Review and Measurements of Benchmark HPA4 Headphone Amp/Pre

[amirm]

Also I find it interesting that the harmonic distortion in the loopback test seems to have decreased when using an external cable, or is it because of the slightly higher level (5.7 vs 5 volts) that the ratio of distortion is lower compared to the internal loopback test?

Yes, it is the output voltage being higher. Not because of the cable.

 

[AndrewDavis]

That Hole-in-One Pink Panther doesn't come out too often. I don't have to read the review to know this is a winner!

 

[ayane]

Thank you Amir, for once again delivering an outstanding slew of measurements! I've been waiting to see these for a while now, and as expected, my jaw dropped. It's frankly amazing to see the APx555's limits bottlenecking the measurability of the HPA4's performance.

I'm curious to see the distortion vs power graphs updated with higher precision resistors (edit: for the low impedance loads), and maybe also 16 ohms. With headphones like the MrSpeakers Aeon and Ether 2, both of which have an impedance of under 16 ohms, the amp used to drive them shouldn't be current limiting. I have faith in Benchmark and THX for accommodating this into their design, so I'm not too worried about the HPA4, but almost every amp out there that doesn't have high power output current limits at these low impedances. There are actually some graphs towards the end of the HPA4's manual showing this particular test result, although their testing methodology and analyzer are different.

Seeing these measurements also makes me wonder if Benchmark is working on a "DAC4" to keep up with HPA4's performance, especially considering that there are newer Sabre chips than the one used in the DAC3. (And some newer DACs like the Okto DAC8 and Matrix Element X, which use the newer Sabre chips, have surpassed Benchmark's DAC3 in performance!).

 

[amirm]

I'm curious to see the distortion vs power graphs updated with higher precision resistors, and maybe also at 16 ohms.

Thanks for the kind words. Yes, I plan to add 16 ohm to my revised dummy load I am building.

On more precise load, I should say that it only impacts 33 ohm. 300 Ohm ones use the ones inside Audio Precision so were very good to start.

But yes, point is well taken.

 

[tomchr]

Know example is another design but for Neurochrome HP-1 by @tomchr, he told below about difference for 1/4 jack verse 5 pin XLR, note his HP-1 is also not a ballanced output but single ended:

QUOTE That little bit of shared ground impedance in the 1/4" phone jack actually degrades the channel separation from 115 dB to 90ish dB. That's physics for you. There's no magic voodoo here. It's all Ohm's Law UNQUOTE

4-pin XLR. But yes. To further quote/paraphrase myself: I see no advantage of using balanced outputs. I think the fascination with "balanced" stems from the days of vacuum tubes where balanced circuitry was often used to get lower distortion. The issue with that is that the balanced circuitry will cancel the even order harmonics (those that most find to sound pleasant) while leaving the odd order harmonics (that many find to sound harsh) in the signal. But the overall THD is lower. This is what happen when you get tunnel vision and lock in on a single measurement as the holy grail.

There are cases where "balanced" matters. Balanced interconnects, for example. Or rather: DIFFERENTIAL inputs. Inputs that respond to the voltage difference between two signal conductors while being utterly indifferent to the ground potential. Such circuits, when used correctly, reject any signal that's imposed equally on the two signal conductors (mains hum for example). It wouldn't surprise me if many single-ended/unbalanced pieces of audio equipment operate with a level of mains hum that's just below audible. This creates various intermodulation products and results in a level of 'fuzziness' perceived in the audio. Switch to a differential setup and the 'fuzziness' goes away. I suppose I could write some marketing babble about "blacker blacks" or whatnot, but I'm not much of a wordsmith in that regard.

Some have extended this to, "well, if balanced/differential connections are good, then balanced/differential circuit must be even better, right?!" Well. No. Not necessarily. Differential circuits are used mainly for cancellation of some unwanted signal (such as mains hum). In the case of a differential receiver, the "cancellation factor" (common-mode rejection ratio) depends on the matching of four resistors. Given that ±0.1% tolerance resistors are commonly available ($1/each) and ±0.01% tolerance resistors are no longer carved from a solid block of unobtanium by certified organic virgins ($10/each - for the resistors, not the virgins!) getting good CMRR is quite achievable even at the PCB level.
However, if you are trying to cancel distortion in a circuit by using a balanced design, you're in for a rough go - especially in a discrete circuit where the matching between active components tends to be rather horrid. Thus, if you rely on transistor matching in a discrete circuit to cancel any unwanted signal, you are much more likely to degrade performance than to improve it in a balanced design.
In IC design, the situation is different. The matching between components on an IC (at least assuming the process has been optimized for analog circuits), the matching between components is much, much tighter. The absolute tolerances on components is worse, but the matching is much better than at the PCB level. Thus, IC design tends to derive its performance from the matching between components, which tends to result in differential/balanced designs.

