High performance OPA1602 + TPA6120A2 Composite Headphone Amplifier

A question for BamboszeK, though anyone who knows may answer as well:

I recently got an Oppo BDP-105 disc player, which has a TPA6120A2-based headphone amp. It pretty closely follows the datasheet recommendation, so it has a 10ohm resistor at the output. Obviously, it and all the supporting parts are mounted to the PCB, so redoing the whole thing to make it into the composite amp originally proposed won't work. But can I still use a small air core inductor instead of a resistor on the output, without changing anything else? Is 300nH the magic number to use, or should I go with a higher or lower value depending on the headphones I will use with it?

Thanks all for your help!
 
Hello bonacciomj,
Inductor value depends on capacitance on output, mostly headphone cable capacitance in practice. 300nH should be safe for most of real-world loads. At least in my circuit as simulations suggested. In yours Oppo it might be slightly different due different compensation.
Keep in mind that air core inductor and small physical size is quite contradictory ;)

I would suggest winding your own inductors and then test stability of amplifier using scope. 300 Ohm dummy load with something between 220 pF - 10 nF in parallel should do the trick.

Regards,
Miłosz
 
Hi Miłosz, Thanks for your help! Sounds like it should be OK. I will have to double-check the datasheet and see what resistor values Oppo has used in the feedback loop, but I think it should be OK to use a commercially available inductor. I have found a company called Coilcraft that sells small air-core RF inductors with the correct value. Oppo has, for some reason, used two through-hole resistors at the output rather than an SMD component, so fitting it shouldn't be too difficult.

As a final follow up question, would there be any penalty in using a larger value? I do not have an oscilloscope (yet) and would rather play it safe than try for the smallest possible component. I do not care about some tenths of ohms output impedance -- I don't know much about headphones, frankly, but I am surprised the amplifier even works at all with 10 ohms at the output (much less the "recommended" 39 ohms). With my 32 ohm Grado headphones, that gives a damping factor of just 3.2! Still, it sounds pretty good, though I'd rather have a lower output impedance than that.

Anyway, could I go a bit higher (~330nH) for better stability? Have I got my theory wrong, and lower (~270nH) would be more stable? Thanks again!!

Best,
Matt
 
Not long after AX published Michael Steffes composite OPA1612/TPA6120A2 design in March, 2021, I decided to give his circuit a try. I initially saw both the low input impedance and the signal inversion as potential drawbacks. I went through several iterations and settled on JFET buffers to raise the input impedance. Close listening with a test setup using A-B output polarity switching to the headphones revealed that I could not hear the difference in output polarity. I also found that Jan Didden's Super Regulator Power Supply circuit gave a noticeable improvement over a standard regulated linear supply design.

I'm surprised it took me this long to find this thread but it reawakened my interest. I immediately tried 9" of 22ga enameled wire wound on a 3 watt 10 ohm resistor body and soldered in parallel. This yielded just slightly over 300nF of inductance. This assembly replaced the 10 ohm resistor I had been using in series with the output. I notice a slight but obvious improvement in low bass definition, more so on my Emotiva GR1 cans (32 ohms) than on my Sennheiser HD 650 cans (300 ohms). A worthwhile upgrade! The schematic is attached for those interested in the final design.
 

Attachments

  • CFA HPA V5 Schematic.pdf
    89.7 KB · Views: 125
Member
Joined 2011
Paid Member
Oops, your .pdf schematic and the text of your post #64 mistakenly refer to "nanofarads of inductance" which isn't correct. Inductance is measured in henrys rather than farads.

I personally would have wound the inductor on the body of a 2.2 megohm 3W resistor, and plugged the resulting LR parallel circuit into my inductance meter. The huge resistance won't interfere with the measurement of inductive reactance. Then when the measured inductance is known, I'd move the coil to the 10 ohm resistor. Or just wind another identical coil on the 10 ohm 3 watt resistor.
 
Thanks, Mark, for catching the typos. I'm not sure how Mr. Szykuta came up with 300nH as an inductor value. The circuit was stable without any inductance, only 10 ohms resistance, and seems it would now have effectively only the resistance of the inductor wire at audio frequencies. Google says the inductor has 0.1884 ohms impedance at 100KHz and 0.03768 ohms at 20KHz and less below that. The chart says 9" of 22ga has 0.011 ohms of resistance so it appears the circuit sees the inductor as essentially a short across the resistor at audio frequencies.

I just went out to the shop, wound another inductor on a 10 ohm resistor and left one end of the coil disconnected. Connecting and disconnecting the free end to put the resistor in and out of the circuit made absolutely no difference to the inductance measured, at least with my impedance meter, a PEAK LCR 45 which measures inductance at 200KHz.

Corrected schematic is attached.
 

Attachments

  • CFA HPA V5 Schematic.pdf
    89.6 KB · Views: 125
  • Like
Reactions: 1 user
Anyway, could I go a bit higher (~330nH) for better stability? Have I got my theory wrong, and lower (~270nH) would be more stable? Thanks again!!
Exact value isn't crucial. I guess anything from 200nH to 1uH should work.
Inductor value was chosen with LTspice simulations. It was a quite time ago but I recall simulating several capacitive loads, compensation and inductor combinations. Around 300nH was adequate to obtain good gain and phase margins. Higher inductances shouldn't be required and physical size of higher value air coire inductors might be an constrain.

Nice to see some other similar circuits constructed. High input impedance obtained with JFET buffer might be useful in some cases, especially with weak sources.
Did you tried to measure THD+N?
I'm afraid that noise part of THD+N might be affected by thermal noise as you have 50K pot and some relatively high value resistors in series, potentially degrading amazing ultra-low noise performance of composite amplifiers.
 
I don't have test equipment capable of measuring THD+N at such low levels. Years ago I found that a 50K pot sounded best to me as a load for my various sources so I have more or less standardized on that value. The 2.2K input resistor puts the wiper in my favored part of the log curve on the Alps RK27, around 10 o'clock on the dial for my average listening levels. I might try bypassing that 2.2K input resistor and see if I can hear the difference.

I attribute the extra detail and clarity I perceive listening to the inverting configuration to M. Steffes excellent low noise design. That said, I have built several non-inverting designs similar to yours but using the OPA1612 and find that I prefer them for long term listening. The Steffes circuit is cleaner and more detailed but more fatiguing to listen to. I'm not sure why that is. I've ruled out polarity since I have several pairs of headphones with balanced connections to the drivers available and thus lent themselves to a multi-pole polarity reversal switch that allowed me to hear the difference between polarities directly. Although I can hear a difference on loudspeakers, I cannot on headphones, perhaps because loudspeakers allow the L and R channel audio to combine in the room space in a way that does not happen with headphones.

Another difference as to why the non-inverting design is less fatiguing to my ears is perhaps the lack of JFET buffers. I added the buffers because the sound of the inverting design without buffers is very dependent on the setting of both the input pot and the connected source capabilities. Consumer source equipment just doesn't seem to typically like a load below 1k or 2k. Removing the pot might work with a DAC or other device that has a low output impedance and variable output volume control.

I may try a build an inverting version with some J113 or even some LSK170 devices (if I could afford to buy them) and see if that makes an improvement. I did try your inductor idea on my non-inverting circuit and it helped improve that sound in the low bass on 32 ohm headphones. So I'm still searching for the Holy Grail combination of the open and detailed sound of the Steffes inverting design but with the smoother character and easy listening of the non-inverting design.