Neurochrome HP-22 - The $90 high-performance DIY headphone amp

[tomchr]

It turns out that high-end DIY actually can be affordable. Who knew?!

Build your Neurochrome HP-22 headphone amp for less than $90!

I have just opened up for preorders of the HP-22 circuit board and will offer a preorder discount until the boards arrive in stock. I expect to start shipping them by December 22, 2020. If you use FedEx or UPS for the shipping method, you'll be ready to build just after Christmas. You can preorder your HP-22 circuit board here: HP-22 – Neurochrome

I am intending for this thread to become the ASR build thread for this amp, so should you choose to build the amp, please post your build pictures here.

The HP-22 mounts as a shield atop a Texas Instruments OPA1622 evaluation module (EVM) to form a complete high-performance headphone amp. The amp provides 200 mW into 300 Ω, 360 mW into 50 Ω, and 200 mW into 32 Ω at -130 dB (0.00003% THD). The multi-tone IMD residual measures below -146 dBr (referenced to 100 mW into 300 Ω).

The OPA1622 EVM comes with rubber feet, so the HP-22 will sit neatly on a tabletop even without a chassis. As it contains the input RCA connectors, the volume control, and the output connector the HP-22 can be used without a chassis if desired. It can even be powered by a pair of 9 V batteries. And, naturally, if you wish to mount the HP-22 in a chassis, you can easily mount it to the front panel using the volume pot mounting hardware and connect the RCA connectors with wires.

The images below show my prototype build. The hockey puck was included in the picture as a scale indicator. The HP-22 is rather tiny. Maybe if I was a better marketeer, I would claim that the proprietary Neurochrome puck is made from a unique and patented rubber compound, which provides vibration control and resonance dampening.

You can find all the performance graphs and specs on the HP-22 product page: HP-22 – Neurochrome.
All measurements were performed using an Audio Precision APx525 and with the HP-22 powered by a Neurochrome Preamp Power Supply.

Tom

 

[abdo123]

This should be in the desperate dealers forum.

 

[tomchr]

My bad. Perhaps the moderators can be persuaded to scoot it over if a seller-started build thread is inappropriate here.

Tom

 

[restorer-john]

Is the HP socket on the EVM a 1/8" or 1/4"? I can't tell from the pictures.

 

[tomchr]

It's 3.5 mm.

Tom

 

[abdo123]

My bad. Perhaps the moderators can be persuaded to scoot it over if a seller-started build thread is inappropriate here.

Tom

No problem, I sent you an email containing some questions about the modulus 86, your products are very intriguing!

 

[tomchr]

I've responded. Thanks.

Tom

 

[solderdude]

The new O2 (but without the clipping input issue)

Finally someone who posts clipping behavior.
How's clipping behavior when the current limiter engages (at 32 Ohm) ?

With a 120dB/V sensitivity head/ear-phone the 120mV switch-on-pulse will result in a 100dB 'pop'.
Something to take under consideration when using this amp with high efficiency headphones.
Even a very common 110dB/V will still pop with 90dB.
It won't damage your headphones though (about 0.15mW in low impedance headphones for about half a second) but the ears may not like it.
What's the reason the turn-on delay pulse in the data sheet is much smaller in amplitude and duration ?
Maybe the 'slow' turn-on can be made faster with a delay circuit that switches the enable signal faster can be a solution ?

When using it with 2x 9V batteries how is the behavior when 1 battery is almost dead and the other one is not ?
This was a thing for NwAvGuy.
The fact that it can run on +/- 2.5V (using prescribed opamps) will surely help with this though.

Nice choice of parts.
Can I ask you why 2 PCBs and why the power supply board jumping isn't done with jumpers but audio part is (serviceability perhaps) ?
Surely most DIY'ers won't mind a slightly bigger PCB ... do you have a standard enclosure in mind ?

 

[tomchr]

Great questions actually. Thank you.

The OPA1622 is able to provide more current with a positive swing than with a negative, so not surprisingly it clips sooner on the negative rail than the positive when forced to current limit. The clipping behaviour is clean, though.

The behaviour with asymmetrical power supply is clean as well. Once you get down to +1.5/-12 V the amp will turn off as its enable signal goes low. I ran it at +2/-12V and +12/-2 V in the plots below. A 9 V battery will provide around 5.5-6 V when fully discharged, so there's plenty of margin there.

The reason the HP-22 provides a larger pop than what you see in the data sheet is because the data sheet figures show the behaviour when the enable line toggles while the power supply remains stable. My graph shows the behaviour when the power supply is turned on and off. I did fiddle quite a bit with the delay on the enable line and what I show was the best performance I could get without getting into elaborate circuit design.
Start-up and shut-down are surprisingly hard things to get right. While at TI I spent quite a while designing a power-on reset circuit for a digital circuit. It's one of those "can't you just use an RC?" types of circuits that ends up being a few weeks spent simulating the circuit over various corner cases (process, voltage, temperature) to ensure that it produces a reliable reset, even on a power glitch or brown-out. Oh, and please fit it in zero area, because it's only active once...

