Tube Rolling: Does it Make a Difference?

[captainbeefheart]

It amuses me that people can claim that they can hear low level distortion without a thought about the level of masking room noise or distortion created by the driver.

This is the main talking point I try and bring up in conversations about many things.

People are losing their minds over the SINAD score which is fine if you know how to interpret it. My biggest issue is that's noise + distortion, noise is by far a big no-no and I do believe a low noise floor is important but I'm not going to lose sleep over .1% THD (especially if it's 2H dominant) which would give an awful SINAD score. It's a "trash" amplifier that only makes distortion they say, but in the grand scheme of things the transducer is going to be biggest offender by far. I just don't understand the need of a power amplifier that is .0002% THD, I mean it's great in regard to academic progress but it's not going to be much better than a tube amp with .05% THD because all is lost with the speaker. I'm not advocating 10% THD tube amps or even 5% THD, I think we can do much better than that but it's quite easy to get satisfactory distortion levels with tubes where the sound quality will not be degraded.

What many don't realize is these amp manufacturers want some of their amps to stand out in the crowd, so 1%-5% THD can be common to see and these amps are going to color the sound. It's relative to power and sensitivity of speakers also, take a 1%THD 100 watt tube amplifier that is running with highly efficient speakers and it's loafing around at 1 watt, distortion should be much lower than the 1% and the amp will sound just fine. But take a 5% THD 300b amp at 7 watts with medium efficient speakers, yes you will probably get some coloration during normal levels, at higher levels you might even get into clipping distortions, not good. Lots of variables involved here.

 

[DualTriode]

I can absolutely believe the measurements. But Nelson's "distortion as colorant" stuff remains completely evidence-free. And with some of @pkane 's software, those distortion profiles can easily be dialed in and actually tested for audibility without laying out a pile of money for actual amplifiers.

Checking out @pkane distortion software. Adjustable distortion phase is not real time just yet.

Thanks DT

 

[Poultrygeist]

This SET amp went from horrible to glorious when I replaced the cheap Chinese 6n3 preamp tubes with WE396's. Chinese amp companies know most buyers will replace the original preamp tubes regardless. So they often come with some of the poorest quality tubes imaginable.

 

[captainbeefheart]

This SET amp went from horrible to glorious when I replaced the cheap Chinese 6n3 preamp tubes with WE396's. Chinese amp companies know most buyers will replace the original preamp tubes regardless. So they often come with some of the poorest quality tubes imaginable.

Quality control must have been pretty bad for it to go out like that. The 6N3 is a nice tube but the problem is they make so many of them unless screened you can get a dud like any modern mass produced tube from China/Russia.

I have compared 6N3/5670/2C51/WE396a with a distortion analyzer and I was surprised with how well the 6N3 performed, but there were a couple duds in the batch so screening is very important. On the flip side the best of the bunch were quite linear and tested lower distortion compared to JAN GE 5670. The handful of WE396a tested were all consistent in performance so Western Electric had quality manufacturing or at least quality control before going out to vendors. The vetted 6N3 tubes were on par with the WE396a in regard to distortion, gain etc..

The company that makes your tube amp purchases boxes of 6N3 tubes from China at wholesale and who knows what their quality control process is. My guess with most of those types of companies is they don't screen the tubes, they get pulled from the box and installed into the amplifier, then the amplifier goes through a very quick quality control test to make sure there is output into load, slap a sticker on it and it goes out. They aren't going to go through rigorous performance tests of each amplifier before going out, basic "it works" tests and that's it.

I just don't want people interested in the 6N3 tube to think they are all bad, it's just the WE396a is more consistently made. If you purchase 6N3 tubes I don't recommend getting random cheapos from ebay unless you have the means to test them or since they are so cheap just get a bunch of them and I guarantee you'll find some nice ones in the lot.

 

[captainbeefheart]

Weird, the reviews and US sales are with 12AX7 and 12AU7 preamp tubes not 6N3. Are you located in Europe?

 

[srkbear]

My experience with valves primarily comes from building point-to-point-wire reproductions of vintage Fender Tweed guitar amps. And although the implementation and subjective targets desired from tube amplification in these applications differ considerably from those in Hi-Fi audio, there is a substantial amount of research, literature and accepted lore in the guitar industry describing unmistakable differences in tone, sustain, and complex harmonics among an endless compendium of tube varieties, with nuances as granular as brands, plate colors, year of manufacture, age, and testing characteristics.

