Discussion in 'Blind Testing and Psychoacoustics' started by purr1n, Dec 8, 2019.
It is a fault light, well at least on x1 gain, or when I'm not tired.
Two more comments, then I'll stop crapping up your thread:
Feedforward, AKA "error correction" is definitely not new. Sumo used it in their amps from the 1980s, and their implementation was based on earlier work by Malcom Hawksford and Bob Cordell. The difference between conventional feedback and feedforward is that you're only feeding back the error signal, not the entire signal (and, in Sumo's case, the error correction was confined to the output stage.) Bob Cordell's paper goes into great detail why it is used and how to implement it in a discrete output stage: http://www.cordellaudio.com/papers/MOSFET_Power_Amp.pdf
And yes, you can have both feedback and feedforward. Sumo amps used multiple feedback loops and feedforward. Magni Heresy uses a ton of feedback, much more than Magni 3+. Magni 3+ has very limited open-loop gain--certainly nowhere near the 114db of the OPA1662 and 130dB(!) of the OPA1688.
Below, I've marked the "area of interest" for THD+N vs Vrms for levels (volume position) set for the ZMF Verite. The reason I wanted to show this is to demonstrate why a singular SINAD number (ASR methodology) at seemingly arbitrary output levels of 4Vrms, 2Vrms or whatever is completely meaningless.
Even if absurdly good SINAD numbers actually meant something, we still need to consider the follwowing:
how loud a person listens at
how efficient a headphone is
the impedance of a headphone
the gain setting of the amp
and even what types of recordings (compressed loud K-pop vs. classical).
The Heresy does have about 1/2 as much distortion compared to the Magni 3+. That's 6db better.
Magni 3+ high gain 300-ohms
Magni Heresy high gain 300-ohms
I wonder if the whole headspace thing is more about phase shifts, than it is about crosstalk.
Now that the cases have been covered for a 300-ohm dynamic driver load, are you planning to move on to something different like the Verums? I know you've recommended the Fulla for Verum before, and it does make sense considering op-amps are designed assuming a resistor load.
Personally I suspect that the quick rise of planars is partially due to their offering a more consistent performance between amps/sources with the same amount of power. Would be interesting to see if the differences become less or more pronounced.
A lot of folks (particularly black & white "objective" minded ones) are probably going to miss this, but a THD+N single number is more of a sanity check. A much better and complete picture comes with a THD+N vs Vrms sweep, which cover the linear operation range performance.
A THD+N 1 kHz PSD at a particular level is more of a trouble shoot plot if something is off in the THD+N vs Vrms at a particular level.
LOL. Remember, this is the guy who invented Amir-bits, where ENOB is defined at the level where linearity deviates 0.1db, this compounded by the fact that he didn't implementant a narrow bandpass at the 1kHz he was looking at. He later did, but didn't go back to revisit, correct, remove, or indicate the circumstances of the early results. So much for science.
The unfortunate thing is that most readers won't know which results are borked and which are not (post Amir being schooled in linearity measurements by Jude / AP tag team).
I really wish he did. By doing half ass corrections to his characterization, he is doing quite a diservice. He tends to leave his mistakes at the front page, and then an obscure link at the bottom to some random post later with corrections.
It's lazy. This is the dude that wants the analyzers to have an "easy button".
I'll do that if and when I get back from jury duty. Already got excused from a long 10+ day civil trial with six lawyers on each side. I probably would have been dismissed anyway because I thought both the defendants and plaintiffs looked like disagreeable assholes and would have stated so. You know, the wrinkly white guys in suits so tight that their faces were red from lack of circulation. If I were judge and jury, I'd give them swords to settle it and watch for my amusement.
Now I'm back in the pool to see if they need me for the second trial. I not sure if the "sucks to be you" and "my dad taught me that everything was always my fault" outlook is suitable for being a jurist.
The judge was charismatic enough though and almost convinced me to not plead hardship (no vacation time, aging parents, medical bills, ongoing illness)
Ok. Tried with RAD-0. Generally, the same impressions. By this time, I've become very accustomed to the sound of the amps, which in the larger scheme of things really isn't all that different, except for Heresy at unity gain which I dislike. I think part of this is because the RAD-0 doesn't do the headstage thing as well as the Verite, so the differences in headstage between the two headphones isn't as large as with Verite.
Getting JDS Atom in, so might be interesting to see if I can tell the difference between that and the Heresy.
Archel 2 or 2.5 are others to consider if anyone wants to do a battle of the IC-based headphone amps, though only the 2 is under $150.
I just tried this with my HD600 and you are absolutely correct - at the same volume level, the Heresy sounds much better in the high gain mode. In fact, I deliberately tried setting the volume output higher in the low gain mode and it still lost to the high gain mode. Basically what happens is that the amp becomes significantly less dry and sterile sounding in the high gain mode and the soundstage opens up. It is still not as holographic or meaty as the Magni 3+. Btw, I had to ship my Magni 3+ back to the mothership because it had an intermittently faulty power switch. What are the odds? I am missing that amp.
My "right arm of the free world" is ready and willing. I have some tracer rounds that should look good on video.
That's one for @LetMeBeFrank. I hope you guys how much of my temple-of-hifi or jamming-with-my-son time I am missing as a result of my "reviewing" and blind listening to all this crappy souless 100db feedback 120db SINAD gear.
I hope that @schiit will come out with a $99 amp that is the complete opposite of Heresy. I know for a fact that the discrete Magni 3+ could have been much more, but that Jason decided to go for measurements on the 3+ as well. Yes, I would like a Magni 4 Anti-Heresy with a mere 95db SINAD that sounds great.
enjoyed the Magni 3+ as well with Auteur and Code-X.
but really - not opening my kid's piggy bank to steal quarters for a mini SS amp unless the SINEAD is exactly 101dB.
and yes, i meant to type "sinead".
because- nothing. compares. 2 u.
Was this mainly a pain because of inconsistency with small pots? Since the specs are so similar I'm wondering if you could have set both to max and used a passive preamp.
Small diameter knobs is part of it, but matching to 0.0005 mV is hard with any mechanical analog device.
We should have a subject ranking named SINEAD. Just need to figure out what the letters stand for.
In retrospect, it's obvious that song was written by Prince.
The inherent gain for the two are just different enough that this doesn't work for both. Max volume setting (low gain for both) results in 0 dB gain for Heresy while producing approximately -0.8 dB for Magni 3+. That difference is easily heard by those with advanced listening skills.
In my ABX I testing I set Magni 3+ to max volume and adjusted Heresy until the match was within 0.05 dB.
And yes, that is a bit challenging with small pots / knobs. One also has the pleasure of dealing with the backlash when letting go of the knob. Makes me long for the days of 8:1 or 10:1 Vernier knobs.
Seriously In Need of Electrical Acoustic Distortion. Make it SINAED and A can become Added.
My condolences on the new Atom. That amp nearly drove me out of the hobby. I put music into it & the output was as enjoyable as listening to a sine-wave. It was ok with a NOS DAC but these chip-amps are abysmal in the budget systems they'll end up in 99% of the time, systems where you want some coverup & don't want an amp that empathizes glare.
Separate names with a comma.