designs on an individual basis, apart from the Xover a lot of the hard work is done. Most decent wood shops should be able to cut work to his dimensions
Measurements to a laboratory standards takes hours and a lot of rigour in setup and preparation and an anechoic chamber would be nice.
Lets say a few hundred people here have the capability and necessary knowledge to take good quasi anechoic measurements, with mics they have calibrated every year? (Guilty of that one)
But after they have the fundamental measurements done and are chasing down component options and going to the MK1 earhole for final voicing, things may get a little less rigorous.
Post the initial measurements maybe those measurements can be a bit loose or taken on a quick look basis as verification they are heading in the right direction. Post design trying to put all the elements back together and provide good design evidence can be tricky. You definitely need to capture data throughout the design. Trying to do this some weeks downline and exactly remember what measurements and how it came about can be tricky. Just sayin.
At the end of the day more words and information would be helpful to reassure people who understand enough to question the info provided, but as I imagine that would be less than 10% of his clientele is it worth the effort on his part ?
As other have noted old designs have full detail and the design choices/options are certainly food for thought.
Just like to add that I think the Scan-Speak Disco series are good drivers and I hope they keep supporting the DIY community.
Lets face it most of the component choices in commercial speakers are quite often several levels below what MR G uses in his kits. Also over at GR research they have a great time taking measurements without too much data describing the measurement setup and the smoothing used, and again sell some big and presumably more expensive parts.
I see it as similar to a bike or car customising shop, at the end of the day they have to make a living, and its up to the buyer to decide.
Back to the measurements depending on where the mic was and how close the woofer was to the floor or corner at the time of the measurement, and if the mic was on the bass axis or at the Tweeter/Mid axis I imagine you would have a very different response. Additionally as mentioned the actual Xover itself may give a bit of a rise in the bass.
Maybe some body could investigate this if they wanted to drive VituixCad and enter driver data. diffraction individual driver spacings and do it that way. I think it should be possible to reverse engineer his designs on an individual basis, apart from the Xover a lot of the hard work is done. Most decent wood shops should be able to cut work to his dimensions
Measurements to a laboratory standards takes hours and a lot of rigour in setup and preparation and an anechoic chamber would be nice.
Lets say a few hundred people here have the capability and necessary knowledge to take good quasi anechoic measurements, with mics they have calibrated every year? (Guilty of that one)
But after they have the fundamental measurements done and are chasing down component options and going to the MK1 earhole for final voicing, things may get a little less rigorous.
Post the initial measurements maybe those measurements can be a bit loose or taken on a quick look basis as verification they are heading in the right direction. Post design trying to put all the elements back together and provide good design evidence can be tricky. You definitely need to capture data throughout the design. Trying to do this some weeks downline and exactly remember what measurements and how it came about can be tricky. Just sayin.
At the end of the day more words and information would be helpful to reassure people who understand enough to question the info provided, but as I imagine that would be less than 10% of his clientele is it worth the effort on his part ?
As other have noted old designs have full detail and the design choices/options are certainly food for thought.