High performance OPA1602 + TPA6120A2 Composite Headphone Amplifier

[xrk971]

Well, I was referring to something like this and no need to include it in the feedback loop.

TPA6120 Headphone Amplifier Board HiFi TPA6120A2 Double Channel Headphone Amp Tu | eBay

 

[BamboszeK]

I agree with Chris. My amplifier or very similar Topping L30 isn't exactly best option for modding. Feedback in composite amplifiers needs precise tuning, otherwise even small changes can make them unstable.
By replacing first opamp in composite amp with discrete one, you would basically need to redesign whole amplifier. Even then you can't be sure about stability. I highly doubt that discrete opamp manufacturers provide SPICE simulation models.
If you find some kind of distortion sound pleasing I would suggest another way of introducing them to signal path. Personally I prefer faithful playback gear and all manipulation done in DSP.
If I want something that "render" music instead of reproduce it, I build guitar effects and tube amplifiers

 

[Mark Johnson]

Thanks for mentioning the L30, ggetzoff. I ordered one from Amazon and look forward to plinking around with it.

 

[ziga90]

Hi Bamboszek,


Can you share your BOM? I will try to prepare shematic for PCB but in case of any issues on the way, do you have a spare PCB to sell?


Is there any major differences between Topping L30 and this amp? Headphone set is Sennheiser HD598.


I would use my HP lap top as audio source for this amp, is there anything that needs to be adjusted?


Thank you.




Kind regards, Ziga

 

[BamboszeK]

I still have a few spare boards. No exact BOM with part numbers as I ordered lots of parts for prototyping from different vendors.
PCB layout for amplifier with such high speed device as TPA6120A2 isn't trivial.
L30 is more or less similar. Much lower input impedance, slightly different feedback/compensation and added output relay.
12dB of gain should be plenty even for laptop level output.

 

[MASamplifer]

Good morning all,
I was not actually aware all this composite amp work using the TPA6120A2 was going on. In parallel, talking with Jan Didden, I had embarked a similar trek that now show up on page 42 of the March 2021 AudioXpress. In the early Comlinear days as apps manager, we were doing a lot of composite amplifiers as the only source for high voltage CFA (1985 to 1990 about). Yes, the ADSL (or more generally differential wireline drivers) are intrinsically very low distortion of necessity. That was my main area for many years - all of those standard DMT (discrete multi-tone) - think of them as little radios in 4kHz slices (for xDSL). The key test for linearity is called and MTPR test - multi tone power ratio - the lineal descendent of the old microwave NPR test (noise power ratio). Essentially, you fill up the test spectrum with carriers with a notch of no power and look at how much those carriers fold into the notch (a comprehensive intermodulation test). Intersil actually was the dominant line driver supplier for a lot of this and as apps manager we did publish one of the very few public discussions on these issues - this largely came out of a more interesting development I was doing as a 14bit 500MSPS digitizer board we were doing to highlight the somewhat incredible ISL55210 FDA. In any case, those 2012 article have apparently migrated all over the place, but here I found part 1,
Linearity testing for G.hn PLC line drivers, pt.1

and then part 2 somewhere else - pretty odd actually.

https://www.eetimes.com/linearity-testing-for-g-hn-plc-line-drivers-part-2/

The essential point is, yes, you are going in the right direction with dual CFA ADSL line drivers for low distortion, high output, audio.

 

[BamboszeK]

Thanks for pointing out. Looks like I need to get my hands on newest AudioXpress and see your results. What input opamp did you used?
I've actually considered some Renesas (Intersil) parts but if I recall correctly there was something missing for this particular application as composite headamp.

 

[chris719]

Good morning all,
I was not actually aware all this composite amp work using the TPA6120A2 was going on. In parallel, talking with Jan Didden, I had embarked a similar trek that now show up on page 42 of the March 2021 AudioXpress. In the early Comlinear days as apps manager, we were doing a lot of composite amplifiers as the only source for high voltage CFA (1985 to 1990 about). Yes, the ADSL (or more generally differential wireline drivers) are intrinsically very low distortion of necessity. That was my main area for many years - all of those standard DMT (discrete multi-tone) - think of them as little radios in 4kHz slices (for xDSL). The key test for linearity is called and MTPR test - multi tone power ratio - the lineal descendent of the old microwave NPR test (noise power ratio). Essentially, you fill up the test spectrum with carriers with a notch of no power and look at how much those carriers fold into the notch (a comprehensive intermodulation test). Intersil actually was the dominant line driver supplier for a lot of this and as apps manager we did publish one of the very few public discussions on these issues - this largely came out of a more interesting development I was doing as a 14bit 500MSPS digitizer board we were doing to highlight the somewhat incredible ISL55210 FDA. In any case, those 2012 article have apparently migrated all over the place, but here I found part 1,
Linearity testing for G.hn PLC line drivers, pt.1

and then part 2 somewhere else - pretty odd actually.

https://www.eetimes.com/linearity-testing-for-g-hn-plc-line-drivers-part-2/

The essential point is, yes, you are going in the right direction with dual CFA ADSL line drivers for low distortion, high output, audio.

Just wanted to say that I've come across several of your articles and posts on TI's forum and they are always top notch.

 

[JensH]

Hi Michael,

I have read your interesting article in the March 2021 edition of AudioXpress.
A couple of things surprised me a bit.

