Fenrir MLTL Build

BrannigansLaw

So as some of you know I've been slowly building a set of Wolf's Fenrir, slightly modified in that MLTL box Paul K designed for the R176p drivers. I am at the testing stage finally, and wanted to get a bit of feedback on whether my measurements look right. 

First - FR taken in-room, at 1m, with 5.5ms gating:

BrannigansLaw

These are 1/24 octave smoothed. One speaker seems to have slightly lower tweeter level than the other. That speaker had to be measured at ~10deg off axis due to room constraints, so I'm thinking I can chalk the difference up to that? The magnitude of the difference is ~1-2 db, biggest difference is 2.3db at 9khz. Can I get a sanity check on this please? Should I be concerned and adjust any crossover values? 

BrannigansLaw

Next, impedance sweep from DATS



This shows the tuning frequency; target was 32hz, I'm quite happy with this

BrannigansLaw

Any feedback would be most appreciated; I'm at the point where the next steps will make the crossover inaccessible so I'd like to make sure all my ducks are in a row!

dcibel

When taking a gated measurement, all the little wiggles through the mid and low frequencies are indicating a reflection has made it through the gate. Have a close look at your impulse response and set the gate width (time) accordingly. For indoor measurements with 8ft ceiling, placing the speaker at 4ft height, the minimum gate would be 4.75ms if measuring at 1m. Move the mic closer and the gate time can increase. For a far field measurement, I'd recommend measuring at 4x the cabinet width.

You loss in the high end is likely due to measuring off axis. If you want to compare the two speakers to see how well the match, set up the mic and measuring location and swap the speakers, that way all other influences or reflections that are included in the measurement remain constant, and you only see the difference between the two.

BrannigansLaw

dcibel said:

When taking a gated measurement, all the little wiggles through the mid and low frequencies are indicating a reflection has made it through the gate. Have a close look at your impulse response and set the gate width (time) accordingly. For indoor measurements with 8ft ceiling, placing the speaker at 4ft height, the minimum gate would be 4.75ms if measuring at 1m. Move the mic closer and the gate time can increase. For a far field measurement, I'd recommend measuring at 4x the cabinet width.

Thank you for this advice, I have reset my distance to 4xbaffle width (40" - not too far off 1m) and have redone measurement. Just want to make sure I get the impulse/gating right, Here's my Impulse with the gate at 4ms, which is where I see the first 'blip' in the response. It's pretty low down though, so I'm unsure if this is correct?

dcibel

The idea behind the 4x baffle width distance is that you need to be as close as possible to allow for as wide of a gate as possible, but far enough away to capture the diffraction effects. For smaller speakers this means that you can measure closer in room, and the larger the speaker the larger the room is needed for a clean indoor measurement.

Yes, the first blip after the initial impulse would be the "first reflection", so around 4ms is pretty normal. With 8ft ceiling height, the best thing to do is try to get the measurement height around 4ft to maximize the distance between speaker and floor and ceiling reflections. 

Not sure if you tried posting a screenshot of the IR, but I don't see it.

BrannigansLaw

Sorry about that, must have forgotten to attach:

dcibel

Hows the response look with the shorter window?

Also, how loud are you measuring? If you increase the volume a bit the reflections will be more easily discerned against the background noise.

BrannigansLaw

I think I may understand a bit better now. Based on this:

http://techtalk.parts-express.com/forum/tech-talk-forum/1369108-rew-gate-settings-help-please

I am now looking at the step response instead of IR. Step response looks like:


g
So I assume I want to set my gate at ~2.1ms instead. Doing that smooths out the response <1khz, so I take it that's a win? 

But, also makes me think that maybe measuring these at the LP is not going to be enough 

BrannigansLaw

Ok, so I moved things around a little. Here's some new measurements, this time measurement conditions are as follows:

40" distance, mic on tweeter axis
Each speaker in exact same location, mic didn't move
2.5ms gating (using step response to find first reflection as per above)

The measurements



On the plus side: the measurements look cleaner, thank dcibel for the suggestions! And the FR match looks much tighter, which is good news.  On the minus side, obviously I'm perplexed by the dip at 2.8 khz, it's a fairly big one. I tried the obvious and reversed the phase on the tweeter, here's that comparison:




A broader dip below the XO frequency and what looks to be better summation above it. 

dcibel

No, step response is not what you look at to find reflections. look at impulse response.