However, despite my having acquiesced to these doctrines for decades now, I’m also innately inclined to trust Amir’s evidence—which I try to consider with as much attention to methodology, design, and controls for potential confounds that I can without being an expert in audio science. And after many months on here I have found his approach to be uniformly rigorous, ingenious, and compelling. It’s hard to argue with the conclusions he demonstrated to the hypothesis he challenged here.

So this one really, really bugged me. I had to weigh a highly uncommon example of objective evidence in this topic against the long-held beliefs that informed the three large cases of tubes I have proudly displayed on the shelves of my studio, along with the $5,000 Hickok tube tester I bought to calibrate what Amir is suggesting might be apocryphal. Are all the perfectly matched NOS Mullard, Telefunken or Genalex 12ax7s that are considered highly collectible and easily fetch four figure sums from eminent musicians really no “better” than a set of cheap ones manufactured today in China or Russia? Have all my heroes been fleeced all these decades by the mysteries of cognitive dissonance and groupthink?

I’ve decided I can’t come to a conclusion here, primarily because further studies must be performed to solve this and I’m not certain they’re possible to design. And that is primarily due to the fact that in the guitar amp, the audible result we aspire to achieve with tubes, and what we consider divine, is how they perform when pushed beyond the limits of their design. All the beautiful tones, sustain and complex cascades of second and third order harmonics emerge when we drive tubes, or more precisely overdrive them, into their farthest limits of tolerance and distortion—which is antithetical (for the most part) to what is desirable in audiophile applications.

And I’m not sure how one could distill such an abstract and elusive target down to a tangible variable that would be suitable for a scientifically-valid ABX test with a statistically-significant sample size, nor am I sure what the study arms would be, or how I’d go about controlling it. And to carry this forward to the topic of this thread, since distortion (on a far subtler scale) is also the desirable endpoint for valves in audiophile applications, I’m not sure how that could be standardized for ABX testing either. I’m open to suggestions. Until then I’m stuck with a massive new conundrum that I wasn’t prepared for, and man does it piss me off!

 

[SIY]

My experience with valves primarily comes from building point-to-point-wire reproductions of vintage Fender Tweed guitar amps. And although the implementation and subjective targets desired from tube amplification in these applications differ considerably from those in Hi-Fi audio, there is a substantial amount of research, literature and accepted lore in the guitar industry describing unmistakable differences in tone, sustain, and complex harmonics among an endless compendium of tube varieties, with nuances as granular as brands, plate colors, year of manufacture, age, and testing characteristics.

However, despite my having acquiesced to these doctrines for decades now, I’m also innately inclined to trust Amir’s evidence—which I try to consider with as much attention to methodology, design, and controls for potential confounds that I can without being an expert in audio science. And after many months on here I have found his approach to be uniformly rigorous, ingenious, and compelling. It’s hard to argue with the conclusions he demonstrated to the hypothesis he challenged here.

So this one really, really bugged me. I had to weigh a highly uncommon example of objective evidence in this topic against the long-held beliefs that informed the three large cases of tubes I have proudly displayed on the shelves of my studio, along with the $5,000 Hickok tube tester I bought to calibrate what Amir is suggesting might be apocryphal. Are all the perfectly matched NOS Mullard, Telefunken or Genalex 12ax7s that are considered highly collectible and easily fetch four figure sums from eminent musicians really no “better” than a set of cheap ones manufactured today in China or Russia? Have all my heroes been fleeced all these decades by the mysteries of cognitive dissonance and groupthink?

I’ve decided I can’t come to a conclusion here, primarily because further studies must be performed to solve this and I’m not certain they’re possible to design. And that is primarily due to the fact that in the guitar amp, the audible result we aspire to achieve with tubes, and what we consider divine, is how they perform when pushed beyond the limits of their design. All the beautiful tones, sustain and complex cascades of second and third order harmonics emerge when we drive tubes, or more precisely overdrive them, into their farthest limits of tolerance and distortion—which is antithetical (for the most part) to what is desirable in audiophile applications.

And I’m not sure how one could distill such an abstract and elusive target down to a tangible variable that would be suitable for a scientifically-valid ABX test with a statistically-significant sample size, nor am I sure what the study arms would be, or how I’d go about controlling it. And to carry this forward to the topic of this thread, since distortion (on a far subtler scale) is also the desirable endpoint for valves in audiophile applications, I’m not sure how that could be standardized for ABX testing either. I’m open to suggestions. Until then I’m stuck with a massive new conundrum that I wasn’t prepared for, and man does it piss me off!