Some fret about balanced connections from the input to the output of a box. I'm not one of them. As I showed in my HP-1, it is perfectly possible to get stellar performance with single-ended signalling within the enclosure. You just have to pay attention to the PCB layout.
Now, in my upcoming HPA-1 (under the Tom Christiansen Audio brand), I use differential signalling all the way to the volume pot. Why? Cuz I can... It doesn't cost me any extra to put the differential receiver by the volume pot and it gives me a marketing blurb, "balanced all the way to the volume control!" Will it make any difference in the performance? It might, but most likely it won't. And most importantly: It will not degrade performance.

NB: Observant readers will note that I'm using the terms differential, balanced, and balanced/differential interchangeably above. I'm well aware that not all balanced circuits are differential and that not all differential circuits are balanced. I assume that the reader is savvy enough to infer that I'm referring to well-engineered implementations of differential circuits (which are also balanced) and not some hack.

BTW: Those interested in learning more about component matching (and layout matching) should take a look at Marcel Pelgrom's work.

And in the manual
"The right-hand jack is a 4-pin XLR headphone jack that features dedicated returns for each channel.
The main purpose of the second connector is to provide a high-performance alternative to the traditional TRS phone jack."

It's impossible to have perfectly same balanced out and single end out if the using the same circuit. The noise has to be doubled, output impedance has to double, output voltage swing(power for high impedance)will be doubled(quadrupled).

Correct. The only advantage of the 4-pin XLR is that you skip the few mΩ of shared ground in the 1/4" phono connector. You better hope that the layout designer of the headphone amp use a ground plane as using a ground trace would degrade performance even more.

Also correct that you cannot use the same output for balanced and single-ended. That would short the two (-) outputs of the balanced driver together and now you're back to single-ended.
I would further argue that using a second (inverting) driver to form a balanced (but not differential) output just so you can claim to have "balanced out" on the XLR (and get to claim higher output power) is 99% marketing, 1% stupidity, and 0% engineering. Twice the noise (= 4x the noise power!), twice the output impedance, likely higher distortion (due to poor matching between the two amp halves), twice the cost. Yeah. Not my cup of tea. But you get more output power...

Tom

 

[tomchr]

On more precise load, I should say that it only impacts 33 ohm. 300 Ohm ones use the ones inside Audio Precision so were very good to start.

Why 33 Ω? Yes, I know it's an E12 standard value, but why not 20 Ω and 12 Ω in series? Both are standard values (E24). You could then use two 32 Ω in parallel for 16 Ω.

You could also use two 16 Ω (E24 standard value) in series, but that misses the 20 Ω tap.

Thinking some more: 16+2+2+12+18+10 in series (all E24) would give you 16, 20, 32, 50, and 60 Ω taps. 16+2+2+12+18+100 would bring you to 16, 20, 32, 50, and 150 Ω.

I understand the difference between 33 Ω and 32 Ω is academic, but when the standard impedance is 32 Ω, I think it would make sense to set up your test rig to support that.

BTW: Make sure to use resistors that are grossly oversized on the power rating and have a low TCR. Otherwise, you'll likely get distortion from the load itself.

Tom

 

[akarise]

Despite the amazing measurements and technical performance of the HPA4, were you able to actually hear a noticeable difference between it and the THX AAA 789? Or is all this improvement in measurements beyond the human ability to hear it?

 

[Jimmy]

A really nice statement product, Benchmark rarely dissapoints, but I doubt that there will be any meaningful audible differences with the THX789 or even the Atom.

Anyway if you want to have a no compromise solution in terms of build, connectivity and aesthetics, this is the ticket, and if you compare the price with the one of many (IMHO overpriced) flagship headphones this can even seem like a bargain.

By the way, what are the two 20 pin TI? chips placed close to the HP outs and next to 2 x 10uf electrolytics?, the images on the net are too low res to tell.

 

[amirm]

Despite the amazing measurements and technical performance of the HPA4, were you able to actually hear a noticeable difference between it and the THX AAA 789? Or is all this improvement in measurements beyond the human ability to hear it?

I didn't try to do an AB test between them. But I don't expect to be able to tell the difference. So shop on basis of features, availability, usability, etc.

 

[Ralf Stocker]

I have one that I did that is close enough:

View attachment 29765

A little peak at 60Hz. hehe! Does AP using linear or switches psu in their AP55?

 

[amirm]

A little peak at 60Hz. hehe! Does AP using linear or switches psu in their AP55?

I don't know. The previous series used a linear supply. But yes, that 60 Hz peak was the first thing I noticed when I evaluated it.

 

[Noob]

I didn't try to do an AB test between them. But I don't expect to be able to tell the difference. So shop on basis of features, availability, usability, etc.

