Yeah, but this is the ultimate implementation of this technology, which very few people have attempted.

 

No pain, no gain as they say...

True....but hey I really didn't meant to come this far to begin with
I thought active amping thing was for some serious nutters...lol


Sent from my iPad using Tapatalk

I know what you mean, "when and how did I become the crazy hifi guy!?"

 

I have been following Keith's progress, but taken an easier path with the same technology

1 hour ago, Keith_W said:

@davewantsmoore that is another good point. Tomorrow I will go and buy an Earthworks M30 microphone and redo all the measurements and probably choose a higher crossover point. I was thinking 4kHz, since that is when the response starts to really drop off. 

1

 

Which mic were you using?  If you already have a reasonable one, then the M30 won't make significant difference.

 

4k or higher looks better.   It will depend on the coverage pattern of the two drivers being blended together as to exactly where works best.   (ie.  if you match them on one axis, do they match when measured from others)

47 minutes ago, davidro said:

Yeah but it takes....

 

 

.... ticking the box that say "put computer to sleep after X minutes"

30 minutes ago, AudioGeek said:

which very few people have attempted

 

 

Many have, it's just not something people talk about a lot, being quite complicated.    Posting a "here's how I did it" isn't a short story

 

 

Many have, it's just not something people talk about a lot, being quite complicated.    Posting a "here's how I did it" isn't a short story [emoji6]

That's like saying hey many listen to contemporary classical music


Sent from my iPad using Tapatalk

1 hour ago, davidro said:

Yeah but it takes money, effort, time, knowledge etc. just look at what the op is going through.

 

 

And you think that being an audiophile is not about time, effort, money, and knowledge? Any hobby is! 

 

Besides, it's not that hard. I'm just abnormally stupid - that's all!

Edited September 6, 20169 yr by Keith_W

11 hours ago, Keith_W said:

 

And you think that being an audiophile is not about time, effort, money, and knowledge? Any hobby is! 

 

Besides, it's not that hard. I'm just abnormally stupid - that's all!

 

 

I don't think so, Keith.  You just chose the hard road (PC & Accourate)! 

 

As you will be able to hear on Saturday ... my 4-way active vinyl system delivers the goods!  But it is miniDSP-based ... not "computer-based"!   And sure, I will have to buy another miniDSP box when they come out with a higher-specced unit ... but at USD600 or so, it is not over the top.  And someone with less of the 'ready green' will buy my cast-off ... so my changeover cost is not outrageous. 

 

 

Andy

 

Edited September 6, 20169 yr by andyr

Oh Andy, I would like to think that I choose the most rational road

 

In any case, Uli got back to me with the diagnosis. If you look carefully at the screen shot, you will see there is a drop down menu to choose the sample rate. The sweeps and crossovers were done at 48kHz, and I used that to generate all the filters required (44.1kHz, 48kHz, and 192kHz). The filters are created as .DBL files, and you have to convert them to .WAV files for use in HQPlayer's convolver. I simply loaded up the 192kHz .DBL file without changing the sample rate in the drop down box from 48kHz to 192kHz and created the 192kHz .WAV files. 

 

This is the result: 

 

 

The blue line is the linearized but uncorrected crossover filter for the tweeter. It was generated at 48kHz. 

 

The green line is the linearized and corrected filter at 48kHz. You can see that its actual crossover point is 2kHz, which is what I set it at. 

 

The red line is the linearized and corrected filter at 44.1kHz. You can see it is shifted to the right. 

 

The brown line is the linearized and corrected filter at 192kHz. You can see it is massively shifted to the left. 

 

How was I supposed to know, eh! Not exactly intuitive! 

 

In any case, tonight I reworked the crossover settings. Changed the tweeter crossover point to 4kHz, as recommended by forum members. I listened to both crossovers (the earlier one crossed over at 2kHz, and the new one crossed over at 4kHz) and cranked up the volume to see if the tweeter would clip. It is a massive difference - the new crossover sounds so much cleaner. So thank you @lusk for pointing that out to me. I am still on a learning curve and I really appreciate helpful comments like yours. You should PM me and arrange a time to come over for a listen. 

 

@davewantsmoore I have not yet measured the off-axis response of the speakers. For that, I would need to take these speakers (110kg!!!) outside and hoist them into the air. Not something I want to do. And even if I found out that the off axis response is far from the Linkwitz ideal, there is not much I can do about it. Take a look at the picture of my room and you will understand. 

 

Plan for tomorrow night: time alignment. As for tonight, i'm off to bed. 

7 hours ago, Keith_W said:

@davewantsmoore I have not yet measured the off-axis response of the speakers. For that, I would need to take these speakers (110kg!!!) outside and hoist them into the air.

 

 

You need to take the speakers outside (or to a bigger room) in order to remove the room from the measurement.      

