Tear down of Neurochrome HP-1 High Perf Headphone Amp

[amirm]

This is a tear down of the Neurochrome HP-1 High performance Headphone amplifier. My review of HP-1 showed exemplary distortion and high power handling. With kind permission from its designer, here is a look at what is inside.

Here is a lower resolution shot of the PC board:

For 4X higher resolution, click on this attachment:



You may have seen my tear down of massdrop THX AAA 789 amplifier which had plenty of empty space. Not so here. As you see, nearly every inch of space is used.

Unlike the massdrop THX 789, the power supplies for the HP-1 are internal. We have two switching supplies from the reputable Mean Well Taiwanese company. I am assuming one provides positive and the other, negative supply rail. If an external switcher is used as in THX 789, then you have to generate the negative rail internally which can be of lower quality. I am happy to see the two thermistors (red and black) for surge protection together with fuse protection. Nice bank of capacitors and inductors clean up the output of the switchers. To adjustable regulators (LM337 and LM317) provide lower output voltages (?).

Looking at the block diagram, we nicely see the instantiation of the same on the PC board:

The only difference is that the DC servo (feedback) loop is implemented using OP2197 rather than OPA2140 as noted above. Or else my eyesight has gone down the hill more than I think they have.

Normally the feedback loop around an amplifier is passive (resistors and capacitors). Here, an active "DC servo" is used in the form of OP2197 (2140 on block diagram) to null the output DC. I believe this is a similar circuit to what is used in my Audio Precision analyzer.

Everything is doubled here to get more power by paralleling the two amplifier stages (OPA1611 and LME49600).

The OPA1612a opamp is used both for differential input, and the gain stage.

Nice to see proper protection circuit to guard against DC into headphones.

Since Tom is here, I am going to stop pretending I have fully reverse engineered this unit from just a picture and let him explain more.

Here are specs for some of the key ICs used:

LME49600: http://www.ti.com/lit/ds/symlink/lme49600.pdf

Notice the incredibly low THD+N rating (of course the challenge is to make it perform this well in a real circuit).

OPA161a: http://www.ti.com/lit/ds/symlink/opa1612.pdf

OPA2197: http://www.ti.com/lit/ds/symlink/opa197.pdf

Note how nicely lined up and organized the parts are. Quality of assembly and soldering is high. Likewise, happy to see top brand capacitors (Nichicon).

 

[restorer-john]

Likewise, happy to see top brand capacitors (Nichicon).

The main caps are Panasonic from what I can see. A few Nichicons. Good quality all round.

So do we have another product with an IEC three pin panel socket where it is not actually earthed? (The earth tag appears to have nothing attached to it from your shot) I see from the timelapse build video, it has a strap that goes down to earth.

 

[amirm]

Indeed: https://www.neurochrome.com/hp-1-ultra-high-end-headphone-amp/hp-1_buildprocess/

 

[derp1n]

Interesting that the thermal pads below the various chips are sufficient for such a powerful amplifier and heatsinks are only required for the LM337 and LM317 pair.

 

[tomchr]

Unlike the massdrop THX 789, the power supplies for the HP-1 are internal. We have two switching supplies from the reputable Mean Well Taiwanese company. I am assuming one provides positive and the other, negative supply rail. If an external switcher is used as in THX 789, then you have to generate the negative rail internally which can be of lower quality. I am happy to see the two thermistors (red and black) for surge protection together with fuse protection. Nice bank of capacitors and inductors clean up the output of the switchers. To adjustable regulators (LM337 and LM317) provide lower output voltages (?).

Yep. The HP-1 runs on ±18 V. Mean Well makes a 15 V and a 24 V version of the IRM-series. No 18 V. So I regulate to ±18 V with the LM317/337 pair. That also provides additional filtering of any switching hash that makes it past the CLC filter on the output of the Mean Well supply bricks.

The red component is a surge suppressor. The black disc component is the inrush limiter.

The only difference is that the DC servo (feedback) loop is implemented using OP2197 rather than OPA2140 as noted above. Or else my eyesight has gone down the hill more than I think they have.

