Take the blind challenge! 300B SET vs. Straight Wire with Gain

[GXAlan]

The challenge:
How does a budget 300B SET compare to a "straight wire with gain"?

You tell me. It's time to put your ears to the challenge.

Poll is open for 14 days. No peeking at the results.

Test Setup:
I connected a Topping D90, set it at 0 dB volume, to a budget 300B single-ended integrated tube amplifier "based on the 91B" and connected it to a pair of Infinity Infinitesimal IV speakers (rated at 90 dB/1W/1m) and set the volume to a comfortable listening level for music. This ends up being in the <1W range for me. I then switched out the speakers for a E1DA Cosmos ADC with a resistive load and recorded real music at 32-bit 44.1 kHz. The out-of-the-box Psvane tubes were used for the recordings. @John Atkinson 's comment on a different 300B amplifier applies to this 300B amplifier also: " Its measured performance is what I would expect from an amplifier with a single-ended output stage that uses a single 300B tube."

Then, for comparison, I took the Topping D90, set it at 0 dB volume, and recorded it directly to the E1DA Cosmos ADC (via RCA adapters) using the DIP switches to get volume matching that was closer than 0.5 dB. The recording was then further digitally volume matched according to peak dBFS as calculated by @pkane's DeltaWave. for my initial 30 second recordings. This is your theoretical straight wire with gain.

I then cropped the music according to copyright law, and exported to 24-bit FLAC.

A headphone-based quality check of the recordings confirmed that the volume was reasonably matched and the channels/phase were correct. A random sequence generator was used to assign each file a number.

Fair Use Statement:
The copyrighted material is being used for research and scholarship. Less than 10% of the total length of each piece is used. Commercial recordings are required for this experiment to reflect the types of music that one would really listen.

Group A: "1 vs 3"
A 1980's pop rock/new wave classic. Instantly recognizable within the first two seconds. The era of analog.

Group B: "2 vs 5"
One of Mark Levinson's direct to DSD recordings made at the dawn of SACD. It's still available as a digital download and the modern era of audiophile pixie dust.

Group C: "4 vs 6"
A selection from the 2017 Academy Award winning soundtrack for La La Land. The modern mainstream era of multitrack PCM recording. Justin Hurwitz graduated from Harvard, has 2 Oscars, and is millionaire, but chooses to use the JBL 3 series in his home studio (and presumably the 708P as he offers a quoted testimonial) and has been photographed with the every-man Sony MDR-7506's. If you want to talk about using the same gear that the composers are using...

Instructions
For the poll, you have to pick a winner from each of the groups. This means you pick 3 choices.

The numbers do make things confusing. This is because the slots were randomly assigned and is necessary to strengthen the results of this experiment.

@amirm @pkane @SIY @Kal Rubinson @dualazmak @MattHooper @atmasphere @computer-audiophile

Edit: Renamed the title to blind challenge. You don’t need to ABX with the formal X. Just listen and decide if you like one or the other or you have no preference.

Edit 2: Thanks to @danadam for these added comments for formal ABX testing

https://audiosciencereview.com/forum/index.php?threads/take-the-blind-challenge-300b-set-vs-straight-wire-with-gain.43983/post-1564124

FINAL EDIT:
Files removed. Answer Key:

Post in thread 'Take the blind challenge! 300B SET vs. Straight Wire with Gain'
https://audiosciencereview.com/foru...vs-straight-wire-with-gain.43983/post-1580424

 

[Sokel]

3,4,5 without scrutinizing anything,totally subjectively.

 

[danadam]

Take the ABX challenge!

This seems more like a preference challange, whatever that means

 

[danadam]

The numbers do make things confusing. This is because the slots were randomly assigned and is necessary to strengthen the results of this experiment.

How? What would change if you named them A1, A2 and B1, B2 and C1, C2?

The recording was then further digitally volume matched according to peak dBFS as calculated by @pkane's DeltaWave.

For group C there's almost 1 dB LUFS difference:

-14.5 LUFS, A1.flac
-14.8 LUFS, A2.flac
-26.4 LUFS, B1.flac
-26.3 LUFS, B2.flac
-14.5 LUFS, C1.flac
-15.4 LUFS, C2.flac

I then cropped the music

Would be nice if they were aligned better. There's about 0.1 second difference in group A and 0.5 second difference in B and C.

 

[Blumlein 88]

A case where using 3 choices each. Two the same and one different might have worked better.

We don't know if some prefer and hear an SET and others prefer and hear no SET.

 

[GXAlan]

How? What would change if you named them A1, A2 and B1, B2 and C1, C2?

Technically, random number generators are based upon the seed, and the seed often is based upon epoch time which increments by seconds. If I ran it three times in a row, I wouldn't really know if the seed failed to increment up or if the repeated random number generated the same value as before.

By generating the whole sequence based upon the seed, it at least is random. You are right -- that I could have renamed 1-6 to Ax, Ay, Bx, By and that doesn't change anything. It was just the method of random number generation that I thought was important and then I didn't opt to change it. Hopefully this makes everyone think a bit harder.

For group C there's almost 1 dB LUFS difference:

-14.5 LUFS, A1.flac
-14.8 LUFS, A2.flac
-26.4 LUFS, B1.flac
-26.3 LUFS, B2.flac
-14.5 LUFS, C1.flac
-15.4 LUFS, C2.flac

Would be nice if they were aligned better. There's about 0.1 second difference in group A and 0.5 second difference in B and C.

It is interesting, because the 1 kHz test tone is very closely matched, but the variance differs between all. The matching was done via the PEAK not the RMS to account for any discrepancy in timing based off a larger sample. I have found that to be more reliable/harder to detect differences from previous ABX testing.

