Doh! You're right and I misread it. Thanks for correcting me, and therefore also for answering the question of build-out resistors.
Much thanks,
Chris
not prejudging - just suggesting a modern "benchmark" to compare to
not even considering a 400 mA current feedback DSL driver? - I don't think you can come close objectively with 1/2 dozen discretes if +/-15 V supply is adequate - as it is for most dynamic headphones
the TPA6120/THS6012 DSL driver chip is built on a GHz process, closed loop Av +10 gain corner is higher than many medium power audio amplifier driver Q ft
the DSL driver application is an interesting modern day reprise of the original motivation for low distortion amplification in long distance FDM telephony that led Black to invent negative feedback
today the QAM digital modem signal densely packs the Shannon-Hartley "Channel Capacity" with a high level noise-like signal over 100 kHz to low MHz, demanding ultra low IMD to prevent cross channel S/N loss from the distortion products
essentially the working mode of these chips is a very demanding "noise fill"/Noise Power Ratio test
while driving 12- 25 Ohms at the line matching xfmr primary
not hard to see why they might make good audio line driver, headphone amplifier candidates
and there are other professional electronics desingers with decades experience here too John - by posting in a public forum you are inviting a conversation
not even considering a 400 mA current feedback DSL driver? - I don't think you can come close objectively with 1/2 dozen discretes if +/-15 V supply is adequate - as it is for most dynamic headphones
the TPA6120/THS6012 DSL driver chip is built on a GHz process, closed loop Av +10 gain corner is higher than many medium power audio amplifier driver Q ft
the DSL driver application is an interesting modern day reprise of the original motivation for low distortion amplification in long distance FDM telephony that led Black to invent negative feedback
today the QAM digital modem signal densely packs the Shannon-Hartley "Channel Capacity" with a high level noise-like signal over 100 kHz to low MHz, demanding ultra low IMD to prevent cross channel S/N loss from the distortion products
essentially the working mode of these chips is a very demanding "noise fill"/Noise Power Ratio test
while driving 12- 25 Ohms at the line matching xfmr primary
not hard to see why they might make good audio line driver, headphone amplifier candidates
and there are other professional electronics desingers with decades experience here too John - by posting in a public forum you are inviting a conversation
Last edited:
"Class A discrete Output" isn't a "proper" performance based design constraint
it is a prescription of how to do something - not a set of objective (or even subjective - as long as you only use your ears) requirements on how well it performs
may well be dictated by Audiophile market prejudices - demanded by your customers that you can call some part of the amp discrete
the expected 5-10 mA internal bias of the TPA6120 output Q gives 10-20 mA push-pull Class A - over 100 dB SPL in Class A for many headphones - proportionality much more than a 200 W power amp’s "first Watt" Class A from several 100 mA output Q bias
I've actually paralleled and biased the 2 amps in each TPA6120 chip against each other thru 2x 1 Ohm current sharing R
the bias V was controlled to give 200 mA standing bias for +/- 400 mA Class A push-pull operation
the bias was switchable - I and much younger ears - some who claimed to easily identify different op amps in headphone amps couldn't hear any difference in informal tests switching the bias on/off
with the TPA6120 inside the input op amp feedback loop I couldn't measure "any" distortion with indirect IMD test which could resolve -160 dB - with Class AB output op amp operation
so I really don't believe the chip’s output itself is lacking compared to discrete Class A - the only advantage at these performance levels is that Class A is more forgiving of layout, poor PSRR - any poor power/ground routing decisions just give linear gain, crosstalk errors with Class A
PSRR is addressed by sub-regulating the input op amp supply as I mentioned
to get extremely good performance with Class AB the nonlinear "half wave rectified" +/- power supply pin currents coupling to everything else needs care
Bruce Hofer's slides at the Audio Precision "Master Class" seminar showed some layout tricks for smt parts, op amps I hadn't seen before
but I don't see "Class AB" PS, common impedance distortions in my current layout "only" informed by the usual low noise, precision design books, articles, my experience putting Av 4000 strain gage amps on boards with x86, Ethernet, DSP and digitizing the analog at 16 bits, looking very closely for cross coupling, spurs, sub lsb linearity
it is a prescription of how to do something - not a set of objective (or even subjective - as long as you only use your ears) requirements on how well it performs