The best argument against the sort of baseball card collector mentality you describe is looking at the amount of variation in characteristics, noise, and microphonics tube to tube even of the same nominal subcategory. Of course, if someone gave evidence that the baseball card stuff were systematically audible…

A good ears only test would probably use a sorting format.

 

[srkbear]

The best argument against the sort of baseball card collector mentality you describe is looking at the amount of variation in characteristics, noise, and microphonics tube to tube even of the same nominal subcategory. Of course, if someone gave evidence that the baseball card stuff were systematically audible…

A good ears only test would probably use a sorting format.

Well, mine was not so much a “baseball card” collector analogy—I bought them in the service of building repro amps that were as close to the originals I could possibly achieve, and almost every component in the amps I build is NOS or pulled from the spare parts of vintage gear.

My point was that countless storied and revered musicians have sworn by the “tones” alleged to be achievable by specific lots of vintage NOS tubes, and there is certainly a monolithic, unstoppable market out there of “desirable” tubes that are spoken of very gravely as being the Holy Grails of guitar tone. I bought into that dogma long before I discovered the merits and principles of this forum, and I’ve had no quandary with eagerly toppling previously held myths I held about digital audio with the lessons I’ve absorbed here.

This is simply the first time I’ve ever been confronted with the validity of a completely separate musical interest of mine from Hi-Fi audio—one that is effectively considered settled fact amongst countless celebrated musicians through history and one that I’ve never exposed to scientific query.

 

[captainbeefheart]

My experience with valves primarily comes from building point-to-point-wire reproductions of vintage Fender Tweed guitar amps. And although the implementation and subjective targets desired from tube amplification in these applications differ considerably from those in Hi-Fi audio, there is a substantial amount of research, literature and accepted lore in the guitar industry describing unmistakable differences in tone, sustain, and complex harmonics among an endless compendium of tube varieties, with nuances as granular as brands, plate colors, year of manufacture, age, and testing characteristics.

However, despite my having acquiesced to these doctrines for decades now, I’m also innately inclined to trust Amir’s evidence—which I try to consider with as much attention to methodology, design, and controls for potential confounds that I can without being an expert in audio science. And after many months on here I have found his approach to be uniformly rigorous, ingenious, and compelling. It’s hard to argue with the conclusions he demonstrated to the hypothesis he challenged here.

So this one really, really bugged me. I had to weigh a highly uncommon example of objective evidence in this topic against the long-held beliefs that informed the three large cases of tubes I have proudly displayed on the shelves of my studio, along with the $5,000 Hickok tube tester I bought to calibrate what Amir is suggesting might be apocryphal. Are all the perfectly matched NOS Mullard, Telefunken or Genalex 12ax7s that are considered highly collectible and easily fetch four figure sums from eminent musicians really no “better” than a set of cheap ones manufactured today in China or Russia? Have all my heroes been fleeced all these decades by the mysteries of cognitive dissonance and groupthink?

I’ve decided I can’t come to a conclusion here, primarily because further studies must be performed to solve this and I’m not certain they’re possible to design. And that is primarily due to the fact that in the guitar amp, the audible result we aspire to achieve with tubes, and what we consider divine, is how they perform when pushed beyond the limits of their design. All the beautiful tones, sustain and complex cascades of second and third order harmonics emerge when we drive tubes, or more precisely overdrive them, into their farthest limits of tolerance and distortion—which is antithetical (for the most part) to what is desirable in audiophile applications.

And I’m not sure how one could distill such an abstract and elusive target down to a tangible variable that would be suitable for a scientifically-valid ABX test with a statistically-significant sample size, nor am I sure what the study arms would be, or how I’d go about controlling it. And to carry this forward to the topic of this thread, since distortion (on a far subtler scale) is also the desirable endpoint for valves in audiophile applications, I’m not sure how that could be standardized for ABX testing either. I’m open to suggestions. Until then I’m stuck with a massive new conundrum that I wasn’t prepared for, and man does it piss me off!

As you mentioned often guitar amps are pushed into clipping and Amir's test isn't going to show if there are differences when overdriven or not. My theoretical side says no, there isn't going to be a difference because well we know what the results of clipping will be, first and foremost is the grid behaves like a diode, it goes from extremely high impedance to a forward biased diode clamping the signal. All tubes behave this way regardless of pedigree so they can't be different in this regard. On the output side asymmetric clipping leads to even harmonics and symmetrical clipping of both top/bottom peaks lead to odd harmonics. This behavior is again the same regardless of the pedigree of the tube and is determined by the circuit.