 

To measure the from a different angle you could simply turn the speaker on the spot, to see a measurement from a different angle - of course, it would still include the room (just like your original measurement does)

 

7 hours ago, Keith_W said:

....  even if I found out that the off axis response is far from the Linkwitz ideal, there is not much I can do about it. Take a look at the picture of my room and you will understand. 

 

 

The idea is that knowing what the response from all angles is, is an essential part of selecting the crossover point for a speaker design.

 

....  So what you would do with this knowledge is change the crossover frequency and/or slope.

Dave, moving all that hifi gear outside just to measure the off-axis response is not something I am particularly keen to do. All this stuff weighs a tonne - speakers are 110kg, one Cary monoblock is 20kg, the SGR is about 10kg, and then there is the PC, all the cabling, and so on. And then I would have to spend hours doing sweeps at various positions whilst hoping it does not rain. I did think about renting an anechoic chamber to do the measurements (there is one at Monash University and another at the Ford factory in Geelong), but then I would have to also hire a van to transport all the stuff there. I even chatted to Stuart of SGR and Leigh of Kyron Audio to see if they would be interested in helping me (for a fee, of course) but even though I let them name their price, both of them politely declined. 

 

It might be easier if I asked Acapella for their off axis response measurements, but they are not good at answering emails, even when I need to order parts from them. 

 

Given that the speaker can be dismantled, It may be possible to measure parts of the speaker instead of the whole. I will have a think about this. 

1 hour ago, davewantsmoore said:

 

You need to take the speakers outside (or to a bigger room) in order to remove the room from the measurement.      

 

so all that gated measurement etc is no replacement/substitute for true anechoic chamber measurement?

Nope, not a substitute unless you have a really big room.

 

The problem is this: sound moves 30cm in 1 millisecond. It would therefore take 5 milliseconds (approx 1.5m) for my right speaker to hit the right wall and bounce back. 

 

5 milliseconds is also how long it would take to generate a 200Hz tone (1000/5). Therefore it is not possible to gate anything below 200Hz in my room, because the wall would be reflecting the front of the wave even before the speaker has finished generating the full wave. 

 

If I wanted to measure the off axis speaker response, I would have even less time to do the gate, therefore the minimum frequency I can measure will start to go up. 

 

Dave is right, the only way to do it is outside. Or a large room. Or an anechoic chamber. 

But above 200Hz is doable right?

16 hours ago, davidro said:

 

Yeah but it takes money, effort, time, knowledge etc. just look at what the op is going through.


Sent from my iPad using Tapatalk

 

 

Having a HTPC and a PC that acts as a Digital Crossover are two different things. But I don't disagree about the effort involved. Money? Compared to what some people spend on cables, HTPC's are generally cheap. 

Edited September 7, 20169 yr by Satanica

3 minutes ago, Satanica said:

 

Having a HTPC and a PC that acts as a Digital Crossover are two different things. But I don't disagree about the effort involved. Money? Compared to what some people spend on cables, HTPC's are generally cheap. 

 

Well if I do wanna get some PC in the rig it will also do the XO duties. So that will be a lot of work. 

I don't play with cables. 

11 minutes ago, davidro said:

 

Well if I do wanna get some PC in the rig it will also do the XO duties. So that will be a lot of work. 

I don't play with cables. 

 

How about DEQX instead?

Just now, Satanica said:

 

How about DEQX instead?

 

out of my budget. besides, other than speaker correction i don't know what it offers over my current setup being Dirac Live + miniDSP 4x10. I will need a couple of DEQX units to achieve 8ch too, I understand.

9 hours ago, davewantsmoore said:

 

You need to take the speakers outside (or to a bigger room) in order to remove the room from the measurement.      

 

To measure the from a different angle you could simply turn the speaker on the spot, to see a measurement from a different angle - of course, it would still include the room (just like your original measurement does)

 

 

The idea is that knowing what the response from all angles is, is an essential part of selecting the crossover point for a speaker design.

 

....  So what you would do with this knowledge is change the crossover frequency and/or slope.

 

Yes, for us normal folks, such measurements are major headache! I guess, though, that it's only relevant for drivers operating above >300Hz and, even then, could be irrelevant if one minimises the off-axis sound field (via near-field set-up or room treatment) or if one knows that the drivers operating in this region have constrained directivity. Or am I just being optimistic? And, assuming that the speaker designer has dealt with off-axis (a big assumption but presumably the case for Keith's high-end horns) I guess that once could use the x/o points of the manufacturer albeit with (presumably) steeper / linear phase cross-overs and be reasonably confident?

