"Am I doing it wrong?"
Short answer: No.
Long answer: No, but it's complicated.
The HD650 is a weird headphone. I tried feeding it a signal¹ with a very wide dynamic range (i.e. not compressed with tubes or other methods) from an amplifier² with ample power to drive the sub-bass region that's 600+ ohms in the HD650, and there was no bass. I don't mean it was weak bass, I mean it appeared to roll off at an embarrassing ~150 hz. So I tried to EQ the low end to a higher relative level, which immediately made it so much WORSE that I couldn't listen for more than a few seconds. It didn't just fail, it failed spectacularly. It failed to sound good in precisely the way a $1 speaker from Radio Shack fails to sound good, it seemed overmatched and outclassed in every way.
However, the 650s reacted very differently from other sources. I plugged them directly into my Hifiman HM601 player, and with the same track playing the 650 sounded fantastic! The bass wasn't the boomiest, but it was boomy enough to be satisfying, and it extended well past 50 or 60 hz, it was easily low enough to present the whole track, not just the 2/3 of it I got on my good system. With the famous HD650 warmth, it sounded so good there that I strongly considered keeping the 650, just to pair with that player.³
So what is going on here? Why would I get such radically different results with the same pair of headphones? If you know, please do share! All I have is a guess.
My guess (which doesn't actually make sense, but it's the best way I can describe what I'm hearing) is that some DACs--the one in the Hifiman player and the Modi Multibit both exhibit this, for example--will not only compress dynamic range, they also compress the frequency range, at least at the bottom end of it. A 50hz wave appears to be brought up to 55hz. Those numbers are made up to illustrate my idea; maybe it's 50.0hz to 50.2hz, whatever. This doesn't seem realistic because it would then be off key without a software pitch correction, correct? And what would be the point of adding an expensive DSP chip to a product that wouldn't need it if it just did a straight up conversion?
That is what I'm hearing from those sources, though. Instead of clipping off the ends of the frequency range and tossing them out, they get squeezed up into a narrower one. It could make aural and economic sense in something like a cheap Bluetooth headset, where the driver's limitations are known and you already have the onboard processing power to effectively optimize around it. The result would be the ability to convey the sub-bass information present in a recording through headphones that can't reproduce anything below, say, 70hz.
My description of the result seems right, even if the only method I can imagine that would achieve it sounds crazy. But whatever actually is happening here, it isn't necessarily a bad thing subjectively, considering how much I enjoyed EDM on an HD650 in the one instance I did enjoy it.
-------------------
¹ & ² - I was running a hub-powered ODAC to a 45Wpc Kenwood integrated tuner, driving the headphones directly off the speaker taps. I've found this setup to be very good at driving power-hungry headphones, from planar magnetics (Alpha Prime, HE400/500, LCD2) to high ohm dynamics (250 ohm DT880 pro, HD800, even down to moderate impedances like the AKG Q701, though the last one has a bit of hiss). Not a single one of the above has even a hint of the problems the 650 demonstrated, nor did any other I've tried, including an ATH-M50, a Koss ProDJ100, and whatever else I've played around with. TL; DR the problems observed in the HD650 are unique to it.
Using a speaker amp in this application works because those 45 rated watts are with 8 ohm load. I [very roughly] estimate it's max into a 300 ohm load to be maybe a watt, but the important thing here is that the volume pot is comfortably up into it's nominal range.
³ - I didn't keep the HD650 the end, because the Q701 sounds just about as good from the HM601, and it doesn't also have the severe limitations that the 650 suffers from.