Open baffle plan/layout

[JoshK]

is that waveguide the 1086.

[Saurav]

Yes, it's the 1086.

it'll be better than what you'll pick up not having the mid flush-mounted.

OK, I just woke up, and I can't parse that So you're saying I should flush-mount which driver, and why?

[fbov]

First off, congratulations - you made something, and learned something. I hope both continue.

Tool limitations are real; a router and circle jig is of great value when making certain things, and speakers are one of them. Same for table saws, but as you note, that can be contracted out - have someone else do it. So can just about every aspect of a customer speaker project: if you want professional results without tools and personal skill, find a cabinet maker. Speakers are trivial compared to custom furniture.

Tool limitations are also a chance to learn how to get around them, as you tried with the Dremel. How about this?
- Use the dremel to rough-out the recess, being careful about the radius
- put sandpaper on a disk and sand the recess to the desired depth
- use wood putty to fill in any remaining low spots.
The trick is using the right size disk to do the sanding. It should be supported by enough of the recess well enough that it sands like a flat surface. I used the piece I cut out inside the recess.

I figured this out when I was a little shallow cutting a tweeter recess, and my normal centerless circle routing workaround wouldn't work because I'd cut ears into the hole for terminals (RS28A). Rather than scallop the edge of the recess, I tried sanding and patience, and it worked!

Laminating a mask to the baffle is a similar idea; if you can't make a cut here, try adding a layer there. These may be workarounds, but when done well, they can contribute to the beauty and uniqueness, or perhaps be hidden in the finished piece.

HAve fun,
Frank

PS I just took a closer look and realized you used plywood! Scrap the Dremel at this point and start breaking the plywood at the ply interface that's the right depth for ht erecess. The Dremel gave you a good outer radius; a few shallow radial cuts and the bulk of the recess will clear with a screwdriver, if a wood chisel isn't handy. Then sand.

[Saurav]

Been working on this a little. I noticed that the largest bit on my dremel-router thingy looks like a 1/4" chamfer, while the largest roundover is 1/2 that size, so I've been going through and chamfering the edges of the baffles. The first time through I tried keeping the speed low, and was taking off too much, and really heated the dremel up. It seems to work much better if I keep the speed high, and make multiple passes. Makes sense, I guess, that little motor has RPM, but no torque. Anyway, as a result, the baffle edge will be closer to the WG than these photos, which shouldn't be a bad thing, though I doubt it'll make any significant difference to the diffraction profile. I'm mostly just doing this so the panels look like someone did some work on them

Thinking ahead... how do I stain/finish panels that will be visible on both sides? I'm thinking of using just L-brackets on the ground to start with, so I don't want to apply any finish while those are attached. I saw some suggestions of a board with nails sticking out of it, and you lay the work piece down on top of that. And if I try stain, I think I should do both speakers together, or I'll have a hard time getting the same shade on them?

And finally, I can't make up my mind if I should do something different for the top and bottom pieces, or just stain them the same. If I want to try different stains or go darker on one, is it OK to finish the panels before gluing them? I'm not sure wood glue will work if I put on a poly top-coat, but if I've just used stain/tung oil finish/something like that, can the panels be glued after that?

And finally... I found some wood glue and Minwax pre-stain conditioner in my garage. They have to be over a year old, since I can't remember having done any refinishing projects recently. Are those still usable, or should I ditch them and just buy new stuff?

[ThomasW]

Thinking ahead... how do I stain/finish panels that will be visible on both sides? I'm thinking of using just L-brackets on the ground to start with, so I don't want to apply any finish while those are attached. I saw some suggestions of a board with nails sticking out of it, and you lay the work piece down on top of that. And if I try stain, I think I should do both speakers together, or I'll have a hard time getting the same shade on them?

I hang stuff from the garage door tracks. That way it's easy to get at both sides.

And finally, I can't make up my mind if I should do something different for the top and bottom pieces, or just stain them the same. If I want to try different stains or go darker on one, is it OK to finish the panels before gluing them?