may well be dictated by Audiophile market prejudices - demanded by your customers that you can call some part of the amp discrete
the expected 5-10 mA internal bias of the TPA6120 output Q gives 10-20 mA push-pull Class A - over 100 dB SPL in Class A for many headphones - proportionality much more than a 200 W power amp’s "first Watt" Class A from several 100 mA output Q bias
I've actually paralleled and biased the 2 amps in each TPA6120 chip against each other thru 2x 1 Ohm current sharing R
the bias V was controlled to give 200 mA standing bias for +/- 400 mA Class A push-pull operation
the bias was switchable - I and much younger ears - some who claimed to easily identify different op amps in headphone amps couldn't hear any difference in informal tests switching the bias on/off
with the TPA6120 inside the input op amp feedback loop I couldn't measure "any" distortion with indirect IMD test which could resolve -160 dB - with Class AB output op amp operation
so I really don't believe the chip’s output itself is lacking compared to discrete Class A - the only advantage at these performance levels is that Class A is more forgiving of layout, poor PSRR - any poor power/ground routing decisions just give linear gain, crosstalk errors with Class A
PSRR is addressed by sub-regulating the input op amp supply as I mentioned
to get extremely good performance with Class AB the nonlinear "half wave rectified" +/- power supply pin currents coupling to everything else needs care
Bruce Hofer's slides at the Audio Precision "Master Class" seminar showed some layout tricks for smt parts, op amps I hadn't seen before
but I don't see "Class AB" PS, common impedance distortions in my current layout "only" informed by the usual low noise, precision design books, articles, my experience putting Av 4000 strain gage amps on boards with x86, Ethernet, DSP and digitizing the analog at 16 bits, looking very closely for cross coupling, spurs, sub lsb linearity
Last edited:
Ah thanks for the heads up jcx - I hadn't twigged that the TPA6120 was in reality just a re-badged xDSL driver. I recently picked up some desoldered AD8016 at a bargain price and was struck by the datasheet similarity of this to the TPA part without realising they were actually designed for the same application...
Now let me point something out here. I am designing a headphone amp, and one of the most annoying things that I have experienced is the 'burning' in my ears due to a 'compromised' HD drive amp. Therefore, I choose Class A, at the very minimum, in order to reduce any 'listener fatigue'. This is why I chose the approach of adding an external discrete output stage. All else being equal, the inherent higher order distortion reduction by not stressing the IC output stage, and the LOWER thermal feedback (ref. Solomon) should make a design above and beyond the normal. Who knows? Maybe, even an A rating in 'Stereophile'? That's what makes a successful audio design, a little concern for the 'subjective' factors. '-)
Really John I’d like to think you can recognize a description of a composite op amp audio line drive similar to Walt Jung’s circuits in his Op Amp Applications book – the TPA6120/THS6012 is just a little advanced over his AD811 example output op amps
I'm pretty sure my posts already pointed to the likely internal thermal feedback distortion at low frequency with the TPA6120 on the graphs showing it delivering ~1 W
why the THD rises to -90 dB!
any guestimate on your discrete Class A stage distortion? – I’d guess that -80 dB THD is very good for a complementary follower working at a reasonable fraction of its bias – is your common emitter stage going to do any better (hint: Dr Edward Cherry’s last JAES article lets you answer this easily)
I am proposing using the TPA6120 as just the output stage - using "good", “audio” op amp singles for each channel for the input, wrapping global feedback around the TPA - fixing up the slow, low order thermal errors of the TPA, any channel crosstalk
that gets the same advantages you are talking about with the discrete Class A output stage of negligible load (300 kOhm, 1.4 pF for the TPA6120 + input) and thermal isolation of separate, lightly loaded input op amps
I did mention taking V gain in the output op amp so the input op amps can work from subregulated supplies
one way is with constant output op amp local feedback gain == total amp gain – so the input op amp is working “unity gain”
another is to use modified integrator feedback around the output to give huge audio frequency gain – the input op amp only has to swing mV (yes I know how to control the clipping recovery despite the region of ~ 2nd order loop gain)
I'm pretty sure my posts already pointed to the likely internal thermal feedback distortion at low frequency with the TPA6120 on the graphs showing it delivering ~1 W
why the THD rises to -90 dB!
