Oh shit! SINAD can suck my ****s

Could it be that it is a defective L30?

 

Might be worth establishing whether you are measuring the behaviour of the amplifier or of the load. Perhaps try a measurement of the Jotunheim 1 with a small series R between output and measurement point.

 

the l30 uses the tpa6120a2 with local feedback in a composite circuit with an op amp. As such the output impedance is a complex dance between the op amp, the tpa and the compensation components. I wonder if a simpler lme49600 design might behave better. Although it is also used in composite mode with an op amp, there is no local feedback around the lme49600. It is simply a current buffer.

marvey: do you have an atom handy?

 

I am going to suggest the L30 is unstable with many complex real world loads. Best measurements from resistor or planar loads. Otherwise lookout. Not a headphone amp I want in the lab.

Here are a few quick measurements using the amp currently on the bench at the moment, SW51+ with EF805S tubes. At 0 dBu expect a high THD+N which will be 2nd harmonic dominant. When using HD650 as a load the sound level is LOUD. Hearing protection was used.

Upper trace in graphs below is sweep of THD+N
Lower trace in graphs below is sweep of 4+HD+N

SW51+ with EF805S tubes 0 dBu 300R load


SW51+ with EF805S tubes 0 dBu HD650 load - free air


SW51+ with EF805S tubes 0 dBu HD650 load - on head (with hearing protection)


There are some changes but remarkably similar between the three load conditions. Biggest deltas appear between 500 and 1000 Hz.

And SINAD and 1 KHz THD+N be damned.
SW51+ produces damn fine sound to my ears with Sennheiser HD6x0 and HD800 headphones.

 

That's the problem with using a distorting load. Any amplifier output impedance forms a simple divider between the amplifier and its relatively low distortion and the load and its massively higher distortion. As long as the output impedance isn't zero measured distortion will increase purely from the load's contribution, even if nothing changes in the amplifier. A passive network, rather than a transducer, would contribute much lower distortion, though even then the inductors in the network are the weak link. It may be worth comparing the distortion patterns of the headphones you have tried (at the same test level) with the patterns seen in the distortion increase when using them as the load. If they show the same shapes that would suggest the load is the main (possibly the only) contributor.

 

I'm a biologist, and you see the same kind of instability in nested feedback loops inside cells/tissues/organs. They work great under control conditions, but any kind of pathological input and it goes haywire. Too much 'infighting'.

I bet a simple/single global negative feedback loop will fare better into complex loads than these fancy pants nested designs. Zero NFB like the SW51+ will do better again. Zero NFB with a high z-out that can 'soak up' some of the back EMF, and the effects of the load will vanish.

 

I'm going to expose my limited understanding of most things here, being a more technical discussion. But wouldn't there be several things related to back EMF here? resonant frequency of the driver, the size of the driver (and correlated half-wave dimension, the frequency at which the driver most efficiently couples with the air around it) as it relates to the volume of the headphone cavity between the ear and the driver (and, in sealed HPs, the cup volume). and on top of that, the ability of the amp to handle impedance swings caused by this? IOW, where "big iron" transformers shine in their ability to lower distortion in higher impedance swings caused by driver/back EMF issues are the same as where SS amps that can't pass big current or deliver high damping factor kind of fall apart/lose their balls? Or am I totally misunderstanding back EMF and amp interdependence? Then there's the whole BWC idea that the larger the stroked volume over VC displacement gives the amp a fighting chance at not having to deal with huge impedance swings below the resonant frequency of the driver. Or am misunderstanding things? If it's not obvious by now, I have now formal EE training. LOL.

 

@Marvey do you still have HE4XX? That had much higher distortion in the lows. A quick test of @JohnM 's suggestion would be to see whether that (or some other planar with high bass distortion) when used as a load for the L30 still gives a pattern like HE5XX, or whether it looks more like its distortion.

 

That's the one which makes the most sense to me. More sample size of higher and lower Qts transducers will conclude this hypothesis.

Maybe test smaller orthos like the t50 vs some Hifiman 4series. Or that stupid big Chifi dynamic you have (still?) vs a Grado or Elex.

 

Ortho dynamic headphones present a mostly resistive load to an amplifier.

Here is my HE-500 impedance:

40 ohms across with the slight bump to 42 ohms between 40 and 80 Hz.

 

Perhaps this explains the common corollary to HD650 scalability, namely that many planars are not amp picky, provided they get sufficient current.

 

Great stuff!

Slight re-write: we can't say it's the headphone's distortion doing this to the measurement. Although I'd still like to see something like the AKG K7--: a dynamic with a fairly flat (and lowish) impedance through the bass & mids but a bass distortion profile like (going from Tyll's measurements, somewhat worse than) HD650.

 

Correct me if I'm wrong, but if the load is the main/only contributor, wouldn't the distortion be largely identical across all amplifiers?

 

If you think that SINAD is not telling even the half truth about amplifiers (and I'm with you), I suggest you read those two PDFs taken from here:
https://www.diyaudio.com/forums/sol...-books-overview-google-books-post6707846.html