No because then you're only gluing to the finish not the wood

And finally... I found some wood glue and Minwax pre-stain conditioner in my garage. They have to be over a year old, since I can't remember having done any refinishing projects recently. Are those still usable, or should I ditch them and just buy new stuff?

Try them on a piece of scrap.

[Saurav]

Those are good ideas. Thanks.

[Saurav]

This is why I don't build speakers - I have no woodworking skills. It would do me good to remember that the next time I get a crazy idea like this.

I'm kinda pissed at myself, and also laughing - I glued one of the baffles crooked. I mean, that's the stupidest fricking mistake you can make, and I should have checked it before I let the glue get too dry. (Actually, I was surprised that the glue had set by the time I discovered this). And I should have known there was a higher chance of error since I was just gluing two overlapping panels together without a good frame of reference.

Oh well... if the speakers sound good, maybe I'll draw up some plans and get someone competent to build the baffles for me.

[Saurav]

Making some progress.



That's pre-stain conditioner, then 2 coats of mahogany gel stain (with some red polyshades mixed into the 2nd coat), followed by 4 coats of wipe-on satin poly so far. Hanging them from the garage door tracks worked well. You can see some unevenness in the stain, though the flash makes it look slightly worse

I'll probably do another coat or two of the poly on the fronts with the baffles laying flat, that should help it dry a little smoother than what it's doing now.

Birch ply isn't the best thing to start with if you want a dark stain.

[Saurav]

With drivers:



Not too bad for my first attempt at staining/finishing something like this

[augerpro]

Looks good saurav! The black drivers work well with the dark cherry finish.

[ThomasW]

Not too bad for my first attempt at staining/finishing something like this

Give that man a cigar.... :T

[Saurav]

I think I'll use this as my 'project' thread going forward, since I have questions scattered around 3 different threads now.

Anyway... making some progress with the active crossover portion. I'm using LTSpice for the simulations, and I bought this crossover board from Michael Price:

http://pricem.mit.edu/projects/newxover/

I like this board, it's pretty modular so I can pick and choose which circuit 'blocks' I want to use. Right now I'm just using this for EQ, and my Behringer for the crossover. Eventually I'll have all functions on this board.

The woofer EQ circuit so far:

Image not available

That's a shelving low-pass filter from 80Hz to 500Hz to counter the dipole rolloff, and a notch filter around 750Hz to remove the dipole peak. Bob Ellis' "Active Filter Four" page has a helpful calculator for the notch filter, and the shelving filter equations are on Linkwitz's site. Michael has also been doing a lot of hand-holding and helping me out with the math and navigating the schematic. I'm using the components I have in my stash, so the values are slightly off from the ideal calculated values.

Transfer function, calculated by LTSpice (blue) and measured by Speaker Workshop (red):

Image not available

To measure the EQ circuit, I just took the soundcard's output to the filter, and then took the filter's output back to the soundcard. The first time I forgot to turn off the mic calibration, so I got some funny results

Responses. Blue is calculated - the woofer's dipole response was calculated from the nearfield response, and then the transfer function was applied. Red is an ungated/farfield ground plane measurement of the woofer with the EQ applied. Both charts are smoothed to 1/6 octave, and levels adjusted to line them up.

Image not available

So far the measurements match the calculations fairly well, which is reassuring. The next step is to add the notch filter to the midrange, and then pick a crossover point. I've also run out of opamps (I can't get them de-soldered off my old XO as cleanly as I'd hoped), so this'll wait for a bit until I get new opamps.

Here's the predicted response with a 500Hz LR4 low-pass applied to the EQ'd woofer response:

Image not available

Also spent some time listening to a few CDs (in mono). I'll definitely need a subwoofer, this does not go low enough for me. But the bass (in the garage) is cleaner than what I had before in my living room. The midrange-tweeter XO still needs work. There's some sibilance I'd like to remove, the rising top octave needs to be tamed (though I'll wait on that until I measure this with my tube amp in place, since I expect that'll roll off the top end a bit).