I used the OPA2197 as the OPA2140 was out of stock when I had the boards assembled. The two are interchangeable in the DC servo application. I can't measure any difference between the two, so I'll stick with the OPA2197 as it's quite a bit lower cost.

Everything is doubled here to get more power by paralleling the two amplifier stages (OPA1611 and LME49600).

Yep. Getting the OPA1611 to play nicely with the LME49600 is where the secret sauce is in the HP-1. The OPA1611 has a wobble in the phase around 10 MHz due to its internal compensation. That makes it very hard to use in a composite amplifier, such as the HP-1. I was quite satisfied that I got it to work.

Since Tom is here, I am going to stop pretending I have fully reverse engineered this unit from just a picture and let him explain more.

I think you've covered the highlights. The relays are by Kemet. They're optimized for low-level signal switching and have gold point contacts. There's about $40 worth of relays in the HP-1.

Quality of assembly and soldering is high.

Yep. I found a good assembly place. They're about 15 minutes from my house and specialize in low-volume orders. It's a solid outfit! For the soldering process, they use a reflow oven with 13 (thirteen!) temperature zones. This minimizes the stress on the parts and prevents the parts from cracking. The various Chinese outfits often use reflow ovens with three (3) zones! I've run a few hundred boards through them and they've yet to let me down.

Tom

 

[tomchr]

Interesting that the thermal pads below the various chips are sufficient for such a powerful amplifier and heatsinks are only required for the LM337 and LM317 pair.

The tab of the LME49600 is connected to VEE. There's a VEE plane within the board layer stack, so the ICs have nearly all the board area to dissipate the heat into. I did keep the planes away from the mains voltage to maintain the Class II isolation. Thus, the area around the mains inlet and circuitry is the only area not used to heat sink the LME49600s.

Tom

 

[Purité Audio]

Not that I know anything bout circuit design but that does look superb.
Keith

 

[tomchr]

Thank you.

Tom

 

[amirm]

Thank you.

Tom

You should huddle with Keith to see if there is a market/business model for him to sell these in Europe.

 

[Tatteredmidnight]

I'm not an EE or any sort of expert, but I tend to judge designs that are too aesthetic harshly, this device seems to straddle that line very deftly. I appreciate the organization and consistency of the layout, especially considering this is also sold as a DIY kit. Despite this consideration, this board doesn't look like it was designed to be framed on a wall, and instead appears to be optimized as its first priority, and tidy when it won't effect performance. I believe this this born out by the absolutely class leading performance of the unit!

Major cudos to @tomchr for actually delivering a boutique product that performs at a level above mass market units. You have accomplished something countless companies boldly claim to, but fail miserably at when measured.

 

[thunderchicken]

If I buy the kit and want to spec it with the OPA 1622 instead of the 1612, what else would I need to change? Would there be any measurable benefit to doing that or is that a silly goal?

 

[tomchr]

The OPA1622 is a high-power opamp. It wouldn't be my first choice for anything other than a headphone output stage (capable of ~100 mW into 32 Ω as I recall). It's in a LLP/QFN or similar package. Not SOIC. So you'd have to find an adapter that'd convert it to SOIC. Lots of work for no gain.

One question springs to mind: Why?

Tom

 

[thunderchicken]

Lurking around the reviews, there's a lot of hate for the 1612. The 1622 seems to be SOTA and thought they were in the same family and easily swapped. I'm a parts-placer, not an engineer though and can't make those calls on my own. Thanks for responding!

 

[tomchr]

You could easily install an LME49720 rather than the OPA1612. Do leave the OPA1611 in the circuit, though. The stability of the amp depends on those.

Both the OPA1612 and OPA1622 are SOTA. There will always be someone who's disappointed that a piece of gear didn't use their pet part. And vice versa, there'll always be some who confuse correlation and causation (this amp sounds bad, it has the [part of hate d'jour] in it, thus all amps featuring that part must sound bad).

Tom