Made measurements of various tubes using an Audio Precision

[Maxhawk]

I decided to take my Crack into work today to do some measurements with an Audio Precision SYS2722. I made frequency response and distortion measurements using a given input tube with different output tubes, and also a given output tube with different input tubes. The AP has 3 selectable input resistances of 300, 600, and 100kohm. My input signal level was 1Vrms in all cases and I used the volume control on the amp to adjust the output voltage. The level of all the curves is normalized at 1 KHz. The graphs were all created in Excel using data I exported from the Audio Precision. I didn't do any extensive tube warm-up -- I only waited for the output level to stabilize before making measurements.

The various input tubes I tested were:

• Sylvania Gold Brand GB5687
• CBS-Hytron 5814 black plate with angled getter
• RCA cleartop 12AU7
• Sylvania 12BH7
• Tungsram E80CC

I switched the bias resistor for each input tube to give the proper bias for the output tube.

The output tubes I tested were:

• Bendix/Chatham 6080WB graphite plate
• Tungsol 5998
• Sylvania Gold Brand GB6080
• RCA 6AS7G
• Mullard CV2984
• GE 6080

The following 2 graphs show a 5814 with various output tubes. Output level is 1Vrms into 300 ohms. The frequency response barely changes with output tube. At 20 KHz the output varies by only ~0.1 dB.
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FZWxJTca.png&hash=37e508d9b5fbb80f1233bca0ffafbe21519204af)

The following is distortion of the 5814 with various output tubes. The 5998 has the lowest distortion while the Bendix/Chatham 6080WB graphite plate is the highest. I suspect the little wiggles are a measurement artifact.
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FFWphjsJ.png&hash=64ec3e84a6766e7870c6462d4a02dbe4257c2cfe)

This is the 6080WB frequency response with various input tubes:
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2F7wzjlZ1.png&hash=227e0a6ee490e9a0f32b4ba41d54e4b72b501683)

This is the 6080WB distortion with various input tubes:
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FZr3Pz8x.png&hash=0944ebe400504e1cd8fdcddab5bcdb8c56172ea1)

The following 2 are of a 5814 + 5998 at various loads. This shows that the lower frequency response is a function of the 100uF AC coupling cap.
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2Fv6UFg4w.png&hash=c74b1bb18da6ede6de919670b483076ac27a0f0d)
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2Fz0wniin.png&hash=7040ccdb6783fc7bf94e35e2107beb0a7ebed2fa)

This is frequency response of the 5687 + 6080WB at various output levels:
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FuNUIqM8.png&hash=c344d420353dd27e9158e7528f354e177ca6311e)

And distortion at various output levels:
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FbAaZFxx.png&hash=ac6d2a54866734be81155a1c87eac8002720a64e)

And finally an FFT of the 5814 + 6080WB. Distortion is primarily second order. It looks like my power supply would benefit from some additional filtering as the main and 1st harmonic are relatively high level.
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FtP0bJ5c.png&hash=088398b96af54b3a4c37e2d1c6f6a36e74544648)

I've learned that my favorite musical tube combos have higher distortion than many other tubes. I'm happy to see such little variance in frequency response vs tube selection. I'm surprised to see distortion to be directly proportional to output level.

[mcandmar]

Fascinating results, thanks for sharing.  Looking at the distortion levels of the output tubes it pretty much matches my order of preference, 5998 > Mullard > GB6080 > everything else. I'm surprised with the Graphite Plate results as i always found those to be very clean sounding. Would be interesting to see how a GEC and 7236 measure up too.

[JamieMcC]

Thanks its interesting to see some measurements on the different tubes.

[hjc]

Thanks so much for going to the trouble of sharing your findings. Nice to have some real numbers.

[Grainger49]

Very thorough.  Distortion is a bath tub curve.  It rises at very low outputs, probably too low to drive headsets and rises with output.  It usually has a flat part of the curve first then as you hit the limit it rises steeply.  It looks like a diode turn on curve.

[Maxhawk]

Any thoughts on what output level is typical for normal listening level? Obviously it varies with headphone efficiency and impedance. I chose 1 Vrms since it's a standard for line level. Tyll reports that the HD800 requires 254 mVrms to achieve 90 dB SPL, so for my own setup, the distortion is a good bit lower then what I've measured at 1 Vrms.

@mcandmar: I find my graphite plate 6080WB very musical as well and it's one of my go-to tubes, but I do feel like I can hear the increased distortion. I have the solid plates and wonder if the slotted plate measures/sounds different. I'd measure a GEC and 7236 if I had one .

Just for fun, here's what the SYS-2722 looks like looped back to itself. This is with an unbalanced BNC connection as used with the Crack. The XLR balanced connections may measure even better.
(https://forum.bottlehead.com/proxy.php?request=http%3A%2F%2Fi.imgur.com%2FijPuSzS.png&hash=ec3bf2c74c5efd614244d15e4515c2eadb92de63)

[atomicbob]

nice work! I've been meaning to do something like this with my dScope but the day job (acoustic research) keeps me too busy.
I have been monitoring listening levels for visitors to my lab. There seems to be a tendency to listen between 85 and 105 dB SPL for averages with peaks 10 dB SPL higher. My own listening habits are 75 dB SPL average with peaks to 85 dB SPL for music with 10 dB of crest factor. With my HD800 that measures at 24 mVrms average with peaks to 77 mVrms.

[ALL212]

These are so cool!!!      

[Doc B.]

The average listening level we usually plug in to our calcs is 82dB, which I got from my TP partner/mastering engineer's rooms.

What might be interesting is for you to compare the FFT distortion spectra of the various tube combinations. My hunch is you will be able to even better correlate the differences you are hearing to the shift in harmonic content from the various combinations than to the relatively small differences in the THD graphs. Back in the early days of Bottlehead I had the loan of a very nice HP analyzer and learned a lot about what distortion spectra correlated to circuits whose sound I liked.

[Maxhawk]

The average listening level we usually plug in to our calcs is 82dB, which I got from my TP partner/mastering engineers.

That works out to 254mV * 10^(-8/20) = ~100mV.

Calculating for some popular Crack phones @ 82dB SPL:

HD800:  101mV
HD600: 92mV
HD650: 82mV
Beyer T1: 138mV
DT880: 205mV
DT770: 236mV

The last 3 are 600 ohms so their distortion will be lower than what I've measured at 300 ohms. Looks like most folks will operate where distortion is minimal and no where near the >=1% that I measured with a bunch of combinations.

[Paul Birkeland]

Another fun one to do is to pick your favorite tube combo, then do a before and after measurement with the Speedball