This is a 2m ungated measurement of the full speaker, smoothed to 1/16 octave:

Image not available

The dip at the woofer-midrange XO is for 2 reasons, I think - I haven't done the midrange EQ yet, and the Behringer's LR4 rolloff is too much, the midrange response is already dropping around 500Hz so I'll need a shallower electrical slope to meet my target LR4 acoustic slope. Which has to wait until I get my opamps.

[Saurav]

Found the reason for the 500Hz dip. For some reason, I thought I had done the math and decided that the wavelength at 500Hz is a lot larger than my woofer-mid distance. But it's not, the wavelength at 500Hz is a little over 2', which is the distance between those drives. With the mic 6' away and at the midrange driver's height, the path length difference works out to about exactly half a wavelength, so I get a nice cancellation notch right at the crossover frequency.

Yesterday I played with the crossovers some more, made both sides match the 500Hz LR4 target fairly well, and still had the dip. I even checked with reversing the woofer connection, and the dip went away. I was going to write to AE and make sure I had the woofer terminal polarity right, and then realized what's going on.

The right solution is probably a larger midrange driver, and a lower crossover (which is something I want to do as an upgrade/evolution somewhere down the line). But for now, I'm not sure what I should do. I'm not sure keeping the woofer phase inverted is a reasonable 'solution'. I shouldn't do anything until I get these in the living room and measure from the actual listening distance, but I'm not sure what do do about this.

[cjd]

If reversing the woofer solves it, it's probably NOT a wavelength issue but an actual phase issue. In which case that is the right solution... (or reversing the mid/tweet)

C

[Saurav]

I guess I should check at different listening distances to be sure. I'm not sure how it can be a phase issue - the acoustic center would have to be pretty far off to make a difference at 500Hz, I would think? As far as I know I haven't set up any inverting stages in my EQ, and other than that everything is the same in the signal path to the woofer and mid/tweet (using the same crossover, and 2 channels of the A500 amp).

Maybe there are other mechanisms that could affect phase that I'm not aware of / thinking about.

[cjd]

Learn from it... don't fight it. But yeah, understanding if you have a mixed stage somewhere is probably a good thing.

C

[Saurav]

I used to think that the way to check that is the impulse response. But sometimes that's hard to do - a low-pass crossover removes the sharp rising edge, and sometimes I've seen traces where the largest peak is positive, but before that I have a pretty large negative peak, so... is that going up or down?

I was conscious of the polarity thing when I had the tube amp and the gainclone driving the old speakers. I'm pretty sure both should be non-inverting, but it never hurts to check.

Well, either way, I'll be glad if I can fix it with reversing the woofer. And now that I think about it, I already ran an informal test at a different listening distance - when I was listening to the speaker, I was further away, and sitting on a box so lower down. The reversed connection sounded fuller through the low midrange, especially on male vocals. So you're probably right about it being a phase thing.

[Dennis H]

You can check driver polarity with a 1.5V battery. + should move the cone forward.

PS, I only get a path difference of about 1/3 foot. sqrt(6^2 + 2^2) - 6

[Saurav]

Yeah, that's what I get for trying to do square roots in my head at 2am after getting up to feed the baby. "40... close enough to 49, so square root is 7". Nope, 40 is closer to 36 than to 49.

[Saurav]

Image not available

2m ungated measurement from today, smoothed to 1/16th octave. Woofer phase is reversed, so the slight dip from 500-1kHz in the previous measurement is gone. I also adjusted the tweeter padding to bring it up a little (compare above 1.5kHz with yesterday's measurement). This makes female vocals sound a little less recessed, but now cymbals are too bright. Jon Marsh had a small inductor in line with the tweeter in the Isiris crossover, presumably to deal with that rising top end (the 4540ND response is pretty similar to the 4538 that I'm using). I don't have anything in my stash that's a low enough value, so I'll have to pick up something to try here. The values I tried rolled off the treble too much.

