Tube Rolling: Does it Make a Difference?

[Schlippwhip68]

Anyway...back to the topic in hand and the T8 does benefit from a tube roll overall imo. Also the Aiyima A8 with the MA12070 chip really does rock when coupled with the Topping E30 and iTunes on my PC! Its much louder too..

 

[ajc9988]

For folks asking about op-amp rolling, I have done that twice:

Rolling Op-amps in Gustard H20 Headphone Amplifier

I reviewed the Gustard H20 which is a high-end headphone amplifier yesterday. Its owner kindly sent me some high-end op-amps to "roll" in there to see what difference it makes. The task was pretty easy. The top screws come out of the H20 and if you tilt the unit, the lid falls out exposing...

www.audiosciencereview.com

Rolling (replacing) Op-amps in Topping D10 DAC

This is a study of whether changing op-amps in a DAC makes a measurable difference. Taking a lead form "rolling" tubes in tube products, a trend has started in replacing op-amps -- sometimes with much difficulty -- to improve performance of solid state products like DACs. A few months ago I...

www.audiosciencereview.com

You have a problem with your methodology. You are applying a phenomena found in headphone amps, which have on average very low distortion levels from the start because there is less demand on the loads, to then extrapolate that behavior to all amps, including those pushing 100+ watts. This is a problem with grand statements, they are not always necessarily true.

Instead, can you run the experiment one more time, but this time using something that should have a more noticeable effect. Say, for example, trying it with a TPA3255 design. I propose this for an experiment:

Take either a 3eAudio EAUMT-0260-2-B ( http://www.3e-audio.com/amplifier-kits/tpa3255-2ch-260w/ ), which is based on the design from JLE, the Sylph-D200 (https://www.jlelectronicsph.com/product/sylph-d200-tpa3255-pffb-stereo-ready-to-run-amplifier-module), or an even cheaper (or copied) design of this (like this on AliExpress: https://www.aliexpress.com/item/1005003213211009.html?spm=a2g0o.cart.0.0.37013c00JIN0Kh&mp=1 ) and then swap the op amps. Do NOT use the JLE Sylph-D200, because that was the original designer of the line and they have really good, known performance.

Now, this can give you a couple pieces of content to put out: 1) chinese clones of the JLE Sylph-D200, how do they stack up, and what compromises harmed them and by how much, 2) does op-amp rolling help OUTSIDE OF HEADPHONE AMPLIFIERS, 3) does swapping op amps help with the third and fifth order harmonic distortion found in the Sylph (and if found it the other designs as well, does it help with those).

You writing something off flippantly with sparse data (meaning you have proven that in specific headphone amps, which already have decent performance and low noise, you get very little benefit, if any, and it mainly is flushing your money; but what you haven't proved is whether or not it helps with designs outside of headphone amps, whether or not it can help tame certain distortion found in some amps, which you did show it can swap second order distortion for third order or vice versa in one of your headphone amp tests, and how that might impact an amp with a higher distortion level than headphone amps; etc.) is one of the things I find problematic with your style, while I still appreciate your work and the quality of work put forth in your testing. I wouldn't ask you to do a more detailed dive if I believed your outputted data to be suspect. This is more a methodology and conclusions drawn from data critique.

In fact, although I presented an experiment above, choosing any amplifier with distortion above -100dB (so anything like -90dB or -80dB) would suffice to show that changing the op-amps may not help.

In other words, there is a chance good engineered boards with good parts from the start may not benefit, but cheaper boards might. That has not yet been proved or disproved. And although I have been a bit gruff in my demeanor and explanation here, especially when making a request for you to test, it is partly because of my interpretation of your demeanor on the topic. It's also with becoming hyper-sensitive in many domains of people making global assertions on sparse data, or trying to say because it doesn't work in one scenario or works in one scenario, that will be true universally. It takes much more proof for global statements like that.

This is something that could help the community if done properly. Also, if the op-amp can effect the levels of second and third order distortion, a different op-amp in the JLE Sylph-D200 may lower the 3rd harmonic distortion, which is the highest at just above -100dB, thereby giving it a different rating regarding the THD+N ratio, pulling that down closer to -110).

 

[pseudoid]

Ok, does God listen to CD or LP, and what tubes doth he use?

@Holmz, I think you got that question bass-ackwards.
Some may go as far as answering your Q, with "The CD or LP listen to God" first and foremost!
Honest, this reply is not about religion.
Edit (add):

Going back between the 6550s and KT120 tubes... So I found it fascinating that what I seemed to hear to my surprise matched what others reported as well...

Weird that I think I noticed similar differences between KT88 (='fat bottom') replacements for 6550s (='more shapely'), with a McIntosh275.
I would not have had the cojones to make such claims @asr from my impromptu rolling-and-listening-and-rolling-and-listening results. Thank you.
Which bring us back to the question regarding:
Can such subtle subjective vagaries of tube-rolling be fully discovered/distinguished by the standard test suite (i.e. APx555) utilized for objective data gathering?

 

[Holmz]

@Holmz, I think you got that question bass-ackwards.
Some may go as far as answering your Q, with "The CD or LP listen to God" first and foremost!
Honest, this reply is not about religion.

Like the physics of the universe?

…
Which bring us back to the question regarding:
Can such subtle subjective vagaries of tube-rolling be fully discovered/distinguished by the standard test suite (i.e. APx555) utilized for objective data gathering?

Are you suggesting that SINAD does not describe a frequency response graph?
or
That we need some time-domain description like impulse response?
or
Something else?

 

[pseudoid]

Which bring us back to the question regarding:
Can such subtle subjective vagaries of tube-rolling be fully discovered/distinguished by the standard test suite (i.e. APx555) utilized for objective data gathering?

The exact Q (as worded) still stands!

 

[Holmz]

The exact Q (as worded) still stands!

I don’t think that there are forum rules against posting a rhetorical answer to a rhetorical question?

Are you suggesting that SINAD does not describe a frequency response graph?
or
That we need some time-domain description like impulse response?
or
Something else?

Or were you looking for a “No”?

 

[pseudoid]

I don’t think that there are forum rules against posting a rhetorical answer to a rhetorical question?

My question was NOT of the variety you are stipulating.
But then, again, "I don't think that there are forum rules against answering a question that is NOT rhetorical, with an answer that is."

 

[Holmz]

Can you reword it so that it is more clear?

(I can see an answer of “yes” or “no” being valid.)

 

[charleski]

Do not conflate differences between tubes to translate to significant difference in circuit performance.

FETs, for example, vary unit to unit by 5:1 or more, tubes are more like 10%. Yet no-one is rolling FETs...

I suspect the difference lies in the circuit design used. Transistor circuits generally incorporate critical elements (like negative feedback) whose effect is to mitigate the sample variation in active elements. A lot of tube designs instead rely principally on running the tube roughly within its linear region and inevitably have lower negative feedback (because of the lower GBP) or even none (because of fashion). In such a design the substitution of tubes with different linear regions will end up changing the behaviour of the circuit as a whole.

 

[SIY]

I suspect the difference lies in the circuit design used. Transistor circuits generally incorporate critical elements (like negative feedback) whose effect is to mitigate the sample variation in active elements. A lot of tube designs instead rely principally on running the tube roughly within its linear region and inevitably have lower negative feedback (because of the lower GBP) or even none (because of fashion). In such a design the substitution of tubes with different linear regions will end up changing the behaviour of the circuit as a whole.

The lower feedback is a result of lower open loop gain.

Here's an example of worst-case: my MC phono preamp (His Master's Noise). The distortion limitation is the second stage which is a grounded cathode voltage amp with no cathode degeneration. So truly no feedback. It uses a 6DJ8-type in that hole. If I measure the preamp's distortion using different 6DJ8 types (different brands and origins, 6DJ8, ECC88, 6922, 8416, E88CC, 7308, CCa...) the difference between the very best and very worst is a variation from 0.02% THD to 0.04% THD, dominated by 2nd since it's a single ended stage. I think it might be quite a stretch to claim this is audible. And I see that same spread from tubes of the same type, vintage, and origin, so I can't even say, "The Amperex Bugle Boy is better than the JJ."

Now if I go outside the 6DJ8 equivalents and sub, say, a 6KN8, I see a significant change in gain; likewise, if I put in a variable mu tube like a 6ES8, the linearity will suffer, but this is straying far from tube rolling, and deliberately putting non-optimal tube types into a circuit not designed for them.

edit: This is a worst case. For nearly any other tube preamp and amp, the differences will be much smaller because of cathode degeneration; even bypassed cathodes have degeneration at DC, which reduces tube-to-tube variation in performance. My oddball preamp has a fixed cathode voltage, which will maximize tube-to-tube differences.

 

[mhardy6647]

I don’t think that there are forum rules against posting a rhetorical answer to a rhetorical question?

My brain.
Just.
exploded.

Rhetorically and metaphorically speaking, of course.

 

[JohnVF]

Indeed, when you run listening "tests" with no controls, you can reach any conclusion you like.

This excuse is quite the Mary Sue here.

 

[Killingbeans]

You have a problem with your methodology. You are applying a phenomena found in headphone amps, which have on average very low distortion levels from the start because there is less demand on the loads, to then extrapolate that behavior to all amps, including those pushing 100+ watts. This is a problem with grand statements, they are not always necessarily true.

The op-amps used as buffers in power amps do not push 100+ watts. They are not being "stressed" by the following power stage (other than thermal). They live their life just as tranquil as the buffers in a DAC or a headphone amp. Their behaviour is just as much dominated by circuit topology and implementation as the others.

Of cource there's "power op-amps" like the LM3886, but those aren't swappable. There's also sometimes op-amps used for error correction as part of the feedback loop, but the designer would be nuts to make those rollable.

In fact, although I presented an experiment above, choosing any amplifier with distortion above -100dB (so anything like -90dB or -80dB) would suffice to show that changing the op-amps may not help.

Using op-amps as band-aids on subpar designs? The input buffer is not likely to be a significant contributor in those. Input buffers are not the difficult part of designing a good power amp.

Also, if the op-amp can effect the levels of second and third order distortion, a different op-amp in the JLE Sylph-D200 may lower the 3rd harmonic distortion, which is the highest at just above -100dB, thereby giving it a different rating regarding the THD+N ratio, pulling that down closer to -110).

A way smarter approach would be to measure the power stage without the buffer. If it has the same distribution of harmonics (it most likely does), then you are s¤¤t out of luck. The buffer can only add noise and distortion. It can't remove or "cancel" it. That would be magic.

 

[charleski]

If I measure the preamp's distortion using different 6DJ8 types (different brands and origins, 6DJ8, ECC88, 6922, 8416, E88CC, 7308, CCa...) the difference between the very best and very worst is a variation from 0.02% THD to 0.04% THD, dominated by 2nd since it's a single ended stage.

My limited experience with swapping a 6922 cathode-follower (so 100% local feedback, and some global FB in the circuit as well) for a Russian 6N23p showed far more dramatic results. Though my measurements were rather casual the deterioration was on the order of 30dB. You're probably not going to see that in circuit positions that don't stress the limits of the linear range.

 

[pseudoid]

Can you reword it so that it is more clear?

Sorry @Holmz; I did not mean to be so harsh in my reply.
My original post reply question was borne from the reply of valid observations (<< Not w/eyes but w/ears) of @MattHooper regarding tube-rolling and subjective results he had witnessed.
My attempt was to start a sub-discussions in attempts to bridge the divide between objective and subjective results.
Since it appears that there are some discontinuities between the two ends. [IMHO/YMMV]
I did not mean to ask in a rhetorical manner and was not expecting it to result in a Yes/No answer.

 

[SIY]

My limited experience with swapping a 6922 cathode-follower (so 100% local feedback, and some global FB in the circuit as well) for a Russian 6N23p showed far more dramatic results. Though my measurements were rather casual the deterioration was on the order of 30dB. You're probably not going to see that in circuit positions that don't stress the limits of the linear range.

Could you share details? This seems... odd and unlikely. In my Heretical preamp (6DJ8 cathode follower line amp), I could see essentially zero difference between tubes.

 

[charleski]

Could you share details? This seems... odd and unlikely. In my Heretical preamp (6DJ8 cathode follower line amp), I could see essentially zero difference between tubes.

As I said, my measurements were very casual not really worth more than an anecdote.

But there has been some research done in Japan:
Statistical analysis of manufacturer difference for miniature triode tube characteristics based on physical model parameters
This is focused on tubes used in guitar preamp circuits, but I think the findings apply here as well. Figure 13 from that article is probably the most notable:

A, B and E are the averages of two tubes from each of three different manufacturers (B_out is a tube from manufacturer B with outlying basic characteristics). Although there's little difference in H2, the higher harmonics show a broad spread in value, with the worse offenders approaching levels that may well be audible (see Archimago's THD test for instance).

[Edit] It’s probably equally important to note that the variation in THD that they found was dependent on the tube’s position in the circuit (and thus the operating parameters). Fig 14 shows far lower variation in THD when the tube in the 2nd stage was altered.

