This is my first post, and also my first attempt at a speaker design, so please excuse any stupidities that may follow. I designed a TQWP in hornresp around the Tang Band W8-1808 using a "folded voigt pipe" design. The internal dimensions of the main box is 95 x 35 x 25 cm. The divider is 82 cm and placed with a 10 degree angle against the bottom. The port is 4 x 20 x 25 cm.
Here's what hornresp thought that would look like with input parameters (hornresp record attached):
I used a UMM-6 michrophone with REW and did my best to follow the instructions over at ASR to create quasi-anechoic measurement (this is with 12db smoothing, see attached REW file):
That curve is... pretty nasty. Given that this is my first attempt, I'm not too disappointed, but it would be nice to get it closer to what I was modelling from the start. Here's a few things I'm thinking/wondering about:
* Sure the 1808 has a rising frequency response naturally, but the design seem to have made it worse, contrary to what hornresp though. Er... Why? There seems to be a pretty long decay (50ms+) in the measurements, could that be the cause?
* That dip at 70 hz, I'm assuming that's a cancellation?
Any ideas what I can try to fix the slope and or cancellation? I am going to DSP these if I end up using them, which should help a bit but... 🤷♂️ Any constructive feedback is wellcome!
Here's what hornresp thought that would look like with input parameters (hornresp record attached):
I used a UMM-6 michrophone with REW and did my best to follow the instructions over at ASR to create quasi-anechoic measurement (this is with 12db smoothing, see attached REW file):
That curve is... pretty nasty. Given that this is my first attempt, I'm not too disappointed, but it would be nice to get it closer to what I was modelling from the start. Here's a few things I'm thinking/wondering about:
* Sure the 1808 has a rising frequency response naturally, but the design seem to have made it worse, contrary to what hornresp though. Er... Why? There seems to be a pretty long decay (50ms+) in the measurements, could that be the cause?
* That dip at 70 hz, I'm assuming that's a cancellation?
Any ideas what I can try to fix the slope and or cancellation? I am going to DSP these if I end up using them, which should help a bit but... 🤷♂️ Any constructive feedback is wellcome!
Attachments
Last edited:
I suppose just looking at what's there in front of us, you might start tuning by ignoring the variations in the below 100Hz region because they are difficult to measure. Perhaps start with baffle compensation that also takes in the rising response, focussing above 100Hz. Then see what high frequency issues remain.
It really does depend on your tools and skills. One easy but effective way to approach this would be with an equaliser while you listen. After a while, compare the result to a measurement.. rationalise what you have, then go back to listening and tuning.
So with regards to the dip around 70Hz, did you measure very close to the cone? That might help identify a cabinet internal issue.
It really does depend on your tools and skills. One easy but effective way to approach this would be with an equaliser while you listen. After a while, compare the result to a measurement.. rationalise what you have, then go back to listening and tuning.
So with regards to the dip around 70Hz, did you measure very close to the cone? That might help identify a cabinet internal issue.
Well, Hornresp is actually showing the behaviour quite accurately -remember it's only going up to 1KHz & realistically speaking, it's accuracy for rear-loading LF modelling drops off above the driver's mass-corner. What you've got there is partly step-loss from the baffle, but as much the 1808's naturally rising HF response. TB's graphing is small, but shows roughly the same 10dB rising trend in your own (you'll never get them to line up completely). LF measurements < 200Hz even with quasi-anechoic are notoriously difficult so like the above, I wouldn't pay too much mind to that; the 70Hz though could be a 1/2 wave floor-bounce depending on measurement conditions. An LR shelving circuit, possibly with a zobel should flatten things out, but being ain 8in wideband, I wouldn't aim for a flat axial response or you'll probably find the power-response & HF suffer in practice at the listening position.
Yes, the port and woofer was measured nearfield, aligned, baffle step corrected (using the Jeff Bagby Diffraction and Boundary Simulator) and then gain matched with the 1 meter measurement.
I guess you could attempt to cross check this by laying the speaker on it's side on the ground and laying the mic on the ground away from the baffle and equidistant from the driver and port, forming roughly an equilateral triangle. (This removes the floor as a reflective surface and takes in both sources).
If this doesn't show an indication of the dip then I'd move on to more important things.
If this doesn't show an indication of the dip then I'd move on to more important things.
Oh, right, thanks: I forgot hornresp isn't modelling high frequerncies correctly. My bad.
The measurement was taken on the table as the speaker is depicted above. So... Mmm, the port is a lot closer to the floor than the driver. Do you think it is worth re-doing it outside (and preferrably a bit higher up) to see if it changes?
Thanks for the tip!
So yes, the 70 hz dip is gone. Nice!
How much do you think a speaker like this can be EQ'd / DSP'd?
If you set up the equalization correctly, the DSP works wonders.
For example, below is the correction of the frequency response of a wideband speaker.
That is a big question. Suffice it to say for now, that there are some room interactions you can successfully EQ and some you can't. If you run into difficulty we can try to work out what the underlying problem is.
Are your measurements properly gated?
I've played around with RaspberryPI / HifiBerryOS elsewere for DSP, and the original plan was to front these with a MiniDSP Flex.
The impulse reponse is really messy, which worries me a bit. I set the gate to 4.31 ms as that was the only place it looked anything like a reflection to me.
Can you look at the response while you slide the gate? You should be able to sort out gating before you measure.. that is to say you prepare a measurement space where you know when the first one is expected.
I built a mltl enclosure with the 1808 driver. After about 5k the output really climbs. I used a step filter with a .5 mh air core inducter with a 10 ohm resistor in parallel with the coil. It is now one of my favorite speakers. You need a large gauge coil to avoid raising the effective qts.
I tried using Equaliser APO with a quick and dirty REW DSP correction, over a full bandwith measurement in the room (including the 70hz cancellation):
Apart from the obvious squashing of the dynamic range that's looking better.
I'll try to re-measre outside to see if I can gate it better and get rid of the 70 hz cancellation.
Any other experiment you think I should run?
Apart from the obvious squashing of the dynamic range that's looking better.
I'll try to re-measre outside to see if I can gate it better and get rid of the 70 hz cancellation.
Any other experiment you think I should run?
This might be a n00b question: should I expect any perceivable difference between taming the high frequencies using cross over components, as opposed to using DSP?
Listening test after DSP: A bit more sibilants than I'm used to and prefer, but moving off-axis solves that (and I can always EQ that in the DSP). Otherwise, I would love just a few DB better bass extension, but I'll I guess I have to wait for the second speaker to be built and then on their placement in a "real" room before I can judge that since the port is located just off the floor.
But damn, just from a single speaker, this is really, really promising! Instrument separation is already good, the mid bass and downwards appears to me "faster" than I'm used to (whatever that means).
But damn, just from a single speaker, this is really, really promising! Instrument separation is already good, the mid bass and downwards appears to me "faster" than I'm used to (whatever that means).
Show the level of harmonic distortion, it is quite possible that sibilants are a consequence of the increased level of distortion in the mid-high frequency range.
Try to make correction of the frequency response from 300 Hz and higher, with psychoacoustic smoothing in REW. You need get the reference frequency response in the listening area.
Reduce the peak at 100 Hz by only half, there is no need to align it into a line.
Last edited: