-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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!
AV123 / GR Research X-Voce Speaker Review
- Thread starter amirm
- Start date
musicforcities
Senior Member
- Joined
- Sep 18, 2021
- Messages
- 427
- Likes
- 493
You need to buy the upgraded crossover kit from Danny if you want one wired correctly.
They seem very dodgy
They seem very dodgy
That’s really a shame, that isn’t at all how the NFS functions and the white papers on the technology, while quite complex, are out there in the public domain for anyone to learn.
I have personally verified both the resolution and accuracy of the NFS by sending a speaker all the way to Norway and calling in a favor to my friends at SEAS. They took measurements of this speaker in their anechoic chamber and my NFS measurements of the same speaker not only precisely matched, the resolution and accuracy of the NFS exceeded the chamber measurements in the lower frequency ranges.
Look, there is no manufacturer with a perfect scorecard such that every speaker is amazing. You found slot port resonance issues with our Luna and Duo that our measurement system, at that time, did not pick up due to the resonance being extremely high Q. “Gating” to remove reflections also results in reduced resolution for the entire response measurement, which was why we didn’t catch this initially. I trusted my measurements but at no point did I claim our measurements were better or that yours were invalid. It was cause for investigation, not denial.
I spent a ridiculous amount of time researching and LEARNING more about the NFS, which eventually resulted in a very long but thoroughly enjoyable zoom call with Klippel engineers. I am not sure if you were aware, but that Luna you tested was the speaker I shipped all the way to Norway to be measured by a company I 100% trust. End result, MLSSA (which is far more accurate than Clio) could not match the resolution of the NFS under any scenario. I can get close, but – and this is a critical point – the only way to know that my results are close is to actually compare to a reference, and that reference is an NFS measurement or proper measurement in an anechoic chamber. This fact isn’t disputable.
Danny can learn more and embrace the new more advanced technology and standards, or become victim to it. It is really that simple. I’d offer him help with this, but he has publicly disparaged some of our products that he has zero experience with, for no reason whatsoever other than to push his products, so that door has permanently shut…
sam_adams
Major Contributor
- Joined
- Dec 24, 2019
- Messages
- 1,024
- Likes
- 2,551
That would require him to overcome his cognitive bias and come to the realization that everything he believes to be true is false. For the majority of people in this situation, this is an insurmountable psychological obstacle. This is, in this day and age, an issue that many people suffer from and will never get over.
In trying to reproduce automation customer issues over decades, I learned how key it is to match conditions as best as possible.
When I started to work more with speaker measurements, a more experienced designer emphasized that you could only expect to compare results against your own measurements and NOT to expect to reproduce his. As I have worked with others on Directiva, I have kept this in mind. Matching two or more hobbyist’s measurements has always been a challenge in my experience. Usually we have been able to get close matches but rarely are they exact matches. There have always been differences in software, mics or some environmental condition(s) that am convinced made exact matches elusive.
As my own rig is not in a permanent location, I have found it challenging to even ensure results between my own measurement sessions. So I have become accepting of some variances. When you start to share results with others is when you realize how important it is for equipment and conditions to match closely. Directiva r1 could have been done much sooner had I not had to tear down and then debug my test equipment each time. When I say that conditions matter, it is more than just my diatribe. Every time I have been less disciplined, I have lost valuable time and sure others have questioned my results. Seems you have had comparable experiences.
Last edited:
Or designing explosive ordnance.
Last edited:
D
Deleted member 48726
Guest
How does it measure with the towel
No one should have to ask that question. -Ever!
May I ask of what magnitude are these variances, you
I hope you are not chasing just TENTHs of dBs...
- Joined
- Apr 30, 2023
- Messages
- 1,654
- Likes
- 2,511
From the bottom of your heart, did you really thought Danny would have accepted your help or even remotely interested in legit science?
Danny caters to a specific group of uninformed audio consumers, that is his business model.
- Joined
- Mar 17, 2020
- Messages
- 1,071
- Likes
- 4,548
I can second this. Although I can't afford the Klippel asking price, I now send any new model off for a 3rd-party Klippel test to see whether I've messed up. That said, I've also found that very careful outdoor tests with quasi-anechoic measurement software and a properly conducted and and spliced ground plane measurement can come remarkably close to the Klippel measurements. However, winter is closing in and I think I'll continue to ship my designs off to Amir or Erin.