 

[Holmz]

Sorry @Holmz; I did not mean to be so harsh in my reply.
My original post reply question was borne from the reply of valid observations (<< Not w/eyes but w/ears) of @MattHooper regarding tube-rolling and subjective results he had witnessed.
My attempt was to start a sub-discussions in attempts to bridge the divide between objective and subjective results.
Since it appears that there are some discontinuities between the two ends. [IMHO/YMMV]
I did not mean to ask in a rhetorical manner and was not expecting it to result in a Yes/No answer.

Thanks for the gentlemanly response.

Let’s assume that there is a sonic difference.
I sort of hope so as I have 8x KT120s in a box, after someone posted a, “You can thank me later” type of response.

In a sine wave sense we are sort of missing whether the steady state signal is at a different phase coming out of a tube amp than out of a solid state amp. But I would assume that an amp with a lot of negative feedback and low damping factor would be “driving” the speaker more than a tube going through a transformer would?

Then there would be gain/power differences from a different tube set.
And
Whether that shows up in an altered frequency response?

I sort of don’t think that a 1kHz could capture that, but an impulse response should.

 

[ajc9988]

The op-amps used as buffers in power amps do not push 100+ watts. They are not being "stressed" by the following power stage (other than thermal). They live their life just as tranquil as the buffers in a DAC or a headphone amp. Their behaviour is just as much dominated by circuit topology and implementation as the others.

Of cource there's "power op-amps" like the LM3886, but those aren't swappable. There's also sometimes op-amps used for error correction as part of the feedback loop, but the designer would be nuts to make those rollable.



Using op-amps as band-aids on subpar designs? The input buffer is not likely to be a significant contributor in those. Input buffers are not the difficult part of designing a good power amp.



A way smarter approach would be to measure the power stage without the buffer. If it has the same distribution of harmonics (it most likely does), then you are s¤¤t out of luck. The buffer can only add noise and distortion. It can't remove or "cancel" it. That would be magic.

See, that is an answer I can respect better than a person saying they tested it in low noise and distortion designs and did not see anything, then blanket applying it universally without explanation. Thank you for that response!

Moreover, you missed my point for the 100+ watts. I was saying amplifier designs that are not used in headphone amps or dacs as you have, for example, the TPA3255 being used in amps outside of it (as well as many other amps that can be used to driver loudspeakers, not just headphones) which utilize them and the amplifier, as a whole uses more than 100W per channel. Twice in your response do you reference there use outside of what I said or point to headphone amps alone. Part of that was me not making myself clear, but part is ignoring other areas of my original comment, which in context, would have made my point clear.

Thank you for the description of testing by bypassing the input buffer. That is a great way of teaching someone how they can identify if changing op amps would do anything at all. That is the type of information salient to the conversation in a palpable way, and mostly negates me spelling out the nuance in the last paragraph.

With that said, thank you, once again, for your thoughtful and detailed response.

 

[Deleted member 43441]

I recently reviewed the Mhdt Pagoda tube DAC. Owner was kind enough to send me three other tubes to test with it. Here are the set of tubes I received:

View attachment 168967

Let's start again with the tube supplied with the unit (GE 5670)

View attachment 168968

Now let's switch to Tesla 6CC42 (GA):

View attachment 168969

Other than slightly more mains hum at 60 Hz, I see no other difference. Gain is reduced negligible amount.

Next up is the Western Electric JW2C51:

View attachment 168970

Distortion is 0.7 dB lower. So nothing significant there.

Finally I tested the Western Electric 396A tube which produced a surprising result:

View attachment 168971

Channel 1 is the same but channel 2 has 6 dB lower distortion. Is half the tube better or is there some asymmetry in the design?

That's all I have for you.

Conclusions
It is clear that by far the source of distortion is the design and not choice of tube. In three cases the difference is negligible and inaudible. In the forth example, the WE 396A, there is reduction of distortion in one channel. In the larger picture, there is so much distortion here that reducing it with this tube is not going to make a difference one way or the other.

Notice the level of instrumentation it takes to see the impact of the tube. You need to measure to see if there is a difference and not just assume there is.

Personally even if I signed up for a tube product, I would not waste time "rolling tubes." You are likely to suffer more from "rolling placebo" by your brain than any variation in such tubes.

----------
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/

How could you say rolling tubes does not make a difference? That’s it I’m taking my NOS tubes to a different playground lol. *can someone please help me move these they are kind of heavy?*