"The Wire" Ultra-High Performance Headphone Amplifier - PCB's

[opc]

A new run of these boards is now available! Please see the following thread for details:

http://www.diyaudio.com/forums/soli...ble-here-bal-bal-se-se-lpuhp.html#post3516741

There is a Wiki here which contains all the project information, documentation and details for all "The Wire" projects:

The Wire - All Boards and Kits Explained Here!

Hi Guys,

I just finished building my latest headphone amplifier project, and this one is definitely worth sharing with the community. After a few prototypes, I went ahead and had some PCB's made, and there are plenty extra available if people are interested.

I basically wanted a headphone amplifier that was as close as could be to a wire with current gain. I didn't want it to impart anything at all to the source signal. That led me to the following design criteria:

- Perfectly flat FR from DC to over 100kHz
- No phase shift from 0-100kHz
- No capacitors (except for PSU)
- Extremely low harmonic distortion
- Extremely low IM distortion
- Extremely low noise floor
- Ability to drive down to 4 ohms with no load dependence
- Enough voltage swing to drive any reasonable headphone set.
- Excellent input CMRR
- Simple circuitry using the best parts available.

I needed a balanced input, and wanted to use a standard 1/4" jack for the output.

The resulting amplifier is basically an instrumentation amp using three LME49990 op-amps and an LME 49600 buffer nested into the last stage. These parts represent the absolute best you can get for this sort of application. I used all 0.1% Susumu thin film resistors, and all X7R ceramic bypass capacitors placed directly on the supply pins. Bulk caps are all solid polymer for the absolute lowest ESR and best HF performance. Layout was optimized for short signal length, low noise and low crosstalk.

The circuit provides differential input, or the option to ground one phase and drive the circuit with an SE input. Gain is set to 1, but can easily be changed to pretty much anything with just two resistors It runs on +/-5VDC up to +/- 15VDC and has enough drive to run anything you can throw at it.

As for sound, having a DC coupled amplifier makes for some of the most stunning bass I have ever heard. I'm driving a pair of Denon AH-D2000 headphones, and there's a world of difference between this amplifier and every other source I have ever tried when it comes to performance below 100Hz. It has incredible impact, depth and cleanliness. In the midrange and top end, it passes on the characteristics of the source like nothing I have ever heard. I've driven it with an Aikido linestage, a BZLS, straight from a DAC and a multitude of other devices, and all you hear is the device driving it. I now use this setup as my primary means of evaluating all preamps and sources since nothing else I have ever used even comes close to exposing the true nature of the sources like this circuit does. It's honestly like it's not even there... hence the name.

I've attached some pictures of the finished product below, and I'll follow up with some measurements done on an Audio Precision. Schematic to follow also.

Cheers,
Owen

 

[opc]

I've attached a multitude of measurements of the circuit that were made using a brand new Audio Precision APx585 and a headphone load test jig.

I'll let the measurements speak for themselves. It's pretty easy to see that all the design goals listed above were easily met, and this circuit has absolutely incredible performance.

Vanishingly low THD+N and IMD levels, incredibly flat FR, no phase shift in the audible band, and an astonishing 145dB SNR.

The 1kHz (1V into 32ohm) FFT tells a large part of the story. The noise floor is down at -145dB, 2nd harmonic is more than -120dB down with the third more than -125dB. 4th and 5th are below -130dB and 7th just peaks over the -130dB line. There is no 60/120Hz PSU noise whatsoever, and the HF range is dead quiet. I'd challenge anyone to post a measurement of an amplifier that can do better than this.

Note dB the scale on the frequency response from 5Hz to 50kHz. You're looking at an amplifier that is +/- 0.001dB from DC to 20K. +/- 0.004dB up to 50kHz. This is at 1V driving a 32 ohm load.

Same notes on the phase scale over the same range. Phase shift is 0.11 degrees at 20kHz, and less than 0.4 degrees at 50kHz. As above, this is 1V into 32 ohms.

Moving on to IMD, performance is still phenomenal and there's nothing to hide here. There's a plot for SMPTE, MOD, DFD and CCIF. Again, this is 1V into 32 ohms.

