Just a little intro post - I'm planning on picking this up in earnest in the new year, but wanted to get some of it written down while I can.
I picked up a used pair of the Fostex FE166En drivers on the usual auction site, with the intention of making a nice pair of small speakers for use as the main stereo in the living room (possibly with LF support), or as reference monitors for when I'm out and doing some recording. I want these to be DSP-free, which is a departure from my usual designs, where I've previously thrown 20+dB of LF boost at a FE126 to get it flat to 40Hz in a 2L enclosure.
I'm taking some inspiration from Morgan Jones' Arpeggio, but with a few tweaks, which I'll detail shortly.
First up, the box needs to be smaller. I'm going to move these to a different location once or twice per month, so I want them usefully small and easy to handle. I'll also be driving them from a solid-state amplifier. Usefully, those two go quite well - where Morgan had to account for a 6.5ohm output impedance (and it's effect on the total Q of the cabinet), we can drive the cabinet size down a bit and keep a similar LF rolloff shape, albeit at a higher frequency. No free lunch, after all.
So, lets have a look at Hornresp. I simulated the driver in a 10L box to see what's what.
Usefully, that's a Qtc of around 0.5 right off the bat. Adding a few ohms of series resistance means Qtc can go up to 0.7 or even higher. A Qtc of 0.9 gives a slight bump before rolloff and gives lower -3dB/-10dB points, with 8ohm series R.
Since we're going to be adding series resistance to play with the Qtc, a Zobel filter will be needed, which can be simulated easily enough with Hornresp. It's quite easy to get the impedance looking nice and flat into the HF. I'll be testing that once the speakers are built since Morgan arrived at very different values.
The oddball in the screenshots above is the 3rd-order sealed box. This adds a large (200uF) capacitor in series with the driver, which gives it a bump below Fc (system resonance) as the capacitor interacts with the impedance peak. More excursion is used around 70Hz, but the capacitor forms a highpass filter so excursion is reduced at very low frequencies.
The large capacitor would be a bipolar electrolytic, bypassed with some kind of film cap in the 4.7uF range to keep ESR down and appease my inner audiophile.
Since these speakers will be used in a large range of rooms and positions, I feel it's important to have an option that gives a bit more bass extension for when the room isn't going to help prop up the LF response.
A lot of attention is paid to diffraction in the Arpeggios, and I can see why. The addition of surface baffle treatments smooths the upper range out a lot, and that's something I plan on investigating further.
Like Morgan, I used The Edge.
... But with a difference. First up, a smaller baffle. Obviously.
I ran two simulations. Firstly with the whole cone radiating. Given that it's a large cone on a fairly small baffle, the upper response came out very smooth - all the distances from any part of the cone to the edge of the baffle are very blurred, so there's not much effect on the frequency response.
After that, I estimated the size of the whizzer cone to be 35mm, and entered that into the program. This gives much more interesting results. Above a couple of kHz, the whizzer cone is doing the majority of the work, so, in my opinion, we should pay particular attention to that.
It took some work, but I managed to get something that looked good. The large cone wasn't very sensitive to baffle placement, but the smaller radiating surface is, so that's the bit I aimed to optimise.
There's an overall need for baffle-step compensation. I'll be doing that at speaker-level, probably with one of two BSC circuits that can be selected or bypassed with a switch. There's also a 2dB hump around 1kHz which might need notching out. Up there, though, you've got to consider the frequency response of the driver itself so there'll be measurements to find out exactly what needs to be done.
Subwoofers are much easier since you're in the realm of T/S parameters and the simulations work well. Higher up, it's all about measurements. Still, the baffle placement looks good with the possibility of a little minor work being required.
Internally, I plan on making the surfaces as irregular as possible to reduce reflections. To test this, I'm going to send a series of impulse signals to the driver, and feed a microphone into an oscilloscope. I've seen this done before where the 'scope shows a series of "return" clicks, and from the timebase you can work out the distance each one has travelled and then treat that surface. I expect a combination of circular dowels and stuffing will help disperse the internal reflections nicely. Of course, the usual basket tricks recommended by Dave at Planet10 will be applied - damping any ringing from the frame as well as smoothing the join between the magnet and the frame are also expected to help.
So, a summary of the plan so far...
- FE166En in a bookshelf-sized sealed box
- Variable LF Q-factor, plus option to go 3rd order
- Variable BSC
- Reduce internal reflections as much as possible
- Investigate baffle surface treatments
That's all for now.
Season's greetings,
Chris
I picked up a used pair of the Fostex FE166En drivers on the usual auction site, with the intention of making a nice pair of small speakers for use as the main stereo in the living room (possibly with LF support), or as reference monitors for when I'm out and doing some recording. I want these to be DSP-free, which is a departure from my usual designs, where I've previously thrown 20+dB of LF boost at a FE126 to get it flat to 40Hz in a 2L enclosure.
I'm taking some inspiration from Morgan Jones' Arpeggio, but with a few tweaks, which I'll detail shortly.
First up, the box needs to be smaller. I'm going to move these to a different location once or twice per month, so I want them usefully small and easy to handle. I'll also be driving them from a solid-state amplifier. Usefully, those two go quite well - where Morgan had to account for a 6.5ohm output impedance (and it's effect on the total Q of the cabinet), we can drive the cabinet size down a bit and keep a similar LF rolloff shape, albeit at a higher frequency. No free lunch, after all.
So, lets have a look at Hornresp. I simulated the driver in a 10L box to see what's what.
