Op-amp Rolling Using Sparkos on Fosi V3 Mono

[SIY]

In my use-case, their life-expectancy hovered around a year or so.

That's extremely short. Possible for poor designs, but my experience with Macs is quite different than that- I have a pair of MI-75s with 15-20 year old tubes that's working extremely well.

 

[CedarX]

In my use-case, their life-expectancy hovered around a year or so.
I must have performed about a dozen "tube-rolling" events, during the 30+ years of a MC275 ownership.

Side note: that’s more like “two years or so…”

 

[pma]

Yeah, tubes are "perfect" devices. "Green" with low power consumption, "small", super "lifetime". Let's use them in computers like back in the forties. And do not forget to mention their tolerance to gamma radiation. Everything living would be dead, but tubes would survive. SOTA devices par ecellence.

 

[solderdude]

Or, maybe you are questioning my liberal use of 'tube-rolling', instead of the necessary and periodic 'tube-replacement'.

Indeed tubes are in sockets for easy installing and replacement.
I would not call replacing faulty tubes 'rolling'. Rolling is more for people who 'hear' different 'magic' between tubes of comparable specs.

[tubes] not so.
In my use-case, their life-expectancy hovered around a year or so.
I must have performed about a dozen "tube-rolling" events, during the 30+ years of a MC275 ownership.

I have an old Graetz tube radio bought by my father about 65 years ago. It still works. Had to clean the switches, the 'eye' has lost 'vision' so is only visible in the dark.
It still has original tubes and capacitors and still works fine. I turn it on for half an hour or so a few times a year.

Have had spanking new tubes (JJ) that died, got noisy within a year or so and old pulls that still keep on working.
I even designed a (hybrid) tube amp that makes rolling easy (auto bias adjust and auto heater voltage select) for people that like to roll tubes and not having to worry about all those things. Quite audible differences from very noisy to rolled-off etc.

Power tubes, especially when they are pairs and require individual biassing may be a good reason to replace or at least check now and then if the bias is still O.K. due to aging.

In circuits with a lot of tubes and overall feedback or loads of local feedback there won't be audible differences unless one is broken/noisy.

Some tube (type)s are more microphonic than others which would be a reason to roll some tubes if that is a problem and external dampers don't do anything worthwhile.

Long live SS.... If one wants 'tube like harmonic generators' J-FETs would be the logical choice to put in a SS.
I even designed one specifically for that purpose but because of the large differences and not many suitable replacements that are drop-in the FETs can't be 'rolled'.
It is even tested by Amir.

Give me SS gear every day...
Tube gear is fun to look at though and, like vinyl, the 'mechanics' of it all are fun and amazing how well all of that can sound... just not practical.
I am a practical guy...

 

[wwenze]

20 years of playing audio and I still don't know what "tube sound" is outside of guitar boxes.

The closest I had to "tube sound" is a DAC with 0.1% distortion and it felt off. Thing was trying to swing 2V with passive IV as a current-out DAC. Swapped the resistor values and distortion dropped 20dB and it sounded normal again.

 

[Count Arthur]

20 years of playing audio and I still don't know what "tube sound" is outside of guitar boxes.

The closest I had to "tube sound" is a DAC with 0.1% distortion and it felt off. Thing was trying to swing 2V with passive IV as a current-out DAC. Swapped the resistor values and distortion dropped 20dB and it sounded normal again.

I've never had any valve gear, but I have tried some valve amp simulator plug-ins with my music player.

What I've discovered, by messing with the settings, increasing the level of harmonic distortion, overdrive, etc. is that when it's set "low" I can't really hear it, and when I increase the effect to where I can hear it, I don't like it.

How accurate, and like a real valve amp, the simulators are, I couldn't say, but it has kind of made me less curious about the real thing.

 

[xNORgate]

Side note: that’s more like “two years or so…”

RE: My apologies for the statement (below), which requires a clarification that is omitted from your math:

I must have performed about a dozen "tube-rolling" events, during the 30+ years of a MC275 ownership.

...was not meant as a continuous usage scenario, throughout.
In those 3+ decades, a variety of other solid-state (Class A, A/B and D) power amplifiers took up the slack... and inevitably lead to many requested Tube/SS listening comparisons.
It broke my heart to finally get rid of my classic MC275 but it had to be done! [sad.gif]

 

[dpuopolo]

Charts are charts, but can you >HEAR< the difference? /s

Only if you need to justify the price.

 

[Jimster480]

20 years of playing audio and I still don't know what "tube sound" is outside of guitar boxes.

The closest I had to "tube sound" is a DAC with 0.1% distortion and it felt off. Thing was trying to swing 2V with passive IV as a current-out DAC. Swapped the resistor values and distortion dropped 20dB and it sounded normal again.

I've never liked tube sounds. They are high distortion typically, but it makes the sound smooth and microdetails are lost.

 

[antcollinet]

So, how do I match level of View attachment 4420372 integrated amplifiers with responses at "flat" setting like this? At 1 kHz? Really?

Of course.

In a blind test that frequency response difference will be an audible difference - one of those that a listening test is used to show audibility of.