 

[opc]

Next up we have drive level dependency. This amplifier exhibits superb performance whether it's driving 1V out, 0.5V out or 0.25V out into 4 ohms, 9 ohms, 16 ohms, 32 ohms, 900 ohms or 10k ohms. That means performance doesn't crumble at lower or higher output level. It also doesn't change with load, which can vary widely with some headphones.

There are three FFT's attached each at different voltage drive levels into 32 ohms. Performance is actually marginally better at lower drive levels, as the 2nd harmonic drops to just over -130dB and the 3rd harmonic to just over -135dB. The noise floor is sitting down at -155dB. I've never seen anything measure this quiet.

The last 6 graphs show drive capability into various loads. In order they are 4 ohms, 8 ohms, 16 ohms, 32 ohms, 900 ohms, and 10k ohms. The LME49600 current limits into very low impedance, and voltage clips into higher impedance loads. These graphs show you what you can expect to get for drive levels into various loads. All measurements were done with +/- 15VDC rails.

 

[abraxalito]

Layout was optimized for short signal length, low noise and low crosstalk.

How do you connect the pin1s of the XLR inputs? Straight to the groundplane ?

 

[opc]

Finally, I've attached the schematic with the top layer layout.

Any comments or questions are welcome, and of anyone is interested in a board, let me know. I made 25 pairs of amp and PSU boards, some of which I'll be keeping for various projects.

Cheers,
Owen

 

[bearberry]

Very nice!

I see you are keeping busy, Owen.
Are you going to bring this to the January meet?

John

 

[opc]

Hi John,

Busy indeed! It will definitely be making an appearance at the next meet, probably driven by the DAC I've been working on. I'll set it up along with the Denon headphones and people can give it a good listen.

I more or less designed it specifically for that purpose, and it works very well in that setup.

I started the whole undertaking 6 months ago when I realized I do 90% of my music listening on my headphones at work. I looked at my $20 Sennheiser headphones plugged directly into my computer's soundcard and though "I should probably put some effort into making this better".

Hence the DAC and this amplifier.

Will you be at the next meet?

abraxalito:

They are indeed tied directly to the plane. Are you concerned about the 0.1mA of current pulling a voltage differential across that 0.001 ohm plane?

Cheers,
Owen

 

[Nikon1975]

If I see that I manage to solder the smd components for the DAC, I would be very interested in getting a board.

Thanks,

Davide

 

[Frex]

Hello,

Opc, very clean and beautiful design.
For HF immunity, maybe it could be better to slightly limit the bandwidth
with shunt capacitor (pF range) and a serial chock inductor.
Without that, your amplifier have a so high bandwidth that it
can receive radio signals and intermodulated with.
Do you plan to organize a PCB group-buy ?
Regards.

Frex.

 

[abraxalito]

They are indeed tied directly to the plane. Are you concerned about the 0.1mA of current pulling a voltage differential across that 0.001 ohm plane?

In my experience, ground currents are often more than 0.1mA. So yeah, given you seem to be aiming for the sky (or should I say the earth's core?) in terms of measurement numbers, it does seem to me to be an issue. Does your groundplane really have such a low impedance as a milliohm? How does it fare at higher frequencies?

I'm curious to see how such a well measuring piece of kit fares when subjected to common mode currents at various frequenices (not merely audio ones) through its pin1s. Do you have equipment that can measure that?

<edit> I agree with Frex about band limiting. I think series inductors or a common mode choke are a good idea. I think it may've been Ed Simon who recommended an inductor on the input pins of National opamps like this one - they are rather RF sensitive. Also a CM choke prior to the mains transformer.

 

[opc]

Hi Guys,

Nikon1975:

There's a board here for you if you want it! As you can see from the picture, I've got a good pile of them.

I'll figure out what I want to do for pricing, and I'll post it up here tomorrow.

Frex:

Very valid concern, and one that is easily addressed if need be. There are two options I can see, the first of which is band limiting the LME49600 by lifting pin 1 (not really band limiting since it only swaps between 180 and 110MHz) or soldering small caps in parallel with the two input R's.

Any band limiting does come with phase issues though, which is why I chose not to do anything. The circuit is perfectly stable as is, and unless you're really sloppy with input and output wiring, then there isn't really any issues with RF interference.

