scorlee
Member
Cool, Looking forward to seeing the final product,
-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
- Thread Starter
- #182
Among the things I've tried...
Of the obvious things, that could point back to some problem with the driver measurements. But here's what I get if I bypass the crossover and disconnect the Purifi, compared with the on-axis DXT response from the driver dataset. The only difference between this and my driver measurement conditions is that the golden colored trace is taken with the PR off (back open).
Great alignment, even as the DXT is dropping off a bit more beyond beyond 13 KHz. No notable difference in the region of concern here, which is a few KHz on either side of 10K.
This has me back to searching for something else wrong with the revised crossover construction, even as every test indicates that things are connected properly and securely.
Gratitude in advance for any ideas!
- Swapped mic cable - no change
- Wind screen on the mic - no change
- Re-downloaded the EMM-6 cal from Dayton website - no change
- Shorted R4 - got a 7.9 KHz notch as expected. Also snipped nylon ties on that cap in case it was being squished. - No change - seems to rule out damage to C3.
- Test at higher and lower signal levels - no change - tends to rule out a struggling amp
- Tried my Umik-1 - responses are not identical (attached - ignore the big high bass swings from the open back) - but see the comments below.
Of the obvious things, that could point back to some problem with the driver measurements. But here's what I get if I bypass the crossover and disconnect the Purifi, compared with the on-axis DXT response from the driver dataset. The only difference between this and my driver measurement conditions is that the golden colored trace is taken with the PR off (back open).
Great alignment, even as the DXT is dropping off a bit more beyond beyond 13 KHz. No notable difference in the region of concern here, which is a few KHz on either side of 10K.
This has me back to searching for something else wrong with the revised crossover construction, even as every test indicates that things are connected properly and securely.
Gratitude in advance for any ideas!
Attachments
a4eaudio
Addicted to Fun and Learning
Take an impedance sweep and see how it compares to the VituixCAD simulation. If it looks significantly different, then something is inconsistent between the simmed XO and actual wiring. (Figuring out the issue can still be a challenge, but the last time something like this happened to me I was able to tweak the parts in VituixCAD to figure out where I had gone wrong.)
- Thread Starter
- #184
Thanks, @a4eaudio. Yeah - that's where I've been grasping at straws. As noted above, what's driving me mad is that with the driver measurements I'm using, VituixCAD properly predicted the response I measured with the original crossover. And I can replicate those measurements, so nothing significant seems to have changed in the mic or measurement system.
I just finished disconnecting C3 to re-measure it with DATS - no change there; value is unchanged and ESR is 0.4 Ohms at 10 KHz. While I had things freed up, I elevated C3, then R4, just in case there was some unexpected interaction with L3. No dice.
I'll button things back up and take an impedance sweep as you suggest!
I just finished disconnecting C3 to re-measure it with DATS - no change there; value is unchanged and ESR is 0.4 Ohms at 10 KHz. While I had things freed up, I elevated C3, then R4, just in case there was some unexpected interaction with L3. No dice.
I'll button things back up and take an impedance sweep as you suggest!
a4eaudio
Addicted to Fun and Learning
Well, frustrating news at the least as you are stuck trying to solve the puzzle.
To be clear:
- You are trying to achieve the VituixCAD simulation in the last (third) image of post #177
- But putting the xo in place and measuring you are getting the results of the last (second) image post #178 (i.e., flatter/better than when you had the wrong calibration file but now with a strange/unexpected bump between about 8kHz and 12kHz)
- What is your measurement amplifier?
- What is your measurement distance?
- Thread Starter
- #187
Yes
Yes - which is a surprise since once I corrected the cal, VCAD did well at predicting the rolloff that I saw with the original design.
The amp is an Extron XPA 1002 Plus - I grabbed a couple of these after Amir's tests. I'm measuring at 1 meter (using the DXT screen) on the tweeter axis. I had the listening distance set to 2 m in Vituix CAD but that doesn't change the on-axis response (or much else in this instance).
a4eaudio
Addicted to Fun and Learning
One last long shot. Are you sure you are measuring the same final speaker as you took the raw measurements for the xo design?
