• Check out the Annual AK Secret Santa exchange Program, signup before November 23rd in the AK Charities & Christmas forum!

loving me some google ai,

stereoud7

Member
So I was having muddy bass from my Tekton Nebo speakers.

My signal path is an Eversolo t8 streamer into a Denafrips Venus 2 dac, into a PS Audio BHK pre-amp, into a rogue dragon amp (Amperex tubes) .

The PS Audio pre had had some NOS Mullards in it. AI pin pointed the problem as the Mullards. I put some MAZDA 12AU7 tubes from an unused Prima Luna cd player in there and it fixed everything. Crystal highs and tight bass.

I can't believe after describing the issue and my system to google that AI could pinpoint the problem and suggested the solution.
 
Last edited:
Register to hide this ad
If you use them to distill all the forums into an answer, they can be pretty good.

Can be.

Also can go way off the rails, but it's often not that hard to sort out.

Glad you got it sorted out! That's a cool fix, and hope you can get more Mazdas. Mullard pre tubes are cool, but in a pipe, slippers, and smoking jacket sort of warm & fuzzy way.

Unless you're trying to nail the "Iron Man" guitar tone, then they're the only way.
 
AI shocks me in how often it gets very hard questions right, as well as how often it gets easy questions wrong. The problem is that it always sound authoritatively correct, whether right or wrong, so you have to do some due diligence to leverage it effectively. One important element in doing that is to look at the sources it used to give you the answer (a source like Reddit is often wrong, for example).

If you interact withAIt, and tell it when it’s wrong (it will respond with something like “that’s a very astute observation” and then correct its path), it can provide some great info.
 
Sorry, having difficulty with the notion that new-testing vintage Mullard tubes were responsible for muddy bass.

Glad you're enjoying the Mazdas, which are excellent tubes, but these issues are not caused or solved by tube brands.
I bought the mullards pretty recently from the company that used to be Brent Jessy Recordings. AI said the problem was mostly caused form worn tubes so maybe one could have aged quickly.
 
Not all “NOS Mullards” are created equal. I experienced a very similar issue with some box plate later issue Mullard 12AX7s. These were stamped as CV4004’s. I think manufactured in the early to mid 70’s and purchased from a very reputable dealer in the LA area. These were used as the output stage for my phono preamp. Sound was muddy thick bass. Swapped those out for a pair of new stock JJ frame grid type 12AX7s and presto! Problem solved.
 
Last edited:
I just used it to verify a generator size we are trying to use for a project where I work, seemed very helpful and verified what I had already came up with, it also showed the calculations and reasoning behind its choice.
 
Last edited:
Mmm. I used it to verify that had the Germans introduced time & contact fused flak shells in 1943 (rather than late '44 when it was too late, they were only using time fused shells), it probably would have ended 8thAF daylight bombing before Schweinfurt/regensburg, or turned those missions into 50%+ losses, significantly delaying the d-day landings. Lots of 'dud' pass-through hits would have been loss-of-aircraft, x4 effectiveness.
 
I don't use a lot of A.I. models, but I have the feeling that the Google search AI is not a good general A.I. for most problems- good for regurgitating the "conventional wisdom" (right or wrong) it has trained on, so it's a great research assistant for reading the whole library and coming back with a summary, but it SUCKS at extrapolating anything beyond that. It can tell me in detail how well a particular op-amp circuit I have drawn is working, but completely misses an obvious output to feedback inversion that would cause the circuit to wildly oscillate. It's important to remember that A.I. does not "think" or "understand". It only regurgitates.
 
In my working life I was asked to allow an AI to take the exam I used as an end of semester assessment for my automotive students in their advanced electronics class.
Grading the AI it was clear that if the answer could be found online the AI would answer correctly.
If the answer had to be determined using logic and prior knowledge the AI would fail every time.
I took the results to mean that the AI was great at finding information but miserable at "putting two and two" together.
 
I have used a paid version of AI for over a year now & it's been extremely helpful with various things..but I would say I use it for DIY 90% of the time.

I prefer the paid version because I've gone thru months of basically programming it where to get information & where not to. To be direct with me but to also break things down instead of just giving a simple answer. I want an explanation of the answer. To sum it up I have spent a lot of time with my AI so that it knows what I expect from it. The paid version seems to always remember these instructions as well as everything we have worked on in the past. It has a pretty good bead on me..knows a lot about me. Which is kind of scary but also extremely helpful.

I have a few of these Aikido circuit/Tubes4Hifi SP13 boards. One I bought as a kit, one I bought as just the PCB board, & one I bought off the used market from someone that could never get it quiet. Which I ended up just pulling the board & using the chassis for a different project.

Paid ver. of AI was very helpful with me taking my kit version that was pretty to dead quiet. Building one of these by kit instructions involves having no safety earth connection. Basically any ground ends all just met back at the IEC ground tab. Isolating the on/off button & attenuator from chassis (which comes pre-built or already installed with the kit) & adding a proper safety earth to chassis turned this preamp dead quiet.