I asked you how loud you are measuring, as the IR beyond the initial impulse is rather noisy. If you are measuring at a very low level, increase the volume and you will see the reflections better.

Have a look at this spreadsheet, it may help:

http://audio.claub.net/software/DaveDalFarra/Micheight_v1.0.xls

BrannigansLaw

Thank I will look at the spreadsheet. I'm doing sweeps loud enough that I have to plug my ears, louder than I would listen normally. REW SPL Meter tool says 90db, but I understand that isn't reliable unless calibrated against an external meter. I'm using a calibrated UMIK from cross spectrum labs. I don't have an SPL meter on hand so I don't think I can give you an absolute value. 

dcibel

That's fine, that's much louder than you should need. For frequency response I usually measure around 75dB in a quiet room which is "talking loudly" volume. Anything nearby such as tables, chairs, etc will be reflected in the impulse, so just try and keep the area clear around the speaker to get a clean measurement and you're good to go.

Wolf can probably help you with the design and crossover if you suspect you have some error there, but your response in your first post, if the wiggles were removed looks similar to what Wolf posted for the Fenrir design, the dip at 2.8kHz anyway.

BrannigansLaw

Thanks dcibel. I'm learning! I definitely need to find a better measurement location eventually.

I do want to say for the record that I have complete faith in Wolf's design, and that the speakers (even in incomplete form) sound great as-is. However, I'm both inexperienced and rusty, and I don't trust my ears.  I also haven't assembled an XO in a decade, so thought I would verify I had everything correct before final assembly.

Someone at DIYaudio suggested it may just be a diffraction issue, and pointed to some measurements from a german magazine that showed a dip there on-axis with this tweeter. So I took a 30deg off-axis plot (purple line):



and modeled the baffle in Edge



Between the two, I'm inclined to think I'm just seeing a big diffraction dip as I'm using pretty small roundovers (3/8"). Can I get a sanity check?

Wolf

It likely is diffraction to a point, as I have a half-inch roundover on all 4 edges, followed by a step of about 1/4" beyond that. I also show a small peak above nominal level at 1.8kHz that you don't have. I'm wondering if we could see the xover network and verify connections, etc, but I'm betting it's cabinet construction differences as your impedance measurement looks very close to mine outside of the TL box tuning.

PWRRYD

The difference in diffraction between 3/8" round overs and 1/2" round overs would be less than one dB.  Not 7 or more dB as shown in the measurement.

Did you only have the one speaker under test connected when you made your measurement?

BrannigansLaw

I tested both speakers and their response was similar. I also did measurements closer and further away, as well as a 30deg off axis measurement. The dip evened out some at 2m, and evened out quite a bit at 30deg off axis. 

Wolf has been kind enough to review my crossover wiring and doesn't see anything obviously wrong, and dip notwithstanding the speakers sound great. I am also not terribly confident about my measurements; the room is too small and there's too many reflective surfaces that can't be moved. I try to gate those out but as you can see above my impulse response is noisy and I'm gating at 2.5ms. I am inclined to finish building these while I have the spare cycles to work on them, but leave the crossover accessible. Over american thanksgiving I expect to have a few days access to a larger space to do better measurements in. If necessary I will do individual driver measurements and model this xo, make adjustments. 

I wasn't aware that DATS can measure crossover components, perhaps I should double check my values as well... I had checked them using an old multi-meter. Looks like DATS is much more in depth. Does anyone know if I can just hook up the leads to components already soldered into my XO? Or will connection to other XO components affect the value across a resistor or cap?

rjj45

You have received some expert advice here. couple of thoughts / suggestions. Just like the advice to set your mike at 40 inches to show baffle diffraction, you can move your mike close to eliminate / minimize diffraction effects. If that dip at 2.8K disappears then, it's baffle diffraction. If it does not disappear, it's a crossover or driver issue.
If you have your intended crossover on a breadboard, check and double check your topology and connections. It's well worth it to double up your jumpers between components. This has caused me trouble many times.