I did find that I was able to pick up radio stations with long input cables and a long headphone jack, and with the circuit strewn across the carpet and not in a chassis, but I don't think it would be an issue with any normal setup in a proper chassis.

Just the same, I could calculate the value of two input caps that would help roll off at say 100kHz, at the cost of some phase shift at 20kHz.

abraxalito:

It does depend on what you're driving it with, and if you're running SE then things definitely change. Not sure what the exact impedance is, but it is technically a large flat piece of copper foil, so I'd imagine it's pretty decent when it come to HF.

If you can draw up what you're thinking in terms of input pin current, then I can certainly look into testing it. In my experience though, the serious culprit in crosstalk is almost always the output jack, and even more often the headphone cable itself. I pulled apart a pair AKG 240 MkII headphones the other day and saw that each channel uses a common ground wire that is tied at the headset itself! That means you're got at least 0.5 to 1ohm of ground resistance which is going to results in terrible crosstalk. The jack itself has a decent amount as well.

If I measure crosstalk at the jack with a 32 ohm load, I get between -98 and -102dB which is pretty decent. Higher impedance loads give significantly better results.

Cheers,
Owen

 

[dggs]

Hi Owen,

I'm interested in getting 1 or 2 sets of the PCB.

Thanks,

 

[abraxalito]

If you can draw up what you're thinking in terms of input pin current, then I can certainly look into testing it.

How about 100uA RMS with frequencies ranging from 50Hz to 100MHz or is that too tall an order? The audio band output noise floor to be monitored while playing some high level white noise in the band 22-100kHz. That ought to be enough of a torture test to see if noise modulation is occuring

In my experience though, the serious culprit in crosstalk...

I'm not so interested in crosstalk as I don't think it has a serious impact on sound quality - LPs have poor crosstalk by design for example. More relevant IME is intermodulation.

 

[qusp]

Looks great Owen, i'm looking forward to the PCB, I might even play around with a portable with one, looks small enough. if you havent sent mine, i'll grab another one from you if thats cool.

 

[DualTriode]

Finally, I've attached the schematic with the top layer layout.

Any comments or questions are welcome, and of anyone is interested in a board, let me know. I made 25 pairs of amp and PSU boards, some of which I'll be keeping for various projects.

Cheers,
Owen

Hello,
If you have a set of PCB's to send south count me in for a pair. I would like to try my hand at surface mount parts.
DT
All just for fun!

 

[qusp]

should we GB the buffers and opamps?

 

[opc]

Hi Guys:

A group buy for the LME parts would indeed be wise. Digikey has the worst pricing for these parts, and they can be had for half as much at other sites.

Either way, I've got 25 boards and that's 150 LME49990 and 50 LME49600.

For 100pcs of LME49990 you've got:

Avnet = 2.57 each
Newark = 3.00 each
Digikey = 4.15 each
Arrow = 2.95 each
Future = 3.12 each

Digikey's single part price is $5.70.

For 50pcs of LME49600 you've got:

Newark = 6.98 each
Digikey = 9.47 each
Arrow = 6.53 each
Future = 6.90 each

Digikey's single part price is $11.56.

I would consider doing a group buy kit for the boards and parts, but I'm not prepared to do it at cost. I just got through that process with 40 DAC boards, and it took easily 50 hours of my time and I ended up more than $200 in the hole.

If there's enough interest though, I'll look into it. The price would end up being less than buying all the parts on your own, but it wouldn't be at cost.

Cheers,
Owen

 

[qusp]

yeah see I dont need anything much except the chips, not using the PSU and wont use the input or output connectors, or probably even the same reservoir caps, although those LF series are pretty good value as polymer caps go. i'm sorted for the rest of the parts as far as I can see, except for perhaps specific feedback R's

yeah DKs extortionate prices are why I asked, the buffer is insanely priced

did you get my email about the boards mate?

so yeah I would participate in a GB for the chips, but thats about it I think, unless prehaps you did an SMD only thing.

I dont even want to talk about the time yet to come for me for the GB haha, at least youve got something on the way to offset the 200, hopefully it puts a smile on your face

 

[johnm]

Hi John.

I'd like a board if there's any still available please?

Kind regards,

- John

 

[qusp]

john?? there is no john here but you