Andysu
Major Contributor
- Joined
- Dec 7, 2019
- Messages
- 3,083
- Likes
- 1,609
Yes, but if I build it they will come.Make sure you don't let @amirm get a scent of your design. He may want to Klippel it...
- Joined
- Jul 4, 2018
- Messages
- 75
- Likes
- 152
Why didn't he just use the "No-Rez" panels GR Research is so proud of on these?
Wow, if amir deepens the investigation a bit more will find who stole my umbrella last wednesday. Great job!
And having told the whole world about it may rise that obstacle even a bit more...
May I ask of what magnitude are these variances, youobserveaccept?
I hope you are not chasing just TENTHs of dBs...
No, I would not usually chase that small a variance and the differences are not always magnitude. I tend to look to see that general patterns are similar for frequency response measurements. For impedance, it is not unusual to see greater differences in magnitude and is more important to have frequencies that match more closely. Distortion measurements are very sensitive to conditions so tend to accept wider variances there. As you may know, Amir is very skeptical of waterfall measurements.
As @AscendDF mentions and we have seen Amir struggle with mic position as not everyone stays on tweeter axis and/or tells you which axis they used. Mic clips can introduce ripples or resonate at certain frequencies, so sometimes it is a matter of accepting or fixing those. Am sure others have good examples too and some of this (like scaling and smoothing) has been discussed elsewhere, so will stop here. If you have more specific questions, feel free to PM me.
AudioSceptic
Major Contributor
This is one of the worst I've ever seen and I'd bet even a cheap smart speaker is better. However, there's a lot of hardware there, so I wonder if replacing all or even just part of the x-over would transform it?This is a review, listening tests, EQ and detailed measurements of the AV123 hybrid open/closed baffle center speaker. It sold for $499 when the company was in business. GR Research seems to be selling the same speaker as a kit ($239 for pats, $250 for flat pack you assemble/finish).
View attachment 326431
No, I didn't just give the speaker a bath. The reason for the towel on top will become clear in the listening test section.
Speaker has unusual configuration. The outer woofers are in rather small sealed boxes. The center three drivers are open to the back, protected by a grill with some stuffing on top of the crossover located there:
View attachment 326432
View attachment 326433
May be able to see it more clearly in the GR Research picture:
It was very easy to get the bent metal to ring or if pushed, pop. Same material is used to screw om a very stout grill to the speaker using the stand-offs you see in my picture.
Speaker is kind of imposing weighing over 60 pounds and being quite deep compared to today's center speakers.
This review has quite a story to go with it so buckle up and let's get into it.
AV123 / GR Research X-Voce Measurements
Worried that the back reflections would create a complex acoustic pattern, I doubled my usual measurement points on Klippel NFS resulting in a scan that took a painful 5.5 hours! The resulting file was some 1.4 Gigabytes. What popped out after the computational phase was quite surprising:
View attachment 326437
What is going on in bass??? Seems like the woofers are just resonating at a frequency (between 80 and 90 Hz) and that is that. No way is it able to blend into mid-woofers that way. Searching high and low online did not result in any measurements of the AV123 X-Voce as the company is out of business and wayback machine did not capture the graphics. So I went to GR Research website and was surprised there was no measurement there either! So I lugged the ungainly and heavy speaker up two flight of stairs to listen to it and finish the review. But then I thought, maybe Danny at GR Research has the measurements for his kit so I asked him for it.
He responded saying he didn't have any as he designed the above for AV123 16 years ago and he has to build his own kit to measure. Kind of strange on both fronts. If I designed a speaker 16 years ago, I would have some electronic record of it in email, old folders, etc. Anyway, he went on to question why I was asking saying. I went ahead and shared the above graph with him (without the notations). He wrote back quickly saying the driver polarity must be wrong and that would cause that deep dip. To me that was a secondary problem to the woofer having such a narrow peak. So I shared with him my near-field measurements of each driver:
View attachment 326438
Here, the mic was just an inch or two in front of each driver, clearly showing that the response from the woofer is wrong regardless of what the mid-woofer is doing. Danny again said that it was clear this was a wiring problem. As much as I don't enjoy ripping open a speaker on a hunch like this with no evidence, I went ahead and took out some 20 to 30 screws and got to the crossover which is behind the tweeter/MTM drivers:
View attachment 326440
I had my son with me and his keen eyes instantly picked up that red wire is hooked up to "LOW-" instead of LOW+. The wires though were all symmetrical from bottom to up, each being white+red so it could be that the PCB label was wrong. To verify, I took out the left woofer and it became clear that indeed, wire colors were correct and the factory installer had wired it wrong. De-soldered the wires, and corrected the mistake and lugged the darn speaker back down two flights and on Klippel NFS. Quick near-field measurement shows little improvement but I thought I do the full scan anyway, this time opting for the standard 2.5 hour one.