Usefully, that's a Qtc of around 0.5 right off the bat. Adding a few ohms of series resistance means Qtc can go up to 0.7 or even higher. A Qtc of 0.9 gives a slight bump before rolloff and gives lower -3dB/-10dB points, with 8ohm series R.
Since we're going to be adding series resistance to play with the Qtc, a Zobel filter will be needed, which can be simulated easily enough with Hornresp. It's quite easy to get the impedance looking nice and flat into the HF. I'll be testing that once the speakers are built since Morgan arrived at very different values.
The oddball in the screenshots above is the 3rd-order sealed box. This adds a large (200uF) capacitor in series with the driver, which gives it a bump below Fc (system resonance) as the capacitor interacts with the impedance peak. More excursion is used around 70Hz, but the capacitor forms a highpass filter so excursion is reduced at very low frequencies.
The large capacitor would be a bipolar electrolytic, bypassed with some kind of film cap in the 4.7uF range to keep ESR down and appease my inner audiophile.
Since these speakers will be used in a large range of rooms and positions, I feel it's important to have an option that gives a bit more bass extension for when the room isn't going to help prop up the LF response.
A lot of attention is paid to diffraction in the Arpeggios, and I can see why. The addition of surface baffle treatments smooths the upper range out a lot, and that's something I plan on investigating further.
Like Morgan, I used The Edge.
... But with a difference. First up, a smaller baffle. Obviously.
I ran two simulations. Firstly with the whole cone radiating. Given that it's a large cone on a fairly small baffle, the upper response came out very smooth - all the distances from any part of the cone to the edge of the baffle are very blurred, so there's not much effect on the frequency response.
After that, I estimated the size of the whizzer cone to be 35mm, and entered that into the program. This gives much more interesting results. Above a couple of kHz, the whizzer cone is doing the majority of the work, so, in my opinion, we should pay particular attention to that.
It took some work, but I managed to get something that looked good. The large cone wasn't very sensitive to baffle placement, but the smaller radiating surface is, so that's the bit I aimed to optimise.
There's an overall need for baffle-step compensation. I'll be doing that at speaker-level, probably with one of two BSC circuits that can be selected or bypassed with a switch. There's also a 2dB hump around 1kHz which might need notching out. Up there, though, you've got to consider the frequency response of the driver itself so there'll be measurements to find out exactly what needs to be done.
Subwoofers are much easier since you're in the realm of T/S parameters and the simulations work well. Higher up, it's all about measurements. Still, the baffle placement looks good with the possibility of a little minor work being required.
Internally, I plan on making the surfaces as irregular as possible to reduce reflections. To test this, I'm going to send a series of impulse signals to the driver, and feed a microphone into an oscilloscope. I've seen this done before where the 'scope shows a series of "return" clicks, and from the timebase you can work out the distance each one has travelled and then treat that surface. I expect a combination of circular dowels and stuffing will help disperse the internal reflections nicely. Of course, the usual basket tricks recommended by Dave at Planet10 will be applied - damping any ringing from the frame as well as smoothing the join between the magnet and the frame are also expected to help.
So, a summary of the plan so far...
- FE166En in a bookshelf-sized sealed box
- Variable LF Q-factor, plus option to go 3rd order
- Variable BSC
- Reduce internal reflections as much as possible
- Investigate baffle surface treatments
That's all for now.
Season's greetings,
Chris
This is interesting... the use of a large cap in series to boost the bass response is not what I would have guessed. Thanks for sharing.
If you are looking for a very non-reflective back chamber, like an acoustic black hole, try the "Dagger". It's a tall (12in long) 3 sided pyramid with the base where the driver cutout is. Stuff progressively dense polyfill/fiberglass starting with loose to dense at the vertex. My measurements show essentially same performance reflection wise as an OB or IB.
For diffraction, rebate the bezel and flush mount driver - reduces ripple by -2dB to -5dB. Make baffle asymmetric like a trapezoid and that reduces another -3dB to -10dB. These are not combined - you can't get -15dB that is. Having large radius round-overs always helps.
Add fuzzy felt or Scotchbrite scouring pads around the driver bezel for several inches, and that smooths out 500Hz to 2kHz region for better imaging - maybe -2dB at key points.
Here is an example (you can hide all of this inside a traditional rectangular box if you wish):
http://www.diyaudio.com/forums/full...-rs225-8-fast-ref-monitor-50.html#post4348154
If you are looking for a very non-reflective back chamber, like an acoustic black hole, try the "Dagger". It's a tall (12in long) 3 sided pyramid with the base where the driver cutout is. Stuff progressively dense polyfill/fiberglass starting with loose to dense at the vertex. My measurements show essentially same performance reflection wise as an OB or IB.
For diffraction, rebate the bezel and flush mount driver - reduces ripple by -2dB to -5dB. Make baffle asymmetric like a trapezoid and that reduces another -3dB to -10dB. These are not combined - you can't get -15dB that is. Having large radius round-overs always helps.
Add fuzzy felt or Scotchbrite scouring pads around the driver bezel for several inches, and that smooths out 500Hz to 2kHz region for better imaging - maybe -2dB at key points.
Here is an example (you can hide all of this inside a traditional rectangular box if you wish):
http://www.diyaudio.com/forums/full...-rs225-8-fast-ref-monitor-50.html#post4348154
Last edited:
Just wondering if you have built this speaker. I also have a pair of FE166's and have been considering the Arpeggio design. Being able to extend the bass would be very useful.
- Status
- Not open for further replies.