any guestimate on your discrete Class A stage distortion? – I’d guess that -80 dB THD is very good for a complementary follower working at a reasonable fraction of its bias – is your common emitter stage going to do any better (hint: Dr Edward Cherry’s last JAES article lets you answer this easily)
I am proposing using the TPA6120 as just the output stage - using "good", “audio” op amp singles for each channel for the input, wrapping global feedback around the TPA - fixing up the slow, low order thermal errors of the TPA, any channel crosstalk
that gets the same advantages you are talking about with the discrete Class A output stage of negligible load (300 kOhm, 1.4 pF for the TPA6120 + input) and thermal isolation of separate, lightly loaded input op amps
I did mention taking V gain in the output op amp so the input op amps can work from subregulated supplies
one way is with constant output op amp local feedback gain == total amp gain – so the input op amp is working “unity gain”
another is to use modified integrator feedback around the output to give huge audio frequency gain – the input op amp only has to swing mV (yes I know how to control the clipping recovery despite the region of ~ 2nd order loop gain)
If I can get a word in edgewise. I EXPECT my headphone amp to draw 100ma or so from the discrete output stage. This way, the beta is relatively flat, the IC doesn't do much to contribute directly to the output. The effective slew rate increases many times. With a good design, the harmonic and IM distortion should be virtually unmeasurable with normal test equipment at ALL levels below 10V peak. Of course, the design will have SOME distortion, but it will be minimal. The thermal distortion ref. Solomon, should be very low. The discrete output devices should be POWER transistors, and they MUST be heatsinked to work well. As I said before, IF I were trying to design a minature or class B type design, I would just select the best IC I could find and just use it.
What I am trying to show here is a relatively unique design approach that is more than 40 years old, and has mostly been forgotten. The ADVANTAGE of this approach is it helps out the IC op amp a lot, and these circuits can sound very good, better than one would expect. That is the primary advantage.
What I am trying to show here is a relatively unique design approach that is more than 40 years old, and has mostly been forgotten. The ADVANTAGE of this approach is it helps out the IC op amp a lot, and these circuits can sound very good, better than one would expect. That is the primary advantage.
All right, let us determine what we need for an easy Class A headphone amp.
First, voltage swing: I can believe that +/- 15V being easy to make, should be enough. IC selection is relatively easy, and power device selection can be VERY easy. I find the T0-220 power transistor package the most efficient and easiest to heatsink.
I am going to presume that 30 ohms is the lowest Z load. This is relatively arbitrary, but I just might make the output resistor that value. In any case 15V/30 = 0.5A This is the maximum design current, for safety, that we have to recognize. Very few useful T0-220 power transistors can't meet that requirement. Selection should be relatively easy.
Now why 100ma for output Iq? Well, we COULD go higher, but things will just heat up and the heatsink will get even bigger. Is it enough? Well, 100ma Iq = 200ma Ipk and this gives 6Vpk or 4V rms out, worst case. That should be enough. However, that sure beats 10ma from the TPA1620. Check it out.
First, voltage swing: I can believe that +/- 15V being easy to make, should be enough. IC selection is relatively easy, and power device selection can be VERY easy. I find the T0-220 power transistor package the most efficient and easiest to heatsink.
I am going to presume that 30 ohms is the lowest Z load. This is relatively arbitrary, but I just might make the output resistor that value. In any case 15V/30 = 0.5A This is the maximum design current, for safety, that we have to recognize. Very few useful T0-220 power transistors can't meet that requirement. Selection should be relatively easy.
Now why 100ma for output Iq? Well, we COULD go higher, but things will just heat up and the heatsink will get even bigger. Is it enough? Well, 100ma Iq = 200ma Ipk and this gives 6Vpk or 4V rms out, worst case. That should be enough. However, that sure beats 10ma from the TPA1620. Check it out.
- Status
- Not open for further replies.