 

9 hours ago, Keith_W said:

Dave, moving all that hifi gear outside just to measure the off-axis response is not something I am particularly keen to do. All this stuff weighs a tonne - speakers are 110kg, one Cary monoblock is 20kg, the SGR is about 10kg, and then there is the PC, all the cabling, and so on. And then I would have to spend hours doing sweeps at various positions whilst hoping it does not rain. I did think about renting an anechoic chamber to do the measurements (there is one at Monash University and another at the Ford factory in Geelong), but then I would have to also hire a van to transport all the stuff there. I even chatted to Stuart of SGR and Leigh of Kyron Audio to see if they would be interested in helping me (for a fee, of course) but even though I let them name their price, both of them politely declined. 

 

It might be easier if I asked Acapella for their off axis response measurements, but they are not good at answering emails, even when I need to order parts from them. 

 

Given that the speaker can be dismantled, It may be possible to measure parts of the speaker instead of the whole. I will have a think about this. 

 

What about Paul at Red Spade Audio? 

An excellent suggestion sir...

13 hours ago, Keith_W said:

Dave, moving all that hifi gear outside just to measure the off-axis response is not something I am particularly keen to do.

 

 

 

1

 

I think you misunderstand what I said.

 

Taking the speaker outside will allow you to measure the speaker without the room effects....   but THAT isn't what you said you wanted to do.  You said you wanted to measure the off-axis response.

 

To measure the off-axis response, it isn't essential that you also remove the room effects....   and so, you don't need to "take the speaker outside".

 

11 hours ago, Keith_W said:

Dave is right, the only way to do it is outside. Or a large room. Or an anechoic chamber. 

 

 

davidro is correct.    Above a certain frequency your gated measurements won't contain the room reflections...  and so you can use these to inform the filter design.

 

Even below the frequency where they being to include the room .... they aren't completely useless as a comparison.     Put it this way, you use full range measurements for the your on-axis data, with seemingly no complaints.   It contains the room below a certain frequency.

 

I think I can see what you are saying about it being worse for the off-axis response, you means because you will move the microphone closer and closer to the wall, yes?!?   .....  Placing the speaker and the microphone as far away from walls as you can get them, and then rotating the speaker, can be a good compromise.     Remember, you are just looking for an indication of where the coverage patterns of the drivers change, so you can decide on what crossover frequency is appropriate - you're not looking to take super-accurate measurement that you might use for your 'driver-linearisation'. 

 

Considering that we are talking about the mid to tweeter transition, which is happening at ~4khz, then you don't need too much room.  

 

4 hours ago, zydeco said:

.... could be irrelevant if one minimises the off-axis sound field (via near-field set-up or room treatment) or if one knows that the drivers operating in this region have constrained directivity. Or am I just being optimistic?

1

 

 

 

 

Yes.   You see people contemplate this alot.    Why does the 'off-axis' sound from the speaker matter if <the reasons you listed> ?

 

The total amount of power radiated by the speaker (in all directions) is all heard (in a typical room) and is very important for how the speaker sounds.   The later arriving sound should be similar to the initial sound.

 

4 hours ago, zydeco said:

And, assuming that the speaker designer has dealt with off-axis (a big assumption but presumably the case for Keith's high-end horns) I guess that once could use the x/o points of the manufacturer albeit with (presumably) steeper / linear phase cross-overs and be reasonably confident?

 

 

 

1

 

Not necessarily....   but if you follow a general procedure of linearising the driver response, and then adding symmetrical high and low pass filters....  then this will give you a good result on the measurement axis which you used to linearise the response.    How it will look from other axis is uncertain.    The original crossover filter may have been setting the responses of the drivers up in a way which balanced the response over multiple axis  (ie  not setting the on-axis response 'perfectly flat', as a trade off to optomise the how smooth the off-axis response was).

Edited September 7, 20169 yr by davewantsmoore

Interesting discussion, gents Horns are pretty directional anyway. And to be honest, I am quite happy with the integration between the horn and the tweeter. What I am really interested in is the integration between the horn and the bass driver. Since I bought this speaker all those years ago, it was the integration between the horn and the bass driver that has concerned me the most. It never sounded right, and that was what sparked off this journey in the first place. 

 

In any case, I redid the crossovers AGAIN last night. My initial selection for the crossover points was 80, 500, and 2000. Then it was 80, 500, and 3500. Last night I changed it to 100, 700, and 3500. Much improved. 

 

I also did a listening comparison between the convolution engines of HQPlayer and AcourateConvolver. Now, this is very interesting. 

 

HQPlayer's claim to fame is that it is able upsample any source material to PCM384 or DSD512 - basically limited by how much CPU horsepower you have (and I have plenty). However, the DAC I am currently using, an RME Fireface only maxes out at PCM192. Not only that, but it is the only convolution engine which is able to take a DSD signal, apply DSP, and output a DSD signal without having to convert to PCM for the math. 

