1/12th is what is always shown in the Distortion view in REW, I'd say no more than that, is fair. but as you've shown, with gating, things can be cleaned up as well... I guess it depends on what one wants to see. I want to see how the lone speaker sounds, so gating might apply, I also want to see how well its dealing with the room, where RAW and Ungated shine. I guess multiple views is desirable really.
Red teaming their output in separate sessions (separate chats) can help with their mistakes. LLMs make mistakes because their self-correction is bad. But LLMs can be good at analyzing results and finding errors in reasoning. You do it in a separate chat so you can put it into a ruthless criticism mode and prevent it from inheriting flawed reasoning.
OpenAI o3/o1 and Gemini Thinking expose their chain of thought (CoT) that creates their output. Anthropic Claude 3.7 is good at analyzing their CoT and output to look for flaws and exceptions in reasoning. For example, I'll have OpenAI or Gemini research something and write a report. Then I'll copy the CoT and report into Claude 3.7 + "Formal" or "Explanatory" mode with a prompt like, "Please analyze this report and look for flaws and exceptions to what is claimed." Then use that to start correcting mistakes and do it all over again.
That is good advice. I've used all of those AI's mentioned at some point, Grok is outperforming them all. @AllenB did you have issue with the response I posted? I have had issues to length, with All of the AI's except for OpenAI, DeepSeek and Grok. I would name them top 3. I used Gemini for the several months until an update really messed with the flow of information. You would think that no would have better training data than Google but go figure. OpenAI is actually fine, but as reported, DeepSeek matches it, and is free. My issue with DeepSeek is that the interactions are throttled. DeepSeek with its "DeepThink" turned on is very impressive. When I tried Grok it was a game changer. I pay the 30 dollars a month and so far have not seen a need to go the API route. The answers are typically so good, text count remains low in comparison to the every other AI I've tried that if there is a limitation on interaction, I've have never reached it... With every other AI except DeepSeek, huge amounts of interaction time is spent correcting logic and mistakes. My efficiency has increased significantly with Grok, I just can't see using anything else until something game changing comes along, that isn't Grok. Its that good, in my opinion.
Grok referencing 1"x1" bracing every 6"s is just another testament, in my opinion, of its prowess. I am still not sure how/where/why it even knows about it. I knew about it already, but how did it know?
Grok referencing 1"x1" bracing every 6"s is just another testament, in my opinion, of its prowess. I am still not sure how/where/why it even knows about it. I knew about it already, but how did it know?
Well they have all the data, I mean its the best search engine in the world, no one is interacting with anything more than Google, on a daily basis. I guess that's not the best premise for "great training data" but, with the way everything is monitor'd/tracked these days, you'd think that Google would have certain advantages given its hugely human based data that I am suggesting it already has. Then again, I don't deeply know what the challenges are. It could be more technical than practical. I also am under the impression that the AI we receive is Dumbed Down... A fully functional AI, that doesn't make stupid, basic, rudimentary mistakes, is truly powerful.
I know what you mean, but I am not saying what I mean, properly. Their search engine is the best, thus, it would seem that all the information that the internet has to offer, like this thread for example, would be easily accessible, and organizable for them, since they are already in the business of accessing and organizing the whole internet. Anything that has flowed thru Google, I think, has been recorded, by Google. Who is dealing with more Data on a day to day basis, than Google? Who outside of Facebook/Apple/Samsung, is interacting with humans on a day to day basis, more than Google. Maybe I am over playing their hand
Just wait till Apple drops its AI. They've been very quiet but they have a hand similar to Google, Everyone is on their Iphone, and just Like everyone else, they are recording Everything that happens there, as well as, any data that is transferring through their devices.
Essentially, most of the Maths and Sciences are available to us and AI ( I think) and AI should become the ultimate simulator, surpassing, HornResp, Akabak, Cosmol, etc etc, just by using the already organized data, logically. Yet, Gemini, for example, will (or did in the past) confuse simple things like acoustical efficiency versus Diaphragm size, literally getting it reversed but then when questioned apologizing.
This is why I know that the AI normal people are using is dumbed down.... AI is the the most powerful thing we will likely see in our life time, with out considering God. What I mean is, BlackRock, the one that we see commonly discussed in conspiracy videos, used AI to become the most powerful corporation in the world... when I was coming out of high school (2000), the emergence of computer enhanced investing, was on the scene.... BlackRock used an investment AI, and won.... Day trading is no longer a thing, because of AI... AI was here before we called it AI.
We are on the bottom, and we receive outdated technology as "new"... Whatever AI the elites have and the military have, is the real deal, and probably making world changing decisions. So the AI we get will be throttled in intelligence because of one, capitalism, and two, unthrottled and unbridled it is simply too powerful for the common man.
