Current Feedback IC input in a power amp

[steph_tsf]

hello,

I like full symetric designs in power amps but I get bored with the double long tailed pair it demands, especially when wanting a JFET input, plus the two current sources that are needed, plus the four cascodes for isolating the FETs from the high voltage supply rails.

What kind of IC would you recommend as input stage for a power amplifier ?

Like many people, I started thinking about the OPA-134 : JFET input, qualified for audio, low noise, low offset. Is there an accurate OPA-134 Spice model I could use ? The Spice model needs to be accurate regarding the supplies input : how much current they sink in function of the current that's delivered by the output stage.

Then came the idea of getting a Current Mode amplifier as front-end. In the form of an IC. What kind IC would you advise ? Are there Current Mode amplifiers with a JFET non inverting input ?

See attached file. This is a Tina 7 Texas Instruments schematic, ready for simulation. Simulation results are not yet trustable, as the TL071 model is not accurate regarding the supplies inputs. The TL071 output is said to deliver 100mA while we still have moderate currents (like 2 mA) in the supplies ! This ruins the simulation.

[peranders]

Google for "AN-211 analog devices"

My headphone amp QRV03 is based on this design.

[steph_tsf]

Here is a Tina 7 Texas Instruments schematic ready for simulation.
The front-end is a discrete version of a diamond buffer followed by a current-feedback amp. With this sub-optimal arrangement, the diamond buffer is open loop, not included into the feedback, so he adds its own distorsion.

One can see the viability of such approach when conduction a simulation.
-3dB bandwidth is more than 3 MHz
Slew-rate is about 300 v/µs
THD is 0.002%

All this with a VAS having less than 5mA quiescent current (Iq).
With a VAS Iq completely independent from the front-end Iq.
Because of the self-biased cascode arrangement used as VAS.

The reason why this design stands out is in the push-pull current mirror arrangement feeding the VAS. The current mirror arrangement can short circuit the VAS base for fast turn-off, and the current mirror arrangement can dump a high current into the VAS base, far more than twice the front-end quiescent current, for fast turn-on.

This is a kind of "Non Slewing Amplifier", however less complicated than Giovanni Stochino arrangements between 1995 and 1998 in Wireless-World (Electronics World).

Now that the concept got proven (on paper), please advise an IC that could be used as front-end, preferably a JFET input one.

Thanks

[steph_tsf]

Originally Posted by peranders

Google for "AN-211 analog devices".

That's the Alexander Current-Feedback Audio Power Amplifier. My design is somewhat different. It adds flexibility using a dedicated VAS.

The dedicated VAS is push-pull driven for fast slew-rates, relies on a self-biased cascode arrangement, boosts the open-loop voltage gain, and enables a VAS Iq that's completely independent of the front-end Iq.

[janneman]

Originally Posted by steph_tsf

Here is a Tina 7 Texas Instruments schematic ready for simulation.
The front-end is a discrete version of a diamond buffer followed by a current-feedback amp. With this sub-optimal arrangement, the diamond buffer is open loop, not included into the feedback, so he adds its own distorsion.

One can see the viability of such approach when conduction a simulation.
-3dB bandwidth is more than 3 MHz
Slew-rate is about 300 v/µs
THD is 0.002%

All this with a VAS having less than 5mA quiescent current (Iq).
With a VAS Iq completely independent from the front-end Iq.
Because of the self-biased cascode arrangement used as VAS.

The reason why this design stands out is in the push-pull current mirror arrangement feeding the VAS. The current mirror arrangement can short circuit the VAS base for fast turn-off, and the current mirror arrangement can dump a high current into the VAS base, far more than twice the front-end quiescent current, for fast turn-on.

This is a kind of "Non Slewing Amplifier", however less complicated than Giovanni Stochino arrangements between 1995 and 1998 in Wireless-World (Electronics World).

Now that the concept got proven (on paper), please advise an IC that could be used as front-end, preferably a JFET input one.

Thanks

I've used the AD844 in that position, it does accurately reflect the output current into the supply pins, but its not JFET input. And not specified specifically for audio, but that's just marketing of course...

jd

[steph_tsf]

Originally Posted by janneman

I've used the AD844 in that position, it does accurately reflect the output current into the supply pins, but its not JFET input.

Thanks for your reply. The AD844 is indeed a kind of speed monster : 60 MHz bandwith (G=-1), 2000V/µs slew-rate (!), 150µV offset, Iq 6.5mA per amplifier. The input bias current is surprisingly low, only 200nA. But it means a voltage of 2mV into a 10K resistor. That's too much for a ICL (input capacitorless) arrangement. A resistance of 10K resistor is, as an example, the 100K volume pot at -20dB. The voltage variation should be less than 0.1mV, which commands an input bias current smaller than 10nA.

I really would like to try a JFET front-end.

Maybe the AD8033 (single amplifier) or AD8034 (dual amplifier) FET op-amp ?
AD8033 | Very Low-Cost High-Speed FastFET? Op Amps | Operational Amplifiers (Op Amps) | Amplifiers and Comparators | Analog Devices
The AD8033 looks more than decent, with 80 MHz bandwith (G=+1), 80V/µs slew-rate, 1mV offset, Iq 3.3mA per amplifier.

The well known OPA-134 FET op-amp looks slow, with 8 MHz bandwith (G=+1), 20V/µs slew-rate, 1mV offset, Iq 4.0mA per amplifier.
Precision Amplifier - Wide Bandwidth - OPA134 - TI.com

Do you see any other mainsteam JFET IC that could be used as front-end ?

[steph_tsf]

Here is a Tina 7 Texas Instruments schematic ready for simulation.
The front-end is a discrete version of a current-feedback amp.
It is an inverting amp.
The input impedance is 1K, quite low. This is the price to pay for dropping the input diamond buffer (see preview version).

[steph_tsf]

attached you will find two simplified versions. If one can drive a 1K ohm impedance, there is no need for an IC font-end. Four discrete transistors wired as a feedback diamond buffer can do the job. Performance is amazing (on simulation).

[RJM1]

It's not current feedback, but I have always liked the L

[jaycee]

A similar topology is used by ARCAM in some of their amplifiers, usually with a TL072 as the frontend. They use one half as a V to I stage, and the other as a servo.