In your simulations you seem to use a purely resistive load (32 ohm). Wouldn't it be necessary to check the stability with a complex load, at least with a capacitor in parallel with the 32 ohm ?

You do mention the "relatively low input impedance". In my view the low input impedance puts some tough requirements on the preceding stage if you want to maintain the very low distortion.
Have you tried a non-inverting configuration, similar to the one Bamboszek has used ? Do you have an opinion on the performance of the one from Basboszek (and perhaps the Topping L30, using the OPA1612 and a slightly different compensation compared to Bamboszeks) compared to your inverting design?

 

[BamboszeK]

One of things I wanted to examine was inverting vs non-inverting configuration effect on distortion. That was reason why I decided to try OPA1602. You probably seen those posts, but anyway:
https://www.diyaudio.com/forums/vendor-s-bazaar/283672-audio-op-amp-opa1622-3.html#post4554515
https://www.diyaudio.com/forums/vendor-s-bazaar/283672-audio-op-amp-opa1622-9.html#post4636374
https://www.diyaudio.com/forums/vendor-s-bazaar/283672-audio-op-amp-opa1622-10.html#post4637092

I already reached performance cap of APx525 so didn't bothered with inverting topology as higher input impedance and preserved absolute polarity is rather useful. But inverting topology is indeed worth trying if we want to squeeze last dBs of THD.

Regards,
Miłosz

 

[jasonhanjk]

Feedback in composite amplifiers needs precise tuning, otherwise even small changes can make them unstable.

That's only partly true.
I can make 5532 or 2068 stable with 4556, after a lot of trial and error. Composite amp schematic online need to have the low noise op amp and amplifier bandwidth to be the same for stability. That's the main reason not to anyhow change ic.

L30 already put a lot of good engineering, so better leave it as is. They add something to make it stable.😅

My current design Tpa6120 use 0 ohm output resistor.

 

[kn0ppers]

BamboszeK is absolutely correct in my opinion. You throw around the term stability and claim to have a design with 0 Ohm output resistor, but how exactly did you verify stability and what kinds of phase margins are we talking into typical headphone loads?

 

[jasonhanjk]

Great question.

I won't tell you the answer directly, however I will ask you just 1 question.
What's the reason(s) for the TPA6120 needing resistance at it's output?

 

[BamboszeK]

It isn't TPA6120 specific. To maintain loop stability, output of the amplifier need to be isolated from capacitive load. Resistor is cheapest way of doing it.
Some opamps are stable up to certain load capatiance without additional isolation, up to couple hundreds of pF usually. That's why you often won't see any resistor at output of headphone amplifiers.
Inductor is the better way of doing it, but it's considerably larger and more expensive. Doesn't limit power and output impedance however. Standard practice in speaker amplifiers.

I really simplified things. More in-depth explanation could be found in most books about analog design or this article Op Amps Driving Capacitive Loads | Analog Devices

 

[chris719]

Great question.

I won't tell you the answer directly, however I will ask you just 1 question.
What's the reason(s) for the TPA6120 needing resistance at it's output?

Capacitive load isolation isn’t some trade secret.

 

[Mark Johnson]

Someplace here on diyAudio is a posting containing several photos of a nosebleed high end manufacturer (Jeff Rowland Design Group)'s high power power amp, which contains eight or twelve LM3886 chips inside. You can see on the PCB that each of the LM3886 chips has an isolation resistor in series between the chip and the loudspeaker output, but none of the LM3886s has a series inductor.

 

[jasonhanjk]

Capacitive load isolation isn’t some trade secret.

Yes.
A few more questions and get it answer will eventually lead to 0 ohm output resistor.

It's also known that inductor cannot isolate capacitance, a ferrite bead should work better. Or better, both in series.

 

[RussellKinder]

Good morning all,
I was not actually aware all this composite amp work using the TPA6120A2 was going on. In parallel, talking with Jan Didden, I had embarked a similar trek that now show up on page 42 of the March 2021 AudioXpress. In the early Comlinear days as apps manager, we were doing a lot of composite amplifiers as the only source for high voltage CFA (1985 to 1990 about). Yes, the ADSL (or more generally differential wireline drivers) are intrinsically very low distortion of necessity. That was my main area for many years - all of those standard DMT (discrete multi-tone) - think of them as little radios in 4kHz slices (for xDSL). The key test for linearity is called and MTPR test - multi tone power ratio - the lineal descendent of the old microwave NPR test (noise power ratio). Essentially, you fill up the test spectrum with carriers with a notch of no power and look at how much those carriers fold into the notch (a comprehensive intermodulation test).

Very true about the ADSL line drivers. I was hired into TI in 2002 to work on what became the TNETD7122, the analog front-end for the AC7 chipset. It drives and receives ADSL signals at the telephone company end of the phone line. I remarked one day that the TX amp would make an amazing headphone amp (-126 dB THD+N into 40 ohms at far beyond audio BW). That's when they told me the audio guys had repackaged the previous ADSL line driver as the TPA6120, and that it was, in fact, very good.

 

[drizdraz]

It's also known that inductor cannot isolate capacitance, a ferrite bead should work better. Or better, both in series.

A ferrite bead is inferior to an air-core inductor in this application. Do you know why? (hint: saturation and hysteresis).

 

[abraxalito]

What's the reason(s) for the TPA6120 needing resistance at it's output?

As far as I could see from the DS, its needed to reduce the gain (into a 32R HP) and thereby goose the noise spec.