Discussion in 'Headphone Amplifier Measurements' started by atomicbob, Oct 6, 2019.
Thanks for added measurements and comments, quite illuminating.
And where else on the internet can one find such extensive, detailed information on a relatively obscure headphone amplifier?
Support your local SBAF!
Additional discussion of SW51+ output voltage vs. distortion.
Maximum output voltage is dependent on load impedance and distortion level tolerated. A much higher Vpp (Voltage peak to peak) may be achieved if greater than 1% distortion is considered.
Several power measurements are presented in post 20 here Voltage output for a given distortion may be determined by reading the dBu on the x-axis for a given THD on the y-axis.
Distortion vs. amplitude plots distortion in either percent or dB below the stimulus on the y-axis and amplitude on the x-axis. My power measurements cited above were made with amplitude in dBu which is an absolute: 0 dBu = 0.775 Vrms. Using the formulae below dBu may be converted to Vrms and then Vpp:
For a 1 KHz stimulus the left channel (blue) measures 0.2% THD+N at -10 dBu. Using the formulae above -10 dBu calculates to 0.245 Vrms and 0.693 Vpp.
For a 1 KHz stimulus the left channel (blue) measures 2.0% THD+N at +10 dBu. Using the formulae above -10 dBu calculates to 2.45 Vrms and 6.93 Vpp.
Bifrost-2 + SW51+ system voltage vs distortion oscilloscope measurements at 440 Hz
DAW 440 Hz 0 dBFS sine output to Dante Virtual Soundcard
Dante RDL spdif output routed to bifrost-2
bifrost-2 RCA 2Vrms to SW51+
SW51+ HiZ output with 300R load substituted for HD800-Jmod
Signal routing diagram:
SW51+ volume knob setting (above) for 1% THD when driven by Bifrost-2 440 Hz 0 dBFS 2Vrms - SW51+ output 3Vpp (below):
FFT associated with SW51+ 440 Hz 1% THD at 3Vpp:
SW51+ volume knob setting (above) for 7.5% THD when driven by Bifrost-2 440 Hz 0 dBFS 2Vrms - SW51+ output 18.4Vpp (below):
FFT associated with SW51+ 440 Hz 7.5% THD at 18.4Vpp:
Explanation of corrections from impressions thread.
How mistakes happen: complicated manual system setup + multiple serious interruptions.
In the impressions thread I had incorrectly stated the SW51+ gain was 1.72 as measured on the dScope. However the picoscope measured approximately 3.26 as distortion was rising. There should be a change but typically the gain drops as distortion rises. Here is how the error occurred.
dScope setup for fully balanced measurements but output derived from unbalanced BNC
Generator has been set for 2Vrms
Note the analog output mode set to Balanced normal - this panel is not usually displayed - it is displayed here to make the point
SW51+ L ch output measures 3.43Vrms
Gain incorrectly determined as 1.715, which is 4.7 dB
Changing analog output mode to unbalanced with the same generator level of 2Vrms
Now SW51+ L ch amplitude measures 6.6556 Vrms
Gain correctly determined as 3.32, 10.4 dB
Here is the gain linearity sweep for -60 to +20 dBu
The gain drops as expected, but only significantly so above approximately +10 dBu.
The dScope output mode either derives unbalanced BNC as half of balanced XLR output or corrects by 6 dB depending on the mode setting. I thought I could pull off a quick manual mode measurement. Bad assumption when the day is likely to be filled with interruptions.
Any chance this has made it up towards the top of your queue ...?
Edit: above photo needs Asian representation.
Not yet. At this time I only have two types of pentodes for SW51+: the stock 6Z51P and Mullard EF184. I should probably acquire a few other models to make the comparisons worthwhile.
It appears multi-tone measurements are in-vogue FOTM metrics of discussion across the intenet. Here are some SW51+ multi-tone measurements for consideration.
First up is a dScope analyzer loopback through a Goldpoint SA2 for reference at approximately -5 dBu.
The above represents ideal response with no extraneous spurs between the 31 tones (AES standard)
SW51+ with CGE6EJ7 tubes adjusted for output at -5 dBu with 300R load - 12 tone analysis for simplified visualization
In the graph above the gold line depicts frequency response for the 12 tones while the blue line highlights spur peaks for intermodulation distortion.
SW51+ with CGE6EJ7 tubes adjusted for output at -5 dBu with 300R load
-5 dBu will produce approximately 95 dB SPL in HD6x0, HD800, HE-500 headphones. Listening at that level for long sessions will invite early adoption of hearing aids. Intermodulation distortion above 1000 Hz is approximately 65 dB below the stimulus tones.
SW51+ with CGE6EJ7 tubes adjusted for output at -10 dBu with 300R load
-10 dBu is approximately 90 dB SPL for the three headphones mentioned. Note how IMD is dropping.
SW51+ with CGE6EJ7 tubes adjusted for output at -20 dBu with 300R load
Above 1000 Hz IMD is below the graph lower limit. If 0 dBu relates to 100 dB SPL for the three headphones mentioned, then -100 dBu relates to 0 dB SPL. Keep that in mind when evaluating this and other measurements. Relating distortion measurements to use case sound levels is extremely important to maintain useful perspective rather than auger into the ozone on performance metrics that are interesting, but not likely to be perceived in actual use. As mentioned earlier distortion sweet spot for this amp is about 80 dB SPL, which is still very loud for long listening sessions.
Same measurement as above but I have added the 80 phon line from the ISO contours of equal loudness, also for perspective.
If calibrating to 80 dB SPL headphone loudness, then the 80 phon line is applicable to this use case. Greatest Human Auditory System (HAS) sensitivity is between 2 and 4.5 KHz. Now notice where the IMD appears on the graph. Hint: it's not there. Also note how HAS sensitivity is reducing as frequency lowers from 1000 Hz down. Those IMD spikes below 1000 Hz are likely to be masked, though they may contribute to a perceived slight fuzziness when compared directly to one of the EC or DNA DHT amps. Without that direct A/B comparison, it is highly unlikely most listeners will notice.
CGE6EJ7 tubes appear courtesy of @gixxerwimp - more distortion measurements coming but I don't have a firm schedule at the moment.
Hey @atomicbob I’ve got a pair of Telefunken EF184 that I can send.
Separate names with a comma.