Based on what you have said so far, it does sound like you have kept things consistent. But, I had this happen a year ago, where I took all of my measurements using one speaker (call it speaker A)...designed the crossover and measured it (which was in reality speaker B, but I didn't realize at the time)...and the results didn't match VituixCAD. I spent days trying to fix it only to find out that when I put the crossover into the the other speaker (the original speaker A it turned out) it was correct all along. There was enough variation between the two mid-woofers response such that the same xo resulted in pretty different SPL responses. Doesn't seem likely in your case, but wanted to suggest it.
a4eaudio
Addicted to Fun and Learning
Maybe (??)
I asked because I have been some pretty significant divergences from expected measurements when using cheap class D amps with no PFFB. But I don't think that is likely with the Extron and also doesn't seem to manifest in the way where you have the wide hump.
- Thread Starter
- #191
Well, this is interesting...
I have a 45 Ohm divider on the Extron output that's driving the speaker to bring the loopback down a little (DAC/ADC is a TASCAM US-1x2HR). REW is set to use the loopback as both timing and cal. [EDIT: deleted irrelevant/errant comments on the TASCAM ADC]
Here's a comparison of the response measured using a DAC channel loopback as cal (blue) vs. using the actual speaker amp output (the speaker, not the "unused" stereo channel) as loopback (purple). The green line is an approximation of a 0.1 dB/octave line. In V.CAD the response is not above the line at all, so this doesn't explain everything. But it sure is interesting and notable that the departure is only in that > 6 KHz region.
I have a 45 Ohm divider on the Extron output that's driving the speaker to bring the loopback down a little (DAC/ADC is a TASCAM US-1x2HR). REW is set to use the loopback as both timing and cal. [EDIT: deleted irrelevant/errant comments on the TASCAM ADC]
Here's a comparison of the response measured using a DAC channel loopback as cal (blue) vs. using the actual speaker amp output (the speaker, not the "unused" stereo channel) as loopback (purple). The green line is an approximation of a 0.1 dB/octave line. In V.CAD the response is not above the line at all, so this doesn't explain everything. But it sure is interesting and notable that the departure is only in that > 6 KHz region.
Last edited:
- Thread Starter
- #193
More cal curves from the Extron amp: No load, 4 Ohms resistive, and the speaker bare and with 1, 2, and 3 Ohms in series. I suppose that the 4 Ohm resistive load is the proper reference, rather than unloaded. The curves with the speaker seem to be headed toward the +4 Ohm curve, but even with an extra 3 Ohms in series, we're not even close to getting there.
Notably, the biggest departure lines right up with where I am seeing the hump in the response.
If I've not missed anything, this leaves the following questions:
Notably, the biggest departure lines right up with where I am seeing the hump in the response.
If I've not missed anything, this leaves the following questions:
- Are my driver measurements valid? I can try a repeat of the on-axis DXT measurement with no crossover.
- Under what circumstances does this happen? Checking straight DXT again will provide a clue.
- Will it be OK to simply use the speaker output loopback?
a4eaudio
Addicted to Fun and Learning
Do you have a different amp, even a cheap class AB AVR?
- Thread Starter
- #195
And finally (?) more calibration curves for the Extron XPA 1002 Plus, under different loads.
Some relevant (to this project) highlights:
Some relevant (to this project) highlights:
- Other than when loaded by the whole speaker, differences are modest. When driving the Purifi, we're up a half dB max, about an octive and a half above the crossover point.
- The Purifi data should be totally usable.
- The DXT is a trickier question. At 15 KHz, the amp output is about 1.5 dB down relative to the DAC+ADC. This isn't horrible, but it makes it much harder to tweak crossovers and leaves final results up in the air until testing. I should probably experiment with replacing the soundcard cal with the calibration appropriate for each driver. When I took driver data I rolled the loopback (DAC + ADC) cals into the traces, so it's more work than just swapping cals in REW.
- Thread Starter
- #196
Yes - a couple of old 1990's consumer workhorses, plus a decent 1970's Nikko amp
- Thread Starter
- #197
I went ahead and applied cal files constructed from the amp output loopback to the measurements I've been using. The DXT responses have some upswings after around 14 KHz that's compensation for the Extron amp rolling off more quickly than did the DAC-ADC alone. But they don't quite look like what a real driver would do:
In any event, I loaded these into V.CAD. Ignoring those high frequency upswings, indications are (1) the crossover as-built is pretty good, especially if that upswing at the end isn't real.