Speaking of features, the preamp capabilities of this unit really set it apart from the Drop 789.

 

[suttondesign]

i know it’s kind of ridiculous, but to me, the benchmark equipment has the feel of precision lab instruments. it flatters me into thinking i am being a scientist or engineer when just fooling around listening to music. sometimes, when i am feeling especially naughty, i wear a white lab coat when using it, and i call my wife into the room to be my assistant. good times!

 

[John_Siau]

Andrew Mason of THX designed the power supply (SMPS) section too...

Andrew Mason designed the SMPS in the AHB2 power amplifier but not in the HPA4. The DC to DC converters on the main board are a Benchmark design. These are the same DC to DC converters that are used in the DAC2 and DAC3 products. These take a 12V DC input and provide +/- 18V, +5V, +3.3V, +1.8V and +1.2V.

 

[RichB]

I purchased an LA4 which is the HPA without a headphone amp.
The XLR connections are:

Oppo UDP-205 -> LA4 -> AHB2's
Oppo UPD-205 -> XMC-1 -> LA4 -> AHB2's

I have compared the XMC-1 Reference Stereo (All analog) to the HPA by splitting the Oppo UDP-205 XLR out to the LA4 and XMC-1 and level matched. With some source material listening at 2.83 volts, it is difficult to differentiate. At even lower levels, the LA4 seemed a bit clearer but it could also be that the LA4 and XMC-1 levels do not track perfectly.

The LA4 allows driving the AHB2's at the lowest sensitivity for best performance.
I am very pleased with the performance and it allows an all analog direct path that is getting harder to find in HT processors.

It is a great product but there are some operational quicks.
In the above configuration, one LA4 input is used for HT (XMC-1) and the other for 2-channel directly from the Oppo Balanced outs.

The LA4 is sold with an optional $100 remote. The remote has discrete codes to select the inputs so I assumed they were selectable.
Unfortunately, they are not, the discrete codes only work when paired with a Benchmark DAC.
Benchmark tech support was able to setup the LA4 to predictably select the first input which then allows a next input to select the other balanced input. I don't think I could recreate this configuration without a call back.
Even so, this is not ideal. If both codes are not recognized the front channels are not selected.
Direct input selection is more reliable.

I have submitted a enhancement request to request a mode where the LA4/HPA4 could be set to use the discrete input codes. I expect other customers will be surprised to find the direct inputs do not work when not used with a Benchmark DAC.

The front panel display is attractive but there is a comparatively large LA4 display and small volume that cannot be easily discerned from across the room. It would be nice to have an option to use the HPA4 area for a large volume display.

Other niceties that could be added are preamp setting for power-on volume and maximum volume. The latter is useful for systems that cannot handle the LA4's maximum voltage output.

- Rich

 

[Topher5]

As an owner of both the HPA4 and THX AAA 789 for my balanced headphones, I'd say the most significant difference is the volume control precision you get with the HPA4.
256 steps with almost no clicks within the headphones (while adjusting volume) is a major technical achievement, in my opinion. I frequently change the volume, and to have basically noiseless stepped attenuation is incredible.
There are clicks from the HPA4 due to its stepped nature, but it sounds cool. If Benchmark wanted, I'm sure they can get clever with some acoustic padding to reduce the clicks within the box but that is overkill.

I like to be able to listen at low volumes and have perfect left/right channel balance. No other headphone Amp I've ever owned could do it like this. After getting the HPA4, I've sold a lot of amps! I'm having a hard time letting go of my THX-AAA 789 however, as I do enjoy it a esthetically, but after the HPA4 two months ago, I never use it at all.

 

[Topher5]

It was completely worth jumping to endgame on the headphone Amp. The price is high, but the peace of mind is amazing. Dollar for dollar however, the THX AAA 789 is still the best bang for the buck.
Even though I enjoyed the THX AAA 789 so much. As soon as I learned that the HPA4 used the same technology, I wanted to give it a shot. I was especially convinced after reading the following on the Benchmark site:

https://benchmarkmedia.com/blogs/application_notes/benchmarks-256-step-relay-controlled-attenuator

If you are really serious about headphone listening, give the HPA4 a shot. If you don't like it, return it! I fully expected to return mine when I first ordered it. But after using it for a few days, I knew I'd have a lot of selling to do.

 

[Music1969]

Dollar for dollar however, the THX AAA 789 is still the best bang for the buck.

JDS Labs Atom is probably up there too, pound for pound.

 

[Topher5]

JDS Labs Atom is probably up there too, pound for pound.

I meant for balanced audio* but yeah the Atom is probably the ultimate bang for the buck. I have not heard it, as I wanted only balanced headphone audio. I trust Amir's thorough review and opinions on it though.

 

[ShiZo]

I have a hpa4 and my friends jds atom sounds pretty damn good.