The other preamp I built off the PCB always had a slight low level hum in the right channel. Otherwise extremely quiet, just this him in the right channel. This has been going on for years. Probably 6 or 7 months ago i started a project with my paid version of AI to try & fix this hum problem. The only difference between the hum preamp & the non-hum preamp using the same PCB is passive parts & remote volume control. The original kit uses 2 ground inputs from the input jacks, 2 ground tabs from the attenuator to linestage PCB input. And I've wired it this way..each channel gets its own separate signal & ground wire..from input jack to output jack.

The hum preamp with the better volume control only has one input ground tab & one ground output tab on its PCB. So from the output of the attenuator board I run a single ground from the attenuator board to one of the 2 ground tabs on the input of the PCB.
The line stage PCB has a single main gnd output on the rear of the board which goes to chassis. All the ground points on the PCB meet there. The 2 input grounds have continuity to various ground points & the main output gnd. Using a meter they are all connected by less than an ohm.

After taking various measurements & me concluding my hum is 120hz AI is convinced we have some ripple leakage, loose or shabby heater wire connection, or soldering issue somewhere in the right channel. On & off for months I'm basically redoing the right channel. We get to the point where AI is like you need a scope bro. I really need one anyway with the amount of projects I do so I'm looking online for something I can afford.

Then on a whim I use regular Google and ask what could be causing a right channel him in my preamp. It proceeds to tell me everything that could be causing it. Google AI asks me which preamp I'm using. I proceed to tell it about the kit PCB & I actually had the board I pulled from the used unit I bought & take a picture of it. It responds, run a jumper from the one input gnd to the other, all will be fixed & running quietly. He says it with such authority. I respond that it shouldn't matter..they are already connected at an almost 0 ohm rating. He explains to me perfectly why that doesn't matter. So I put a simple jumper between the two inputs & the preamp is dead quiet.

I apologize for the long story. You probably can't imagine the hair pulling time I've spent trying to get that very low level hum eliminated. Frustrating. What I found odd, discouraging & with a bit of anger is that one AI, that knew everything about the preamp & work done on it took me on an extremely long route with no positive ending & then another AI that knew nothing really about the preamp & what I had been thru fixed me in a matter of seconds. My paid version said it never considered the jumper because during those months troubleshooting, the return ground reference between both channels never occurred to it. The other, non-paid version took 2 seconds to diagnose the issue.
 
Then on a whim I use regular Google and ask what could be causing a right channel him in my preamp. [...] It responds, run a jumper from the one input gnd to the other, all will be fixed & running quietly.
The AI aspect of "regular" google is largely a natural language interface to their search tool plus ability to collate and reformat results. It answered your question based on what has been said online about hum in that circuit (and there is quite a bit).

If you ask it a question that is not answered within its search corpus, it will give you related (but maybe not useful) information or else just "hallucinate." In comparison, other AI's (including full gemini) may be better at being an "assistant" while not having immediate access to the whole of the internet.
 
So I was having muddy bass from my Tekton Nebo speakers.

My signal path is an Eversolo t8 streamer into a Denafrips Venus 2 dac, into a PS Audio BHK pre-amp, into a rogue dragon amp (Amperex tubes) .

The PS Audio pre had had some NOS Mullards in it. AI pin pointed the problem as the Mullards. I put some MAZDA 12AU7 tubes from an unused Prima Luna cd player in there and it fixed everything. Crystal highs and tight bass.

I can't believe after describing the issue and my system to google that AI could pinpoint the problem and suggested the solution.
It still makes mistakes.

I have had good luck with it making tables to compare transistors, but I always double check the pinout on a spec sheet
 
Half dozen iterations with Gemini last night and I finally have an accurate LND150 MOSFET model to work with. Might do some more checks with it, but now it matches the spec sheet drain current vs Resource resistor
 
So I was having muddy bass from my Tekton Nebo speakers.

My signal path is an Eversolo t8 streamer into a Denafrips Venus 2 dac, into a PS Audio BHK pre-amp, into a rogue dragon amp (Amperex tubes) .

The PS Audio pre had had some NOS Mullards in it. AI pin pointed the problem as the Mullards. I put some MAZDA 12AU7 tubes from an unused Prima Luna cd player in there and it fixed everything. Crystal highs and tight bass.

I can't believe after describing the issue and my system to google that AI could pinpoint the problem and suggested the solution.
i wasn't having any issues with mine but i did ask it all sorts of questions about my system's transparency vs other like gear. as expected, speakers are the next upgrade. next level is about 4-5k, used.
fun part is it walked me through wiims room correction and did a decent job. i was not making progress till then. i was using the wrong parameters.
 
for yucks I asked the Google AI to spit me out a chart comparing specs on two tubes. The one for max plate voltage was wrong on the first line, so I didn't bother to compare the rest. Sheets all say the same thing, so no idea where it got the value it used.
 
Back
Top Bottom