.. and (2) with tweaking it could be even better, even with resistor changes alone (more linear on-axis, LW, and SP):
I went ahead and closed up the box and ran a very hasty set of responses with large intervals between angles beyond 40 degrees or so. With those caveats, here's what I get:
This is pretty good. Yet I do think I need to pause and assess, decide whether to take new driver data, whether to collect data with an amp that doesn't have the complicated behavior, take the time to be sure no topology changes are in order, and settle on component values.
A big thanks to @a4eaudio for prodding me to look at the amp today, and to @PavelV for suggesting due attention to amp interactions.
In any event, I loaded these into V.CAD. Ignoring those high frequency upswings, indications are (1) the crossover as-built is pretty good, especially if that upswing at the end isn't real.
.. and (2) with tweaking it could be even better, even with resistor changes alone (more linear on-axis, LW, and SP):
I went ahead and closed up the box and ran a very hasty set of responses with large intervals between angles beyond 40 degrees or so. With those caveats, here's what I get:
This is pretty good. Yet I do think I need to pause and assess, decide whether to take new driver data, whether to collect data with an amp that doesn't have the complicated behavior, take the time to be sure no topology changes are in order, and settle on component values.
A big thanks to @a4eaudio for prodding me to look at the amp today, and to @PavelV for suggesting due attention to amp interactions.
Last edited:
- Thread Starter
- #198
Sorry that I'd never responded to this question. Yes - all on the same speaker. In my other unit, the DXT isn't broken in so I'm keeping it separate, with trusty blue tape on unit "B" for when they're side by side in the workshop!
So we have explanation to this strange measurement behaviors. Maybe that's why I'm hesitant to try D-class. Anyway I'm happy you found the cause of real life vs CAD and our intuition predictions anomaly. Definitely this amp has 'mind of it's own'
- Thread Starter
- #200
As I noted above, Amir saw load-dependent variations in the response of the Extron amp, characterized by - as I'm seeing here - a couple dB of peaking prior to rolloff. The behavior does appear to be consistent. I have two of these amps, and their responses are the same. Of course, all future measurements will use loopback from the amp output rather than the amp input.
I've also found that my corrected measurements (new cals applied in REW) lie right on top of newly taken measurements with the amp output looped back. So while I'm almost certainly in a position now to take more precise driver measurements (3D printed myself some jigs to keep mic distance very consistent), I'm not seeing re-collecting driver data as best use of my time.
Here is a comparison of predicted vs. measured curves for Sound Power and On-Axis response. As the full six-pack above for the whole speaker would suggest, we're a bit hot right in the range of the final notch filter. For on-axis, there are two predictions. Note the two predicted response curves for on-axis: the one that's up about 1/4 dB in the notch filter region takes into account the measured ESR of C3, which is 0.43 Ohms. I had an "Aha" moment when I realized that I hadn't entered an actual value (didn't know how much to trust DATS values for ESR).
So, while ESR doesn't explain it, it really does seem that the notch filter doesn't go quite as deep as intended. Lowering R4 from 4.7 to around 3.3 would bring the response down a dB or so in that area. I have some 3.7's on hand, so I think I will make that change and re-test the whole speaker.
I've also found that my corrected measurements (new cals applied in REW) lie right on top of newly taken measurements with the amp output looped back. So while I'm almost certainly in a position now to take more precise driver measurements (3D printed myself some jigs to keep mic distance very consistent), I'm not seeing re-collecting driver data as best use of my time.
Here is a comparison of predicted vs. measured curves for Sound Power and On-Axis response. As the full six-pack above for the whole speaker would suggest, we're a bit hot right in the range of the final notch filter. For on-axis, there are two predictions. Note the two predicted response curves for on-axis: the one that's up about 1/4 dB in the notch filter region takes into account the measured ESR of C3, which is 0.43 Ohms. I had an "Aha" moment when I realized that I hadn't entered an actual value (didn't know how much to trust DATS values for ESR).
So, while ESR doesn't explain it, it really does seem that the notch filter doesn't go quite as deep as intended. Lowering R4 from 4.7 to around 3.3 would bring the response down a dB or so in that area. I have some 3.7's on hand, so I think I will make that change and re-test the whole speaker.