Here are the second set of measurements: (original test was with the grill; this one without).
View attachment 326441
We see that the dip is filled in more but the narrow, peaky response of the woofer remains. As I had guessed, the crossover fix didn't deal with the major issue but admittedly that was wrong as well. By this time, the conversation with Danny had gotten ugly with him claiming that my measurements are not high resolution, and that I was wasting my time spending more than 5 minutes on it. What to do now?
I wanted to verify the PCB wiring so googled out there and by accident, landed on a very interest thread on Audiocircle that is "home" to Danny as far as talking to his customers, etc. https://www.audiocircle.com/index.php?topic=162372.0
The owner there had a bunch of GR Research speakers which he liked but found his X-Static speakers boomy. X-Static is similar to X-Voce in that it seems to use the same set of drivers but with the open baffle being on top instead of the middle:
Unfortunately Audiocircle website doesn't allow linking to their images.
"The X-Statik's use a sealed box for the lower woofers. So the natural roll off is pretty smooth and gradual. It doesn't look anything like your in room response.
You'll need to work with placement and room treatments to knock that peak out of it. Your room reflections are out of phase in the dipped area and in phase in the peaked area. So it is possible that if you absorb some of those room reflections you can knock down that peak and at the same time you might bring up that dipped area too."
He didn't know who he was up against as the owner showed in measurement after measurement how this is a property of the speaker and not the room. He eventually narrowed the problem to the drivers being out of spec, experimenting with adding some mass to them and improving things. The thread ends there with no more responses from Danny or the owner. This was in 2019 so would have had to been fresh in Danny's mind. Why he didn't volunteered to me that he had seen this exact problem before is troubling.
Back to our speaker, the owner had purchased this as a back up so had another X-Voce. I asked him to measure it in his room and he got nearly the same response as the poster in the above thread:
View attachment 326461
In absence of measurements like this, one could have blamed this on room modes and such but now that we know the anechoic response of the speaker, it is clear that his sample suffers from the very same problem. Hard to know if he has a crossover wiring issue as well but again, that is a minor problem compared the peaky response of the woofer.
Given all this data that two different productions of AV123 have this problem and so does a sample of similar design in X-Static doing the same years later in 2019, it seems to me that this is a huge production issue. Who knows how many bad speaker kits Danny has sold between these two models (and wherever else he uses the same driver).
Anyway, going with what we have, here is our early window frequency response which naturally looks pretty poor:
View attachment 326450
Notice how the rear reflection is mostly coming from the mid-woofer. Naturally predicted in-room response is poor even though our model is not for rear firing speakers:
View attachment 326451
Directivity is poor of course, making the speaker highly room dependent:
View attachment 326452
We have a classic "MTM" configuration which leads to narrowing of the response as the two mid-woofers cancel each other at some frequencies and angles. Fortunately the problem is not too severe so if you sit back far enough, it may be OK for a few seats. Directivity plot shows the same:
View attachment 326453
Vertically we don't have that problem although it is still best to sit at tweeter axis:
View attachment 326454
Doubling up the woofers and mid-range always results in low distortion which I like:
View attachment 326455
View attachment 326456
Impedance and phase don't show much other the speaker impedance being above average making it an easier load for an amp (although you may get less power):
View attachment 326457
Danny is in the business of modifying other company speakers and routinely complains about "stored energy" in waterfall displays. It is kind of ironic that there is enough of that in X-Voce to power a few houses:
View attachment 326458
AV123 / GR Research X-Voce Listening Tests and Equalization
I wasn't about to lug the speaker upstairs again so opted for living room to listen to it as you see in the review picture. Upon power up, I thought something was seriously broken. Not in bass initially but in how the voices were so recessed. The sound seem to come from behind the speaker as if you were listening to your favorite singer in a public bathroom and you being next door! The back reflections were clearly pulling the perceived image behind the speaker but not a in a good way. Once I got past that I could hear singular bass notes that would come go, and once gone, there was little real bass there.
