A compact, composite I/V stage with OPA1612 in DFN and TPA6120A2 in VQFN definitely sounds like something I would be interested in. I shall simulate some stuff before I go ahead with my ES9028/38Pro layout.
@JensH: Feel free to share your knowledge
Do you mind telling how much gain you had in the inner loop? Just in case you happen to remember it - not asking to dig out the schematics or old sim files.
@JensH: Feel free to share your knowledge
Do you mind telling how much gain you had in the inner loop? Just in case you happen to remember it - not asking to dig out the schematics or old sim files.I tried gains of 1 and 2 in the inner loop. Not much difference. With no load perhaps a small improvement with 2x, but with load I saw a slightly better result with 1x, at least at 10 kHz. At 1 kHz and 50 ohm load I got exaxtly the same THD with OPA1612. So all within a couple of dB difference.
At 10 kHz the distortion with load was around -148dB.
At 10 kHz the distortion with load was around -148dB.
5V total supply rails.
Distortion performance probably isn't up to the task for ES9038PRO.
Last edited:
It's 5V single supply, so +/- 2.5V split, but it's worse than that because it's not close to rail-to-rail.
So, is contingent on you getting a 135 MHz GBW amp stable with a TPA6120 in both loops with a gain of 10 or around there.
I used the same circuit for both measurements.
I didn't try it at 20 kHz. I don't have an ultra low distortion oscillator at 20 kHz, so I didn't even consider it. Perhaps some other oscillator is good enough though.
The TPA6120A2 is a rebranded DSL line driver. TI has a whole family of them, some as cheap as $1.50. They would need to be cheap for CPE applications. Products | Line Drivers | Amplifier ICs | TI.com Most seem to be characterized for low distortion at 1 MHz and able to drive as little as 25 Ohm.
I think an IM test- 19 KHz and 20 KHz would be a more practical and relevant way to test opamps. Its all "in band". You would need a good passive low pass filter to remove the 19 +20 HKz. The good part is that they do not need to be particularly good sources.
I think an IM test- 19 KHz and 20 KHz would be a more practical and relevant way to test opamps. Its all "in band". You would need a good passive low pass filter to remove the 19 +20 HKz. The good part is that they do not need to be particularly good sources.
I think most here know about the 'similarities' between TPA6120A2 and THS6012 and use the SPICE model of the latter for simulations, since TPA6120A2 doesn't have one. TPA6120A2 are much cheaper in low quantities than any of the other DSL Line drivers AND even among the DSL line drivers they are among the best performers, together with THS6182 and some others maybe (I haven't looked at all the datasheets in detail).
TPA6120A2 has sweet low quantity price though.
But the extra feedback at 1kHz will obscure the distortion, and the lowpass filter will get rid of the interesting harmonic skirt around the 19k/29k peaks...
If one of the peaks was 32.768kHz, could a watch crystal act as a notch filter? (although it would probably add distortion of its own). Or a 20k notch filter implemented in the same way as the 1kHz notch filter... after all, getting rid of one peak is enough to make sure the acquisition downstream does not make more IMD...
> The good part is that they do not need to be particularly good sources.
But the two opamps that drive each tone into the summing node through resistors have 19k+20k output currents, so their output stage might join the IMD party.
But the extra feedback at 1kHz will obscure the distortion, and the lowpass filter will get rid of the interesting harmonic skirt around the 19k/29k peaks...
If one of the peaks was 32.768kHz, could a watch crystal act as a notch filter? (although it would probably add distortion of its own). Or a 20k notch filter implemented in the same way as the 1kHz notch filter... after all, getting rid of one peak is enough to make sure the acquisition downstream does not make more IMD...
> The good part is that they do not need to be particularly good sources.
But the two opamps that drive each tone into the summing node through resistors have 19k+20k output currents, so their output stage might join the IMD party.
About composite amplifiers, here's a very simple old technique that I experimented with. It can be easily designed, analytically or by simulation, for any closed loop gain. The only precondition is to have an accurate AC model for the op amp(s), which is usually available today (the models distributed for free suck at transient (large signal) analysis, being otherwise quite accurate).
The idea is pretty obvious: a composite amp, designed as a low pass filter of n-th order, with the additional constraint for the overall filter transfer function to be stable, which is if and only if all its poles are inside the unit circle in the Z plane. Known filter coefficients can be used, so the design is simple.
I did not try over n=3, but I suspect the sensitivity increases quickly with the filter order, to the point where the natural op amp AC parameters variations become critical and everything turns into a gamble.
I did not try over n=3, but I suspect the sensitivity increases quickly with the filter order, to the point where the natural op amp AC parameters variations become critical and everything turns into a gamble.
Further to my previous post, a functioning prototype was presented on another forum:
Пидсылювач "Крокодил" или "цiкавi дослiди"
including the distortion spectra which IMHO look very good.
However, a member who apparently also built a prototype mentioned "problems with steep fronts".
Пидсылювач "Крокодил" или "цiкавi дослiди"
including the distortion spectra which IMHO look very good.
However, a member who apparently also built a prototype mentioned "problems with steep fronts".
Yes, it's his approach to bootstrapping that he introduced in his headphone amplifier
"The Crocodile"
Re. his plots on the Ukrainian forum, the last three are repeated and explained in the post I quoted earlier, i.e.
"The Crocodile"
and the first two refer to a measurement at a frequency of 31.5Hz.
I'll be glad to translate whatever info from the Ukrainian forum if there is interest..
"The Crocodile"
Re. his plots on the Ukrainian forum, the last three are repeated and explained in the post I quoted earlier, i.e.
"The Crocodile"
and the first two refer to a measurement at a frequency of 31.5Hz.
I'll be glad to translate whatever info from the Ukrainian forum if there is interest..
We had something similar in the PGP amp, where we avoided a ton of cascodes by feeding the gain stages +/- power supplies modulated by a copy of the input signal. But coupling was DC, not AC like here.
THD is already spec’d at -114dB @1KHz and G=+1 for the TLV172, what else is to be bootstrapped for an headphone amp?
THD is already spec’d at -114dB @1KHz and G=+1 for the TLV172, what else is to be bootstrapped for an headphone amp?