[rewjack]

Hi

I have myself no problem to time align my full multi-poweramps- multi-channels (7.1) drivers with REW Tools and Jriver's MC19 PEQ delay Tool. It works very good for individual driver at crossover frequency. Time and phase are close to be perfect.

However, my friend has a passive excellent 2ch loudspeaker spks system, using two monoblock poweramps and a 2ch DAC.

Yesterday we tryed to time align left-right spks the way I used REW's IR graphs and MC19 delay DSP tool. For this trial we choosed a 40-200hz sweep signal, we found a hudge 11 ms delay for first arrival?? We fixed it this way, hoping it represents a latency difference in poweramps (a first guess).

The frequency band sweep signal must be too low. Is any frequency signal could represent an reliable average? The manifacturer does not say where are the exact XO position.
The loopback seems to work fine and measurements are repetitive and overlap at exact first arrival position every time for the same spk.

We feagured that this was not very good after earing results. The voices were definitly wrong.

So, is there a way to aling the two loudspeakers with this method?

I hope that this is possible choosing specific freq sweeps in higher bands?


Thanks
jacques

 

[jtalden]

In general, when using loopback timing it is a good idea to include HF in the sweep to maximize the accuracy of REW's location of the loopback IR peak. The higher the freq of the sweep the sharper the IR peak. This makes it easy for REW to pick the exact same time for the reference point. I therefore normally sweep up to 20 kHz just to be safe. It does not matter if I am only testing SWs as the loopback reference IR peak is still nice and sharp. The SWs don't care if the sweep extends higher; they just are not going to have any output there.

That said, I have not personally noticed any issue using more limited sweeps. I occasionally do that for specific needs. REW normally does a very good job of picking the loopback IR peak even if it is a lower freq sweep.

In your case for full range speakers I see no reason not to sweep full range, maybe 40-20k Hz.

To your problem however; I don't see how an 11 ms error could occur even with your current sweep range. Any REW error would be very small so it is not likely to be due to the sweep range used. It is also not possible to be due to the P-amps or the DAC.

The only think I can think of is that the loopback was not actually activated or the loopback cable was shorted or open. If loopback was not active then REW would not have a reference point for the alignment. Maybe there could then be an 11 ms spread in the alignment? If that is not the cause then I am at a loss for other ideas.

If you post the .mdat file I will see if that helps.

 

[rewjack]

To your problem however; I don't see how an 11 ms error could occur even with your current sweep range. Any REW error would be very small so it is not likely to be due to the sweep range used. It is also not possible to be due to the P-amps or the DAC.

The only think I can think of is that the loopback was not actually activated or the loopback cable was shorted or open. If loopback was not active then REW would not have a reference point for the alignment. Maybe there could then be an 11 ms spread in the alignment? If that is not the cause then I am at a loss for other ideas.

If you post the .mdat file I will see if that helps.

That is encouraging anyway to know that it can be done with 3 drivers.

For the loopback issue, I figured that when I am able to repeat the same response all the time, I should have a good loopback reference. This is only my own experience at home. When it doesn't repeat the same I find a problem. In this case it was looking allright.

So I'll hask my friend Gilbert to send me the data, I think we saved them, and post them.
Otherwise, we will do new ones maybe this weekend.

Thank you to stay with us.

jacques

 

[jtalden]

Yes, this quite a stretch on my part and we can rule out a cable problem as REW would have advised that the loopback signal level was low. That only leaves forgetting to activate the loopback function.

An 11 ms spread in location of 2 IRs is possible using a low freq sweep such as yours if there are 2 IR peaks within each trace that are approximately the same size. The L channel may be slightly higher on one peak and the R channel on the other. The selected peaks would be placed at 0 ms though so that would be the tip off that loopback was not engaged. The measurement results would still be very repeatable however.

This was the only scenario that I could imagine that fits the problem. Maybe the .mdat will provide additional thoughts.

 

[rewjack]

Maybe the .mdat will provide additional thoughts.

OK, Gilbert sent me the data.

Here they are. Hope wou'll make them talk

The Parsifal Ovation spks used info.
Note that the woofers are directed from the back in the oposite direction than midrange and tweeters

"Component-wise, the most important driver in the Parsifal Ovation is the custom-made 5" midrange. This one driver, which is manufactured by AudioTechnology, covers the range between 150Hz and 5.4kHz. There’s no crossover at the upper-frequency limit and only a 6dB filter at 900Hz, which compensates for an intrinsic upward tilt in the driver’s frequency response. Instead, Verity relies on the natural mechanical upper limit of the driver. So, for difficult-to-reproduce instruments such as female voice, the Parsifal Ovations have no true electrical crossover and have a mechanical crossover point that's up high, out of the range where the human ear is most sensitive."

http://www.soundstage.com/revequip/verity_parsifal_ovation/index2.html


"Of course there's no free lunch in audio, and relying on a 5" driver to produce that high a frequency invokes a bit of a tradeoff. When I questioned Julien about the behavior of the Parsifal Ovation's midrange at its upper-frequency limit, he mentioned that the driver does indeed beam, meaning that it doesn't produce as much energy off-axis as it does on-axis. However, he explained that their listening tests showed that the benefits of getting the crossover out of the presence region far outweighed the drawbacks of limited off-axis output. An additional consolation, according to Julien, is that because the Parsifal Ovation doesn't produce as much output off-axis, it's possible to place the speaker fairly close to side walls without detrimental sonic consequences.
Another benefit of running the midrange driver high in frequency is that it makes life much easier for the tweeter. A lower crossover point would mean that the tweeter's power handling would decrease and distortion would increase. The Parsifal Ovation's Scan-Speak tweeter, which has extension up to 50kHz, is literally loafing, working well within its comfort zone"

regards
jacques

 

[rewjack]

I forgot to say, but we took 20-20kz too first and then 20-100hz were done with one corrected by 11ms. They are in the response names.

thank you again
jacques

 

[jtalden]

Okay, the .mdat did provide some thoughts.