The tweeter L-pad values that I'd picked based on simulating this in SW resulted in the tweeter level being visibly (and audibly) low. I'm measuring from slightly further away now than when I'd done the individual driver measurements, and I wonder if that's the cause. I need to take individual driver and crossover measurements and check them against Speaker Workshop to see how well they compare.

Anyway... the next task on the list is to do the EQ for the dipole peak on the midrange, and then work on the woofer-mid crossover. I took 1m measurements with the mic centered on the midrange to get a better feel for what's going on there. Black is gated, blue is ungated/smoothed.

Image not available

According to LTSpice that looks fairly simple to notch out, and then the crossover can be closer to an ideal LR4.

[Mark K]

Hi Saurav,

Looks pretty nice.

Are your ungated full range measurements in room? I've found that in room ungated measurements are generally not very good. Outdoors, or in a large room, a smoothed ungated measurement can be used.

[Saurav]

It's in the garage, which is larger than my listening room, but you're right, outside would be better. Below about 400Hz it looks like it starts to get noisier. I'll drag them outside once I'm closer to being done. I also need to put the tube amp that'll be driving these into the mix, and see what that does to the FR.

[Mark K]

Hi Saurav,

I use my garage too for "quickie" measurements. I agree that somewhere in there you should drag everything outside. It's very difficult to really see what's happening under 1k without going outside.


The image below is an outdoor comparison of a gated and ungated (IIRC ~1/6 or 1/12 octave smoothed.) measurements. This kind of closeness never occurs indoors. There are reflections, but the magnitude is so much lower. The extreme low end is suspect as the ground and side of a house start acting as ground planes (and wind noise also usually has the effect of raising the low frequency part of the curve.)

Still, your project looks pretty good and is moving right along...

[Hank]

Saurav, for ultimate brace stiffness, why don't you use square steel tubing? Two pieces, one on each side of the woofer in the lower panel and one piece in place of the wood brace in your drawing. Attach with construction adhesive and screws. This could be an item of experimentation when you do your MDF prototypes.

[Saurav]

That's an interesting idea, I hadn't thought about that.

[augerpro]

So how is the project coming saurav?

[Saurav]

It's been in my main system since I stopped updating this thread, and I've been listening to it since then

I'll get back to it in a couple of months... have a wedding this month, and family visiting. But basically, the XO I'm running with is my rough design, built with whatever parts I had on hand or could pick up at the local surplus place. I didn't do any measurements after I built the XO and put the speakers in the main system (i.e. with the tube amp in place). After a few weeks of listening I did reduce the tweeter level a bit from what I'd calculated from the driver measurements and crossover sims.

The active XO I'm pretty happy with, the EQ transfer functions landed where I want them to, and the measured response between those drivers looks OK. The passive mid-tweet XO needs more investigation.

Subjectively, I like the sound a lot. The bass is much clearer than what I'm used to with the sealed woofers I had in this room before. The treble is a lot smoother too, but that's probably got more to do with a better crossover job than any inherent CD/ribbon differences.

[augerpro]

What kind of eq are you using for the bass? How is the bass >40hz?

[Saurav]

I hand off to a sealed 1259 subwoofer at 60-80Hz (still experimenting). I wasn't going to try pushing this that low. For unknown reasons, the dipole woofer seems to (subjectively) integrate better with the subwoofer than my previous speakers. This is most evident with walking bass lines, where individual notes are clearer and more balanced, where earlier I had some emphasis on some notes while others were missing.

The woofer EQ details are in post 75 in this thread. It's a notch filter plus a shelving low-pass. The first handles the dipole peak, the second handles the rolloff. I only implemented a shelving filter till 80Hz... I guess I could go to 40, I haven't really looked at how much excursion that would take. Based on my measurements in the garage, I think the floor bounce starts to help around 50Hz or below, so I might not need the shelving filter below that if I tried to use a lower sub crossover.