Why two PCBs? Because the OPA1622 is impossible to solder by hand. OK. Maybe not impossible, but darn close. The device measures 3x3 mm and has 14 pads connecting to it underneath. If you have a solder paste stencil, solder paste, and some proficiency using those, you can solder it yourself. I have done the occasional toaster oven reflow (or hot air reflow) for prototypes. But it is not something I would want to support via email. So I leave the hard part to TI.
I did consider having the builder remove the 3-pin terminal block from the TI eval board and connecting the two boards by wires, but decided against that. First off, the lead-free solder TI uses is pretty hard to work with. The EVM is a through-plated board (four layers, I think, possibly two) and if you don't have the ability to heat up all three terminals at once, it's pretty hard to get that connector off. And good luck supporting that by email when someone pulls the through-plating out with the connector. Secondly, the footprint required for the wires would interfere with the RCA connectors. I really like the compact nature of this amp and didn't want the 'shield' to become too big. It's really no big deal to connect two wires to one of the terminal blocks. I use 20 AWG wire. It fits just fine.

I don't have a standard enclosure in mind. I'll leave that as an exercise for the reader. I'm sure Hammond has something that'll work. I'll probably search a bit and see if I can find something that'll fit. I will provide a drill template so you can at least get the holes in the right locations.

Tom

 

[tomchr]

Following discussion elsewhere, it looks like the Hammond 1455R1601 would be a good fit as a chassis for the HP-22 powered by a Neurochrome Preamp Power Supply.

Tom

 

[solderdude]

Great Tom, this thing looks like it had quite some thought going into it.
So the bottom board is a TI evaluation board... that explains the dimensions and different color/looks.
Layout will be optimal for the 1622 as well.
Also explains why the RC circuit is on the top board


Asymmetrical clipping will certainly be audible .
It will create a 'DC component' on the output though with higher imp headphones.

Thanks for the answers.

Someone may be asking you if you can make it balanced and drive high impedance headphones to high levels.

 

[Χ Ξ Σ]

Just asking, what is the reason for testing the multitone at -20 dBrA?

 

[solderdude]

Because the peak levels of all the frequencies, when they happen to be in phase, reaches near 0dB.
The more tones are used the lower the level of individual tones will be.
One could design a MT with the frequencies going up having a 'pink noise' amplitude character which would be more realistically representing music but isn't as easy to look at and won't show issues at higher frequencies that well and that's where the problems usually are.

 

[renaudrenaud]

Excuse me for maybe a silly question but do we have to buy the Ti EVM from our side?

 

[somebodyelse]

Excuse me for maybe a silly question but do we have to buy the Ti EVM from our side?

It looks that way - from the bottom of the Features tab:

Build budget: The total parts cost for the HP-22 PCB is just shy of $50, including the cost of the TI OPA1622 evaluation module.

 

[tomchr]

Asymmetrical clipping will certainly be audible .

Yeah. If you're using battery power I'm expecting you to turn the amp off if it starts to sound bad and plug the batteries into their charger. At least the amp itself behaves responsibly.

Someone may be asking you if you can make it balanced and drive high impedance headphones to high levels.

You could use two kits and drive one with an inverted signal. You'd have to make an inverter (easy) but it's certainly possible.

Just asking, what is the reason for testing the multitone at -20 dBrA?

This is my test procedure for multi-tone:

1. Set sine generator amplitude to produce 100 mW into 300 Ω
2. Set dBrA to the resulting output voltage
3. Load multi-tone signal
4. Run multi-tone test

This results in a test signal with the same peak voltage as a sine wave that produces 100 mW into 300 Ω.

The reason each tone ends up at -21 dBrA is that there are 32 tones.

I have noticed that @amirm places the 32 tones at 0 dBrA. I used to do that as well. But I don't think that is an accurate representation of what's being tested as the highest signal amplitude is 21 dB taller than the tones. I can take that up with him offline. I've been running these multi-tone tests on my circuits since about 2015. There isn't a standard way of running these tests, but it would be nice if we could all sort of agree on how to run them. If Amir has a good reason for putting the tones at the reference level, I'm certainly willing to listen.

Excuse me for maybe a silly question but do we have to buy the Ti EVM from our side?

Yep. You can buy it from Mouser/Digikey or you can buy it directly from TI. TI charges a flat $6.99 shipping fee.

Tom

 

[jtwrace]

I sure hope ASR is getting one so it can be benchmarked to others.

Just love what you're doing here Tom...hope you have more on the horizon!

 

[technoian]

Great concept Tom. Looks to be a fun project that most people could complete with only simple tools.
I would love to see you produce a balanced input version using the same TI eval board.

 

[tomchr]

Greetings from the Winter Wonderland.

The HP-22 PCBs are with UPS. They're in town, actually, though probably stuck at the airport for the moment. UPS will deliver a tad late for some bizarre reason... We had about 30-40 cm of snow dumped on us since sunset yesterday.

Tom

 

[Veri]

So much snow.. that looks fun not gonna lie.