Looking only at the full range sweeps for now:

> The loopback channel signal is showing up in the measurement channel along with the measurement.

• Zoom out on impulse chart to see that both the IR of the loopback and the IR of measurement are present. The measurements are delayed roughly 179 and 180 ms thus having the 11 ms difference in delay that you identified.
  
• The loopback should not show up like this so it is best to find out why and correct it. A small tic may still appear due to crosstalk between channels, but it should be a very small tic, not the large one we see here.

> The large delay suggest the signal is being convolved, so the delay difference between channels may be due to the correction IRs implemented in the convolver. I don't think this is unusual depending on how they were created. So we are back to your original question:
You added the 11 ms delay to align the channel IRs, but then the voices indicate it was wrong.

I am left wondering is the problem is related to the crosstalk between the channels similar to the same issue with the loopback channel showing up in in the measurement channel. If both speaker are playing both channels we would end up with that 11 ms offset between them creating chaos? Do you think this may be the problem?

 

[rewjack]

Okay, the .mdat did provide some thoughts.

Looking only at the full range sweeps for now:

> The loopback channel signal is showing up in the measurement channel along with the measurement.

• Zoom out on impulse chart to see that both the IR of the loopback and the IR of measurement are present. The measurements are delayed roughly 179 and 180 ms thus having the 11 ms difference in delay that you identified.
  
• The loopback should not show up like this so it is best to find out why and correct it. A small tic may still appear due to crosstalk between channels, but it should be a very small tic, not the large one we see here.

> The large delay suggest the signal is being convolved, so the delay difference between channels may be due to the correction IRs implemented in the convolver. I don't think this is unusual depending on how they were created. So we are back to your original question:
You added the 11 ms delay to align the channel IRs, but then the voices indicate it was wrong.

I am left wondering is the problem is related to the crosstalk between the channels similar to the same issue with the loopback channel showing up in in the measurement channel. If both speaker are playing both channels we would end up with that 11 ms offset between them creating chaos? Do you think this may be the problem?

Well we have something to work on. I'll see with Gilbert when he wants to make new measurements. I have to bring my Equipment to his place, therefore he's not able to check it out now.

So this large croostalk between channels is strange, maybe we will find the problem there? It took us a while to find a way to isolate the reference channel mapped on his PC. We could have left something wrong there. What would indicate if the loopback is clear? Like you say, it would do the same error for both and null it at the end.

I'll need to check if the PEQ was on when the measurements were done. He doesn't use convolution right now, I think. The 11 ms means to me like something close to 3.7meters, this is impossible. Is there any other way to see it in lenght delay?

For now I don't see anything else.

In conclusion, you brought enough indications to see that the loopback was well in function but a too high tic at "0" and maybe check Jriver's MC19 PEQ for a clue.

With the spk informations do you suggest a different sweep range?

Thanks a lot
jacques

 

[jtalden]

So this large croostalk between channels is strange, maybe we will find the problem there? It took us a while to find a way to isolate the reference channel mapped on his PC. We could have left something wrong there. What would indicate if the loopback is clear?

The IR chart shows the reference channel IR at 0 ms where it should be but the magnitude is very high (at full level) and higher than the measurement. That IR peak at 0ms should either not be there at all or be very very small compared to the measurement IR at 190 ms. That loopback signal is only supposed to be only on the loopback channel which is not measured. If a similar situation occurs when in playback mode (there is crosstalk between channels then that is a problem.

Ignoring that issue for a second, the fact that the loopback IR is at 0 ms where it should be means that the locations of the measurements are indeed correct. The ~11 ms difference does exist between channels.

Like you say, it would do the same error for both and null it at the end.

Sorry, I don't understand this comment.

I'll need to check if the PEQ was on when the measurements were done. He doesn't use convolution right now, I think. The 11 ms means to me like something close to 3.7meters, this is impossible. Is there any other way to see it in lenght delay?

The delay is actually about 178 and 190.4 ms!! So there is that ~11 ms delay between them that you found. What I don't understand is how we get a such large delays without using a convolver. PEQ will not cause any significant delay. JRiver is the most likely source as you stated, but I am not a user so I don't know where that would arise. Make sure any speaker distance settings or channel delay settings are properly set. If the reason for the 178, 190 ms delays are found and corrected the 11 ms difference will disappear.

In conclusion, you brought enough indications to see that the loopback was well in function but a too high tic at "0" and maybe check Jriver's MC19 PEQ for a clue.

Good summary, but not PEQ instead, some delay setting.

With the spk informations do you suggest a different sweep range?

A full range sweep is the most accurate for aligning full range speakers.
[We could use just a high freq sweep if we wanted and still get the alignment correct, but that way we cannot look a the all the other response characteristics of each speaker.]