My first REW Measurement - DIY Speakers - Did I get speaker polarity wrong?

[nikb]

Given that the speakers (as I understand it) are not in the same place in the room or relative to the mic, I think this looks OK? The dip on the right speaker between 400-700 is a little concerning but could be due to the room / placement.

Yes, I think I am almost there. Listening to them is pleasant now, but I think the right speaker still dominates, even though the measurement shows that is doesn't. Maybe this is my paranoia or even the room? I will have another look at it tomorrow.

Maybe it would be helpful to measure them from a shorter distance, or maybe measure both in the same position?

 

[kemmler3D]

Maybe it would be helpful to measure them from a shorter distance, or maybe measure both in the same position?

Definitely would help narrow down whether there are still differences in the speakers themselves or if it's the room. Nearfield measurements of each driver can give you a much more focused picture of whether the speaker output is what is intended / expected, because it minimizes the influence of the room.

Once that's settled you can wrangle the problems with your room.

 

[Coltr@ne]

Dayton dat v3 for impedance, total and single driver

 

[jmdesignz2]

AFAIK We can't judge polarity unless you also take a measurement with both speakers playing at once, in that scenario you will see extra dips where the two speakers cancel each other out.

However, the response here looks like the woofers aren't connected or maybe the crossover has a serious problem. 100hz should not be 20dB below 10khz, no matter what is wrong with your room.

Clearly something has gone wrong with these speakers or amp, inverted polarity is possible but it's not just a polarity problem.

Alternately, What mic did you use for the measurement here? Maybe it's not calibrated?

I had a huge suck out at 100hz. It was caused by a cancellation mode from the front wall.

 

[Sokel]

just a short update:

This is what it currently looks like.
All changes were reflowing solder connections, rewiring and recrimping crim-on connectors only. No out of place components were found in the crossover, all components measured OK.

View attachment 491441

I took some room out of your measurements so to see the general stuff higher:

R needs work, something is still amiss, either speaker or amp.

BUT, since you're building a speaker all we do in here is next to useless, you need some quality anechoic data to go forward, there's no way around that.
Take them out and measure.

 

[Sokel]

Distortion up high is also off the charts.
What do you run them with?

 

[Keith_W]

Umm, what is going on here?

This is "L reflowed XO" (green) vs. "L recrimped connectors" (yellow). We expect and want the two FR's to be the same, and they are. So no problem here.

This is "R reflowed XO" (red) vs. "R recrimped connectors" (blue). These measurements are obviously not identical. You can see two problems. First, there is a dip in the blue measurement at about 600Hz. Second, there is a huge bass drop-off.

If you compare the step response of "R reflowed XO" (red) with "R recrimped connectors" (blue) you will see the problem. The woofer polarity is inverted in "R recrimped connectors". This explains the dip at 600Hz (crossover phase cancellation) but does not explain the huge bass drop-off. I don't know what is causing that.

Re @Sokel's point about taking them outdoors for a quasi-anechoic measurement. If you look at the Energy-Time Curve, you can see a LOT of loud and early reflections. Not to mention, you are also measuring loud enough to create a lot of driver distortion. If you want to improve your signal-noise ratio (SNR), it is better to take a longer measurement (1M) instead of a louder measurement.

If you plan to correct your speaker response with DSP, taking a quality measurement is vital. But working with a passive XO, you do not need such a high quality measurement since possible interventions are more limited. Regardless, this isn't a very good measurement - it is obvious the measurement was taken in a smallish room, or a very cluttered room. And there is too much distortion. You are probably measuring loud enough to push the speaker into nonlinear territory. About the only good thing we can say about the measurement is that the SNR is very low.

 

[nikb]

Distortion up high is also off the charts.
What do you run them with?

View attachment 491507

Let me give a bit of detail about my chain used in this thread:

audio chain:
source: WIIM Ultra (for streaming and as preamp. I have other sources but no preamp)
amplifier: Purifi Eigentakt based class D amp from audiophonics.fr

This chain was used in all measurements:
source: desktop PC (arch linux) -> audio interface: Motu M2 -> very long (9m) RCA interconnect cables to other room -> pre: WIIM Ultra* ** (analog in), analog out -> audiophonics.fr amp -> speakers

I have kept the WIIM in the chain for the measurements to 1. protect the speakers and neighbours and 2. leave the chain the way I normally use it.

* It is my understanding that the WIIM ultra does a-d and later d-a conversion when used as a preamp for an analog signal. This is of course one more conversion that is not strictly needed. I could connect the M2 directly to the amp, but I'd have to be careful.

** When using the WIIM Ultra analog out, you can set the output voltage level in the WIIM app. I have set the voltage level to the lowest setting of 200mVrms. All measurements were made with the volume control at 75% (of the 200mVrms I presume).

 

[Sokel]

Let me give a bit of detail about my chain used in this thread:

audio chain:
source: WIIM Ultra (for streaming and as preamp. I have other sources but no preamp)
amplifier: Purifi Eigentakt based class D amp from audiophonics.fr

This chain was used in all measurements:
source: desktop PC (arch linux) -> Motu M2 -> very long (9m) RCA interconnect cables to other room -> pre: WIIM Ultra* (analog in)
pre: WIIM Ultra** (analog out) -> audiophonics.fr amp -> speakers

I have kept the WIIM in the chain for the measurements to 1. protect the speakers and neighbours and 2. leave the chain the way I normally use it.

* It is my understanding that the WIIM ultra does a-d and later d-a conversion when used as a preamp for an analog signal. This is of course one more conversion that is not strictly needed. I could connect the M2 directly to the amp, but I'd have to be careful.

** When using the WIIM Ultra analog out, you can set the output voltage level in the WIIM app. I have set the voltage level to the lowest setting of 200mVrms. All measurements were made with the volume control at 75% (of the 200mVrms I presume).