If you look at C1 vs C2 for the musical content and then look around 1 kHz, you can see why it's hard. This is NOT correcting the volume yet. Around 1 kHz, you can see how the area under the curve may be the same, but the blue (which is louder RMS for the recording) may actually be quieter in this range from 942 Hz to 1.1 kHz. It's the non-linearities of the tube amp that make it impossible to match.

You can see the difference here between the cross corellated (match time) and matched (match volume). By increasing the volume of the white, you do a better job matching the left peak, but there are areas where the white now overshoots on the right peak when it worked better before.

Initial peak values Reference: -2.284dB Comparison: -2.284dB
Initial RMS values Reference: -16.216dB Comparison: -17.343dB

Final peak values Reference: -2.284dB Comparison: -1.237dB
Final RMS values Reference: -16.247dB Comparison: -16.273dB

When you make the RMS values equal, then the peaks are different and you might be able to hear the difference in the transient/impact.

You can also see that the PK Metric show they are at the threshold of being audible (a very impressive result once you know the sinad). But my attempt to match the 1 kHz tone is reflected here. The PK Metric is -138 dB at 1 kHz!

I'll leave the recordings as is. If we see a lot of "I can't tell between the first two songs" but there's a clear preference for the louder one, that would still be useful to point out.

 

[Sokel]

My take on this was solely to what was more enjoyable the little it lasted (non of this is my favorite music gender but it's ok).
No searching for "differences",etc.
One useful thing to this test is for younger friends here who have not heard what these things sound like and have imagined monsters coming out of tubes.
Nope.

 

[Blumlein 88]

The numbering seems unnecessary for this kind of test. Listening over Babyface Pro with Sony MDR 7510 headphones. Entered my votes in the poll. I didn't do any abx type comparison. Just listened a few times and picked one.


My preferences: Click on the cloudy spoiler.
I picked 3, 2, and 6 as my preferred versions

 

[Axo1989]

Comparing 1 vs 3 I've no clue what the 'instantly recognizable' track was—before my time I guess ...

Spoiler: the rest is spoilered for those who like a clean slate.

... but 3 has a hum audible before the music starts, which I didn't notice on 1. I have no preference, the music is execrable either way. Is 3 a better reproduction because I can hear the hum on the recording, or is the tube amp adding the hum?

Comparing 2 and 5 the track was interesting. 2 had a hum and a high-pitched ring, 5 sounded more clear/open. I could understand people preferring 2 for warmth/character but I preferred 5. This time I'm assuming the hum/ring isn't in the recording.

Comparing 4 and 6 I'll preface by saying I lasted maybe 10 minutes into La La Land (but I liked the sonics of Chazelle's First Man and haven't yet seen Babylon) the music is almost as hard to spend time with as the first pair. Anyway 4 sounded brighter/sharper and 6 a bit more mellow/rolled off. Prefer isn't applicable but I'll say 4 was a more complete reproduction.

JA's measurements were interesting assuming applicability. I've never heard a tube amp intentionally, it will be interesting to see if my perceptions match in any way! I'm guessing it's 3/2/6 ... unless it was actually masking the hum/ring I didn't hear on 1 or 2 which would be perverse but interesting.

Edit: tried speakers first but it's late so didn't turn them up, comparison is listening from M2 MacBook Pro via AirPods Max.

 

[pkane]

It is interesting, because the 1 kHz test tone is very closely matched, but the variance differs between all. The matching was done via the PEAK not the RMS to account for any discrepancy in timing based off a larger sample. I have found that to be more reliable/harder to detect differences from previous ABX testing.

If you use DeltaWave to compute the difference for the C group, you'll find that it's about 0.93dB. Matching on peaks is probably the least reliable way to match levels. RMS level is next, LUFs is still better, and what DeltaWave is doing is probably still better, since it tries to eliminate all the linear differences between the two files based on all the samples in both.

 

[computer-audiophile]

Once the fine sound of a good 300B amplifier is put through the digital mill, it is ruined.

 

[Axo1989]

Once the fine sound of a good 300B amplifier is put through the digital mill, it is ruined.

Haha, maybe. I certainly wouldn't know!

My playback was about as digital as it gets.

 

[danadam]

I've no clue what the 'instantly recognizable' track was—before my time I guess

 

[Axo1989]

Iconic prog rock then, I'm ashamed of my ignorance.

 

[Blumlein 88]

Once the fine sound of a good 300B amplifier is put through the digital mill, it is ruined.

Ridiculous.

 

[pkane]

Once the fine sound of a good 300B amplifier is put through the digital mill, it is ruined.

It's a good thing that no recordings today are done or mastered using digital... right?

 

[Sokel]

It's a good thing that no recordings today are done or mastered using digital... right?

The ones I love has been mastered with tubes and monkey coffins
(and that's what I consider the modern ones!)

 

[pkane]

The ones I love has been mastered with tubes and monkey coffins
(and that's what I consider the modern ones!)

I prefer wax cylinders myself, for their unparalleled sound quality and for being completely un-ruined by digital.

 

[Sokel]

I prefer wax cylinders myself, for their unparalleled sound quality and for being completely un-ruined by digital.

Tried them but the cough of the audience is not as sharp,time conditions are not as realistic as they should be for the post WW1 people who was suffering diseases .

(ok,I'm going by myself,no need for violence)

 

[pkane]

Tried them but the cough of the audience is not as sharp,time conditions are not as realistic as they should be for the post WW1 people who was suffering diseases .

(ok,I'm going by myself,no need for violence)

Audience cough must be realistic. If I can't diagnose the exact illness of the person coughing from the sound, the recording is just not worth listening to!