 

So AcourateConvolver does not stand a chance, right? It is not able to convert to DSD. It is able to upsample to a max of PCM384. Any incoming signal has to be converted to PCM for processing, and output in PCM. 

 

When I get my NADAC, AcourateConvolver will be even more disadvantaged. The ESS Sabre chips in the NADAC convert all PCM into DSD before D/A conversion. This means, if I were to play back a DSD file using AcourateConvolver, it would involve conversion to PCM for processing and then conversion back to DSD using on chip conversion. And, I have read elsewhere (from someone who had a chat to a Sabre engineer) that most of the reason why Sabre chips excel in DSD and not PCM is because of the on chip conversion. 

 

So, why use AcourateConvolver at all? One huge reason - AcourateFlow. The theory behing AcourateFlow was published in this paper. Basically, the theory states that the "phantom image" (i.e. the 3D illusion created by stereo speakers) is dependent on the amount of crosstalk between the speakers. 

 

In a post on another forum, Uli said this: 

 

We have learnt for many years now that crosstalk is bad. But is that really true? Anyway a digital playback chain is perfect regarding crosstalk. Usually we do not expect that some bits will cross between digital channels during transfer. 

As already mentioned by esklude we can discuss the following situation: we play a modern stereo track created by a mix of pan-potted signals. Let's assume it is a nice track and we love it. A typical example for ILD.
And we play the track perfectly without any crosstalk. Great so far?

It is well known and already described in different papers (Gerzon, Griesinger, Sengpiel ...) that in case of ILD we get a different localization of frequencies played by a constant ILD relationship. So lower frequencies tend to located more close to the center whereas high frequencies get more located closer to the speaker with the higher amplitude.
So how do we recognize a phantom source which creates a mix of frequencies? E.g. we want to locate it at 75% between center (between speakers) and a speaker, thus we apply about 12 dB difference between speaker levels. IMO the answer is that we do not get a pin-point localization. The phantom source is received widened.

So we may think about a panpot law that follows some psychoacoustic principles. So we may e.g. apply a level difference of 16 db (arbitrary number) for low frequencies to get them to the desired position. And a level difference of 8 dB to get high frequencies to the desired position. 12 dB then will fit for e.g. frequencies around 1 kHz.
Indeed this means, that we introduce a frequency dependent crosstalk to ILD panpot. The target is clear, we like to compensate the undesired image widening. A better focusing will relieve the brain in its decoding job. The sound is more relaxed, the brain gets more capacity and thus it is capable to recognize more subtleties in the sound.

All this is not really new. But thinking about the (frequency dependent) crosstalk of turntable pickups I got hit. IMHO by accidence the pickup is more or less creating the required crosstalk. It would explain, why analog playback is often preferred and why a digital playback is often considered nasty or nervous. 

Ok, it is of course possible to try a crosstalk solution on the digital side. The basic task then is to find a suitable panpot law. That's what's behind AcourateFlow. It's name is created by users who described the music to be more flowing.

Stereo playback has its weaknesses as it creates an illusion of phantom sound sources. One weakness is the frequency dependent localization in combination with ILD. A procedure like AcurateFLOW helps. Of course it is not a perfect solution. In case of ITD recordings or combined ITD/ILD recordings we cannot expect perfectness. But anytime when the listening result is more relaxing the playback is simply better or closer to the intended truth 

 

... Uli can be a little difficult to understand, but what he is saying is that the ILD (Interchannel Level Difference, i.e. the difference in loudness between the left/right speaker) and ITD (Interchannel Time Difference, i.e. when the same sound hits your ear) is how phantom images are created and localized. However, the ILD and ITD are frequency dependent. Follow his argument closely and you will see that he says that one reason digital sounds harsh is because it has perfectly zero crosstalk, whereas the crosstalk of a turntable cartridge is much higher. 

 

I was intrigued by this argument, so I shelled out some money and bought AcourateConvolver and turned on AcourateFlow. 

 

What does it sound like? Well, it easily beats HQPlayer. It sounds so much more relaxed. HQPlayer sounds as if there is a spotlight on the musicians, AcourateConvolver makes them sound as if they are softly lit. It just sounds ... more beautiful. On this DAC anyway. 

 

Of course, ideally I would like to retain DSD capability AND have AcourateFlow. But I think it will have to be one or the other. 

I'm not sold on DSD reaaaaaally so that makes sense to me


Sent from my iPhone using Tapatalk

Very interesting Keith. I have gone completely the other way and measured and corrected each speaker inividually (mono) in Acourate before loading the filters into hqp convolution and upsampling to both dsd256 and dsd512.

Sounds exactly as you describe - smoother and more relaxed.

Will have to try a stereo measurement again.

Iam using an rme babyface and a behringer microphone. Did the earthworks upgrade make a difference?