Ok now, there's the problem.
Something at this long chain is way amiss.
Highs are distorted and they seem like they modulating, specially R.

I would keep the chain short ruling out any faults, away from clipping, etc.
My advice would be to start blanc, design anything on paper, double-triple check all your connections with a DMM and then measure.

 

[nikb]

Ok now, there's the problem.
Something at this long chain is way amiss.
Highs are distorted and they seem like they modulating, specially R.

For the measurements, the motu was turned all the way up while REW was playing at -6DB. Could it be that the WIIMs input is distorting/clipping at this level?

I could use a notebook instead of the desktop computer, get rid of the long interconnects and directly attach the motu to the amp.

 

[nikb]

Umm, what is going on here?

[...]

View attachment 491555

This is "R reflowed XO" (red) vs. "R recrimped connectors" (blue). These measurements are obviously not identical. You can see two problems. First, there is a dip in the blue measurement at about 600Hz. Second, there is a huge bass drop-off.

The bass dropoff happens on every first measurent after I turned off the amp to change anything. I usually make several measurements and discard the first. I only included this one by accident. Measurements after the first look like the red line above. Is this bad? Do your measurements all look the same, first to last?

 

[nikb]

Umm, what is going on here?

[...]

View attachment 491552

If you compare the step response of "R reflowed XO" (red) with "R recrimped connectors" (blue) you will see the problem. The woofer polarity is inverted in "R recrimped connectors".

What a great find! I will investigate this later today.

 

[nikb]

If you plan to correct your speaker response with DSP, taking a quality measurement is vital. But working with a passive XO, you do not need such a high quality measurement since possible interventions are more limited. Regardless, this isn't a very good measurement - it is obvious the measurement was taken in a smallish room, or a very cluttered room. And there is too much distortion. You are probably measuring loud enough to push the speaker into nonlinear territory. About the only good thing we can say about the measurement is that the SNR is very low.

I do plan to use DSP (after I got he speakers to play the best they can).

My priorities are:
- fix speakers as good as possible
- find a good listening position
- optimize room a bit (carpet...)
- DSP

I thought it was good to measure loud so noise from outside interferes less with the measurement. Should I try to measure at listening level?

 

[3ll3d00d]

Fwiw building a crossover without being able to measure impedance is really tying one hand behind your back, it's really such a quick and basic sanity check to verify that you haven't got anything wrong in the wiring.

You can buy a device to do it (eg dats as mentioned above) or build a jig yourself (various instructions on line). It's not hard to do.

 

[Sokel]

For the measurements, the motu was turned all the way up while REW was playing at -6DB. Could it be that the WIIMs input is distorting/clipping at this level?

I could use a notebook instead of the desktop computer, get rid of the long interconnects and directly attach the motu to the amp.

As much I can understand your chain.
-do digital gain of Motu all the way up*
-while doing that, run the "check levels" ** utility with REW at -12dB or at lower level (don't go deaf, way far from that) and adjust its analog gain according to the instructions.
-check again all connections.

* ** even better, use WASAPI exclusive for Motu and I/O in general and adjust the analog gain.
You'll find the tool at the Preferences, at the bottom of the first tab.

 

[Keith_W]

I thought it was good to measure loud so noise from outside interferes less with the measurement. Should I try to measure at listening level?

Measurements should always be done at normal listening level. You haven't read that free REW eBook yet, have you . There is an extensive discussion in there about correct measurement volume and how to raise the SN ratio. If you measure too loud, the frequency response may be unrepresentative because you may push the speaker into nonlinearity. If you want to improve the SN ratio, it is better to take a long sweep - 1M. If the SNR still isn't good enough, take multiple 1M measurements and vector average them.

 

[Juhazi]

Yes, KeithW is right, woofer has wrong polarity in measurement 3: R Recrimped
Otherwise measurements look like they should, but are corrupted by many and high level reflections (see impulse overlays) and noise.
High distortion above 2kHz comes most likely from noise near the mic (not the speaker!) Birds, fridge...

 

[Sokel]

High distortion above 2kHz comes most likely from noise near the mic (not the speaker!) Birds, fridge...

Look at his gated one I posted, if this modulation of the R channel was external noise, it should be present at L too as the mic position is the same.

Also,distortion pattern up high is weird, external noises usually create a bubble like distortion, not sharp spikes like that, these look very similar to the kind of clipping people complain about headache and "harsh" sound.

Edit: if his long chain has some gain done bad, even couple of dB will appear like that (that's electrical, only) :


Normal


clipped

It wouldn't be the end of the world audibly (as bad as it looks it's still lower the the 0.1% pro mark,his is way above though) but it looks suspiciously similar.

 

[nikb]

A quick update:
As you know, I have mainly done general maintenance on my speakers, meaning that I have rewired and recrimped everything and reflowed the solder on both crossovers. This has all been done now, and have no obvious things that need doing.

Unfortunately in the latest iteration of measurements, the imbalance between the two speakers has increased again, to the point where they are not usable (I have to increase the level for the left speaker by almost 25% to achieve balance).

This and the fact that I measure a 0.01V difference between the speakers when I play a 1000 hz sinus and measure voltage at the speaker terminals tells me that I should get a jig for measuring impedance (maybe a Dayton DATS v3). Something is obvously still foul here.

My measurement chain seems to be OK, though. I have since connected my audio interface directly to the amp, and there is not much of a difference in the measured curves.
I have to be very careful with gain seetings to not overload the speakers and thereby increase distortion, though.

This is probably my last update until christmas, as I will be out of town for a while. I will report back with updates when I get back here.

 

[nikb]

For my next measurements, I will take longer sweeps and use a windows notebook for measuring - makes things easier