Schiit Ragnarok distortion analysis

I initially wanted to see how much noise the Ragnarok really produces on its own.
With inputs shorted and on high gain I got a figure of around 3.5uV at a 44.1kHz SR. That would be fantastic and I might've made a mistake somewhere, but the OG Ragnarok is a very quiet amp, that's for sure.
I got a maximum output voltage of 20V_RMS, which makes sense given the 100W 4 Ohm rating. 20V*5A.
That should make for a DR of 135dB, which would be crazy good. Again, maybe I made a mistake somewhere.

Unfortunately I couldn't replicate the low OI. The output did drop a bit under speaker loads. Just something I thought I should mention.

Anyway, the more interesting thing turned out to be the distortion under load. We normally don't see amplifier distortion measurements under typical loads. Heck, if we're lucky we might see distortion measurements under resistive loads or special speaker simulation loads. But that's rare.

The Ragnarok might not be so easy to measure and even the sweep may distort things a bit. In order to get as low distortion as possible I played music for a bit and then went for the shortest sweep possible, a 512k sweep at 192kHz sampling rate. A 4M sweep definitely caused the distortion to rise. Even so there was some variation between different measurements.
Here's what I got:

NO LOAD (the scale is intentional):


AMON AS LOAD (even lower minimum impedance)

0dBV corresponds to a level of roughly 90dB at 1m.

We can see that it's not so happy running the large OB speakers. In fact I think the D3 in the lower midrange could be higher than the speaker distortion at this level. The D2 sub 40Hz probably, too.
So far I've only measured them with the Ragnarok, actually. I should confirm with my measurement amplifier.

HATHOR AS LOAD (3 Ohm dips)

15dBV corresponds to roughly 90dB at 1m. Even so it doesn't sound so bad to me at high SPL. But the Ragnarok is indeed no amp for low impedance, medium efficiency speakers. Going by my own measurements this is more distortion than the speakers at this level. Maybe it's just a Ragnarok measurement artifact.

I intentionally did 15dBV so it corresponds to the same volume level. Even just 5dB quieter it's much happier. See the attached measurements. At 0dBV it's really quite good.


Would be interesting to see similar measurements for other amps.

 

I wanted to see for myself how audible this kind of distortion really was, so I recorded a couple songs with and without Hathor as a load at a level that corresponds to the same 15dBV at 0dBFS that you see above. Note that I posted the left channel, the right channel had less D2.

The snippets were recorded with the MOTU M4 as a DAC and ADC, just the Ragnarok in between.

Free Dom

They are level matched at 1kHz, but not corrected for the slight FR difference.

 

I thought it was a bit weird that the Ragnarok was specified to 2Hz -3dB only. My REW measurements (which start at 2Hz) were absolutely flat from 2Hz on. And its design should allow it to virtually amplify DC.

It turns out it does amplify DC!


This is from connecting a 1.5V battery and then disconnecting it again. DC bias did drift somewhat, but it will get adjusted down to 0 again.
Speakers were connected for this test. I did a couple tests and it seems for it to work like this the amp needs to be at thermal equlibrium. 15s is quite long, but it just keeps going. Not sure what's the limit.
Battery doesn't get warm, nothing.

With music the DC offset on the output does get frequently adjusted to zero. It works!



Kinda cool to watch how with music the DC offset keeps drifting a mV or so and then gets hammered down to zero.


Edit:
One advantage of it being able to pass and amplify DC is that the phase response in the bass is flat.
This kind of design is nuts for sure.

 

You might be on to something. The rise towards 20Hz does look suspicious.
Could it be a limitation of my measurement setup?

For reference, here's the impedance of the two speakers. Unfortunately VituixCAD doesn't allow plotting on a log axis.

Amon:


Hathor:


EDIT: Prettier impedance graphs attached.