If there are differences it will not show up in being "overdriven", differences will show in the normal linear region of operation, an FFT plot will easily show the "differences". The problem is anyone that has ever tested a bunch of tubes, mind you I'm not talking about transconductance testing like with your hickok, that's basically useless and will not show any differences is that even within the same production batch of say a Mullard 12AX7 there will be differences. A curve tracer will show linearity but I'm more akin to just using a typical operating point and use a spectrum analyzer, the FFT will tell all. So from say 20 RCA Black plate 12AX7 tubes and tested at the same output amplitude you will get many different distortion spectrum's.

Some of these manufacturers are more consistent so that I think is the reason for the "hype". On average you're more likely to get a good specimen. In guitar we don't really care about low distortion, I screen my tubes and put them in two piles, one for hifi and the other for guitar use. The guitar use tubes are ones that show high distortion, I mark them in order of harmonic profile so it's not always about THD, but more which harmonics make up the total. Tubes that are 2H dominant regardless if the THD is high will be relatively smooth sounding compared to one that tested lower in THD but has higher order harmonic content. The latter type tubes with a lot of higher order harmonics will have a more aggressive tone to them. When I set up peoples amps I do not care about the brand of tubes that gets installed, it's more a matter of finding the ones that test a particular way and use them. So far I have great results from clients, they love that they can get the tone they are after and don't need to spend a ton of money on the roll of the dice that the "vintage RCA 12AX7" is going be worth the price.

Science is a wonderful thing.

 

[dfuller]

we know what the results of clipping will be, first and foremost is the grid behaves like a diode, it goes from extremely high impedance to a forward biased diode clamping the signal. All tubes behave this way regardless of pedigree so they can't be different in this regard. On the output side asymmetric clipping leads to even harmonics and symmetrical clipping of both top/bottom peaks lead to odd harmonics. This behavior is again the same regardless of the pedigree of the tube and is determined by the circuit.

Yes, but also no. I've done measurements (which I don't currently have access to, but am planning on replicating at some point in the future) - the harmonic spray when they go into grid clipping isn't always quite the same depending on the physical design of the tube. It's not hugely different within a type - but it's not nothing. The tubes from "back in the day" are as a whole quite a bit more consistent, though some of JJ's stuff is surprisingly good these days!
Honestly though, I wish there was some more data on this.

 

[Poultrygeist]

Weird, the reviews and US sales are with 12AX7 and 12AU7 preamp tubes not 6N3. Are you located in Europe?

The earlier units came with 6N3 tubes and later Dared switched over to the 12AX7.

 

[captainbeefheart]

Yes, but also no. I've done measurements (which I don't currently have access to, but am planning on replicating at some point in the future) - the harmonic spray when they go into grid clipping isn't always quite the same depending on the physical design of the tube. It's not hugely different within a type - but it's not nothing. The tubes from "back in the day" are as a whole quite a bit more consistent, though some of JJ's stuff is surprisingly good these days!
Honestly though, I wish there was some more data on this.

Yes definitely a difference from say a beam pentode vs a true pentode etc.. Within the same type though say 6L6GC vs 6L6GC the clipping should be relatively the same unless like you say the construction of the tube is very different but then it wouldn't be the same tube type.

One issue with reviews of EL34's is there are very different constructions from today's manufacturers, some actually have beam plates and not a third suppressor grid, some ar grid aligned and some aren't etc... They may give different clipping results but again they are a different constructions.

True Pentodes depending on loading, if there is no zobel at the output have different distortion vs frequency results compared to a beam pentode. Thanks for your response, I should have been more clear about clipping between different families of tubes.

 

[captainbeefheart]

The earlier units came with 6N3 tubes and later Dared switched over to the 12AX7.

That makes sense. Thanks for the clarification.

Do you know if the later units use both 12AX7 and 12AU7? I believe that's how they are being made now.

 

[dfuller]

One issue with reviews of EL34's is there are very different constructions from today's manufacturers, some actually have beam plates and not a third suppressor grid, some ar grid aligned and some aren't etc... They may give different clipping results but again they are a different constructions.

None of the tubes I know of sold as EL34s are beam tetrodes, but some have aligned control and screen grids, and some don't.