So to counter your comment... What has @gedlee and other professionals and experts in this thread, taught, that can't be expressed in Math/Science thats already available? The experts in this thread represent the result of large accumulations of data processed logically. AI's memory is better. Its just a matter of time till we have access to an AI, that will outperform that human experts.
Just wait till Apple drops its AI. They've been very quiet but they have a hand similar to Google, Everyone is on their Iphone, and just Like everyone else, they are recording Everything that happens there, as well as, any data that is transferring through their devices.
Essentially, most of the Maths and Sciences are available to us and AI ( I think) and AI should become the ultimate simulator, surpassing, HornResp, Akabak, Cosmol, etc etc, just by using the already organized data, logically. Yet, Gemini, for example, will (or did in the past) confuse simple things like acoustical efficiency versus Diaphragm size, literally getting it reversed but then when questioned apologizing.
This is why I know that the AI normal people are using is dumbed down.... AI is the the most powerful thing we will likely see in our life time, with out considering God. What I mean is, BlackRock, the one that we see commonly discussed in conspiracy videos, used AI to become the most powerful corporation in the world... when I was coming out of high school (2000), the emergence of computer enhanced investing, was on the scene.... BlackRock used an investment AI, and won.... Day trading is no longer a thing, because of AI... AI was here before we called it AI.
We are on the bottom, and we receive outdated technology as "new"... Whatever AI the elites have and the military have, is the real deal, and probably making world changing decisions. So the AI we get will be throttled in intelligence because of one, capitalism, and two, unthrottled and unbridled it is simply too powerful for the common man.
So to counter your comment... What has @gedlee and other professionals and experts in this thread, taught, that can't be expressed in Math/Science thats already available? The experts in this thread represent the result of large accumulations of data processed logically. AI's memory is better. Its just a matter of time till we have access to an AI, that will outperform that human experts.
Last edited:
System of Stereo Sound's Editor-in-Chief:
Drivers:
Drivers:
Last edited:
A home made fixture, for mounting/dismounting a JBL 150xAL in the enclosure without putting it down on it's back.
Those drivers are heavy suckers, so difficult and risky to mount them.
Not obtainable and no recone kits.
Unfortunately, we do not see what is this tiny horn on top about, especially what driver).
Hello it's me ! 😅
It wasn't me who took the measurements, it was a friend, here is the unsmoothed curve :
It's dirty, no ?
Sincerely,
Laurent
"Since it is recessed, it cannot be removed with bare hands.
A special tool will be brought in to remove the 1500AL.
The 1500AL is 20 years old, so maintenance and upgrades have been required during that time.
This time, it seems that the problem is due to a part that was modified from the original."
They probably find a way to fix the issue, since there's a lot of knowledge about JBL in Japan (and probably parts too).
This was his setup in 2009, so 15 years before the pictures in my previous post. Note the JBLs 435 - including throat-adapter, next to the speakers:
"I went to listen to Mr. T in Tokyo, who introduced a coaxial 2-way driver.
He had previously borrowed an interesting driver from me, so I asked him if I wanted to listen to it, and he sent it to my house.
However, since it was only one at that time, I was unable to confirm the sense of positioning provided by the 2-way driver, but I attached it to my TH-4001 horn and measured and listened to it. As a result, I was impressed by the toughness of the low range, which can be used from about 300Hz as per the catalog, and the high range of 10kHz, which tends to be heavy with my TD-4001 + TH-4001, was very refreshing.
When I told Mr. T about it, he procured two of these drivers and aged them for about a month before listening to them. His speaker system used a JBL unit for the WoodWill box and horn, but the mid-range driver was changed to a 2-way driver called BMS 4594ND. You can see a wooden horn for the tweeter between the horn and the woofer box, but this time we did not use it. This is the 2-way driver. When I borrowed it before, it had a 1.5-inch throat, but this time it seems that he purchased a 2-inch one, and the connection with the horn was much better. Mr. T adjusted the system again with DEQX on this day, and I listened to it right away. The mid-low range cross was set to 300Hz as he wished, and the result was particularly noticeable in vocals, where the low-frequency components of female and male vocals did not become heavy, and the homogeneous tone produced a pleasant sound. The violins in the string ensemble also had a refreshing sound with a silky feel in the high range, a feather-like smoothness, and good cutting. And needless to say, the connection of the sound and the good sense of positioning, which can be said to be the very charm of the 2-way driver, are exquisite. Since all these units are available at a surprisingly low price, it may be that we are finally in an era where you can get ultra-high quality sound at an affordable price by building your own speaker system or using a box designed and hand-made by a professional. Mr. T's report on this product will appear in an audio magazine soon, so please check it out."
This was during the 'DEQX madness', around 15 years ago.
Apparently he later ditched the BMS in favor of the K2-S9800 driver configuration.