I thought I deal with tonality first so dialed in a couple of filters in bass:
View attachment 326459
Tuning bass response of a speaker in room is tricky as we are dealing with a mix of what the speaker is doing and the room. Still, after 20 minutes of trial and error, I settled on what you see above. The first two filters give you tight and correct bass response but there wasn't enough of it so I put in the shallow band 5. That acted as a sloping down room curve bringing very good bass response. Prior to that, i had put in the correction at 9 KHz which was also fixed some brightness.
This should have been it but wasn't. I couldn't put my finger on it but the voices and broad regions of the music just didn't sound right. I went to check to see if the mesh grill was the problem. It was resonating but I wasn't sure how audible that was. I went closer to listen to it and immediately found the reason culprit: the rear radiation from those drivers was so distorted! Moved my head in front of the speaker and sound was beautiful. Move to the top close to the mesh and you heard mid-frequencies that were distorted and unpleasant. Clearly the driver was not designed for open baffle application. As a quick solution, I grabbed that white towel and put it on top and back. The difference was remarkable! Speaker now not only sounded right but that phasey, distant effect I had observed was gone. The sound was now anchored to the speaker but without it being a point source.
I went through a number of reference tracks and almost all sounded quite nice. Overall there was a constant signature of speaker sounding spacious which may get boring with music that is not supposed to sound that way. But with the towel on top, it is probably livable.
On tracks with sub-bass, X-Voce would play them at around 20 to 30% with modest amount of distortion -- much better than majority of small ported speakers.
Conclusions
Clearly we have a design/manufacturing issue here. If the poster is right on the other forum, the design was based on one set of drivers, then speakers were/are being manufactured with drivers with different response, resulting in a speaker that is clearly broken. You could get lucky that your room modes kind of compensate for the errors but you will never know. Automatic EQ is likely to remedy most of the problem as well as did my manual EQ.
The notion of letting out random radiation out of the back is just wrong to me. If you are going to do that, you better make sure the response of the speaker is clean. Danny never measures distortion so this would be a lesson to start now. And to test the drivers being shipped to customers. And oh, while he is at it, he needs to measure down to 20 Hz, not gated measurements that are cut off at 200 Hz.
On crossover being miswired, the back of the speaker has a sticker saying the speaker was tested at the factory. Clearly even rudimentary frequency response was not performed to catch a) the wiring problem and b) driver issue if if is indeed the problem. Either way, it is a big mistake and likely responsible for some number of customers not being happy with the sound of their speakers.
At end, this all boils down to not doing proper things in designing and building speakers. Proper testing is a requirement and clearly that is missing seeing how you can't even get a measurement of this speaker from GR Research.
Needless to say, I can't recommend the AV123 or the kit version of X-Voce from GR Research. Danny needs to test his drivers and build a sample of this speaker to figure out what is going on. Assuming our investigation is right, he owes a lot of people fixes for speakers/kits they have bought from him.
------------
As always, questions, comments, recommendations, etc. are welcome.
Any donations are much appreciated using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
oscar_dziki
Active Member
- Joined
- Mar 31, 2021
- Messages
- 190
- Likes
- 310
Why would someone not measure below 200hz? It's not like it costs more money to measure a broader scope...
Researching builds of this speaker indicate that it has not sold well. This would seem a reasonable guess as the speaker is a very large format center that is mainly suited to match AV123’s niche open box tower speaker. All the few crossover pics I have found are mods of the AV123 product. These were likely built before GR sold NoRez so it was not used. Only have seen a AV123 old X-LS and it had foam damping as it was popular at the time (early 2000s) these products were sold.
This thread looks to be more talk than this speaker has had over the decades since it was released. As there seems more downside to Danny trying to defend this speaker, perhaps he has determined that a non-response will be his best play. These designs were superseded by his NX series (also mainly open baffle). Expect he would be more protective of them. It has only been a few days, so we shall see.
Last edited:
Similar threads