 

[WDeranged]

As you mentioned often guitar amps are pushed into clipping and Amir's test isn't going to show if there are differences when overdriven or not. My theoretical side says no, there isn't going to be a difference because well we know what the results of clipping will be, first and foremost is the grid behaves like a diode, it goes from extremely high impedance to a forward biased diode clamping the signal. All tubes behave this way regardless of pedigree so they can't be different in this regard. On the output side asymmetric clipping leads to even harmonics and symmetrical clipping of both top/bottom peaks lead to odd harmonics. This behavior is again the same regardless of the pedigree of the tube and is determined by the circuit.

If there are differences it will not show up in being "overdriven", differences will show in the normal linear region of operation, an FFT plot will easily show the "differences". The problem is anyone that has ever tested a bunch of tubes, mind you I'm not talking about transconductance testing like with your hickok, that's basically useless and will not show any differences is that even within the same production batch of say a Mullard 12AX7 there will be differences. A curve tracer will show linearity but I'm more akin to just using a typical operating point and use a spectrum analyzer, the FFT will tell all. So from say 20 RCA Black plate 12AX7 tubes and tested at the same output amplitude you will get many different distortion spectrum's.

Some of these manufacturers are more consistent so that I think is the reason for the "hype". On average you're more likely to get a good specimen. In guitar we don't really care about low distortion, I screen my tubes and put them in two piles, one for hifi and the other for guitar use. The guitar use tubes are ones that show high distortion, I mark them in order of harmonic profile so it's not always about THD, but more which harmonics make up the total. Tubes that are 2H dominant regardless if the THD is high will be relatively smooth sounding compared to one that tested lower in THD but has higher order harmonic content. The latter type tubes with a lot of higher order harmonics will have a more aggressive tone to them. When I set up peoples amps I do not care about the brand of tubes that gets installed, it's more a matter of finding the ones that test a particular way and use them. So far I have great results from clients, they love that they can get the tone they are after and don't need to spend a ton of money on the roll of the dice that the "vintage RCA 12AX7" is going be worth the price.

Science is a wonderful thing.

I'm no expert but I do play some guitar. I have a tiny Fender tube amp which was running a pair of no-name Chinese valves. I replaced them with some NOS Mullards and the change in overdriven tone was clear. A large amount of high frequency hash was mostly gone. Otherwise the same general tone, just smoother.

 

[srkbear]

I'm no expert but I do play some guitar. I have a tiny Fender tube amp which was running a pair of no-name Chinese valves. I replaced them with some NOS Mullards and the change in overdriven tone was clear. A large amount of high frequency hash was mostly gone. Otherwise the same general tone, just smoother.

Me too—I wrote an overlong post about the differences a few posts back. I’m not sure the findings here are generalizable to our applications because the tones and sustain we’re seeking occur when tubes are driven beyond their intended purposes. I’m looking into whether there’s any good scientific data on tube performance or valid differences among them when they’re overdriven beyond their normal operating ranges, because I build guitar amps. Certainly there’s a wealth of anecdotal data that drives up the price of rare NOS lots.

 

[captainbeefheart]

None of the tubes I know of sold as EL34s are beam tetrodes, but some have aligned control and screen grids, and some don't.

The Chinese EL34-b is a beam tetrode where the EL34-a is a true pentode.

 

[dfuller]

The Chinese EL34-b is a beam tetrode where the EL34-a is a true pentode.

View attachment 208895

They're not, I have some in my stash. There's a suppressor grid and no beam plate.

 

[captainbeefheart]

I'm no expert but I do play some guitar. I have a tiny Fender tube amp which was running a pair of no-name Chinese valves. I replaced them with some NOS Mullards and the change in overdriven tone was clear. A large amount of high frequency hash was mostly gone. Otherwise the same general tone, just smoother.

EDIT: I made a mistake and even the 6BQ5 is a pentode with suppressor grid, for some reason I had thought they were Beam tubes like the 6V6/6L6.

The American companies had to come up with the beam tubes to get around the Philips patent for a pentode. This is confusing because the 6BQ5 is a true pentode yet American made. I did find this.

"The "American" designation of the famous European EL84.
JEDEC registration to Rogers, Canada, as sponsor for Philips.
Of course all models made before 1956 used the European EL84. "

 

[captainbeefheart]

They're not, I have some in my stash. There's a suppressor grid and no beam plate.

Interesting, weird how they make them and yet call them beam tetrodes. They also are rated for 30 watts dissipation instead of the normal 25 watts. Have you smashed one open or can you see the third screen grid through the plates? I have really no first hand experience with them but I remember when they came out and were touted as a beam tetrode and so suited better for hifi.