 

[antcollinet]

FWIW: Unlike op-amp rolling; tube-rolling had a purpose.

I asked googleGemini about sonic differences between KT88 and 6550 tubes when used with the McIntosh MC-275 tube power amplifier.
Some of the requested 10 bullet-points Gemini provided were strangely close to what I recall of those years of attempting to hear their differences... but never could prove.

Quad match sets of KT88s were more pricey than 6550s. I used to think that it may have been my implicit bias to prefer the KT88s.

How have you verified that the AI has given you factually correct information?

 

[Jimster480]

How have you verified that the AI has given you factually correct information?

AI is mostly useless for these types of tasks. It just makes things up as it goes along and half the stuff that it tells you is just incorrect. It's like asking a 12 year old for the answer to the question.

 

[DualTriode]

Of course.

In a blind test that frequency response difference will be an audible difference - one of those that a listening test is used to show audibility of.

Ant,

Blind testing is old news.

The Machine has been taught, as in an algorithm, what sounds good to humans even at lower than human audibility threshold levels. The Machine can measure it before you can hear it.

Take a look at the AES Document Library to find papers discussing the use of Head Torso Manikins to measure augmented reality and automobile sound systems.

The robot measurements are more reliable and the bots do not take coffee breaks.

Checkout the new Cadillac ATMOS sound systems, designed by the Machine.

Thanks DT

 

[Jimster480]

Ant,

Blind testing is old news.

The Machine has been taught, as in an algorithm, what sounds good to humans even at lower than human audibility threshold levels. The Machine can measure it before you can hear it.

Take a look at the AES Document Library to find papers discussing the use of Head Torso Manikins to measure augmented reality and automobile sound systems.

The robot measurements are more reliable and the bots do not take coffee breaks.

Checkout the new Cadillac ATMOS sound systems, designed by the Machine.

Thanks DT

Sure but we are talking about what would be audible depending on what level of measured performance there is on a given product.

 

[antcollinet]

Ant,

Blind testing is old news.

The Machine has been taught, as in an algorithm, what sounds good to humans even at lower than human audibility threshold levels. The Machine can measure it before you can hear it.

Take a look at the AES Document Library to find papers discussing the use of Head Torso Manikins to measure augmented reality and automobile sound systems.

The robot measurements are more reliable and the bots do not take coffee breaks.

Checkout the new Cadillac ATMOS sound systems, designed by the Machine.

Thanks DT

The topic was level matching.... as needed for blind test.

And the "machine" can only know what is audible, and what sounds good - based on previously carried out blind tests.

And there is still a need for people to be able to do their own blind testing to convince themselves that they are able to trust the "machine"

 

[kyuu]

AI is mostly useless for these types of tasks. It just makes things up as it goes along and half the stuff that it tells you is just incorrect. It's like asking a 12 year old for the answer to the question.

The biggest problem is that you're getting an answer that sounds extremely fluent and confident while being completely indifferent to accuracy. It's like if that 12 year-old had an adult voice, spoke like they had been giving public speeches since they day they were born, and were a prodigal bullshit-artist.

 

[DualTriode]

Sure but we are talking about what would be audible depending on what level of measured performance there is on a given product.

Exactly.

That is the testing result.

The topic was level matching.... as needed for blind test.

And the "machine" can only know what is audible, and what sounds good - based on previously carried out blind tests.

Blind Tests, done in the lab by PhD's

Machine Learning

And there is still a need for people to be able to do their own blind testing to convince themselves that they are able to trust the "machine"

convince themselves?

Trust the Machine?

I trust but verify if I can. Some times I close one eye.

DT

 

[Jimster480]

The biggest problem is that you're getting an answer that sounds extremely fluent and confident while being completely indifferent to accuracy. It's like if that 12 year-old had an adult voice, spoke like they had been giving public speeches since they day they were born, and were a prodigal bullshit-artist.

I don't think I could have described it better myself.

 

[antcollinet]

I trust but verify if I can.

I verify then trust.

The machine (if we are talking about AI here) has shown itself to be most untrustworthy.

 

[DualTriode]

I verify then trust.

The machine (if we are talking about AI here) has shown itself to be most untrustworthy.

Hello,

I invested in AI when AI was IBM Watson. Measurement tools and data processing. The returns from Watson paid for a GRAS 45CA on my shelf.

The machine is measurement hardware as in APx555 and software data processing algorithms. => MATLAB Machine Learning

Experimental design to study human hearing is a bit more problematic. The popular use of "AI" as you find it on Bing is not going to help much when you and the guy next door attempt ABX testing in his garage. Show me the published AES paper. Personally I do not care to manage 100 trained listeners or to buy them lunch. I might buy my neighbor lunch.

I trust the work of Sean Olive and the other PhD's at Harman. They have the human data to support the measurements taken with a Human Head and Torso Bot inside a automobile.

Some things are DIY fun even ABX testing with one eye closed.

Thanks DT

An example of machine testing of Audio Quality:

https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.1387-2-202305-I!!PDF-E.pdf