This is a way to use a single-secondary transformer in a dual rail supplied amplifier with no coupling capacitors.
To achieve that, I used an OpAmp (LTP can be used too if one wants discrete solution) to split single rail into dual-rail PSU.
The 24V AC gives about 32V DC after rectification and filtering and we turn it into +/-16V DC by using a voltage divider (R3,R13) and applying its middle node to OpAmp's input so the feedback loop keeps amp's output at the same potential (which we declare to be a ground so the output sits on zero volts potential and the 32V supply voltage turns into +/-16V maintained by OpAmp).
One thing worth mentioning is that PSRR is really good, the amp is totally silent on 93dB/1W speakers (the output noise is less than 100uV and I can't measure lower than that at this moment) with only 2 x 10,000uF filtration per channel.
I chose inverting topology because my preamp is inverting too, but non-inverting mode can be used with the same result.
I tested different OpAmps (AD797, AD825, AD829, LT1122, OPA637, to name some) but the ADA4898-1 sounds best to me. Of course, each of them needs different compensation scheme to stay stable (AD825 and LT1122 are stable without any compensation - not even C10 is needed - and they sound very good too).
The gain is 20dB (10V/V, set by R7/R2 ratio). Zin is determined by the value of R2 and the feedback loop provides very low Zout (we have 3 digit damping factor here so speaker choice is not critical).
The output power is about 20W_peak at 8 Ohms load and about 36W_peak at 4 Ohms load which i consider more than plenty for home use.
The bias current of the output stage is set by voltage drop at D1-D4 diode string - 4 diodes give about 350 mA through MOSFETs (2.1V of total Vgs). If 3 diodes are used it will be about 200mA and 5 diodes will generate a bit over 0.5A bias current so you can choose in accordance with your preferences and available heatsinking, but more than 0.5A will not bring much since we here enjoy the square law output characteristic of the MOSFETs which enables them to stay "longer" in the class A regime.
Output waveforms look pretty much the same as they do here (post #1)
At the end, it sounds so good that I decided to move it into this small case (post #28) which is now a bit heavier than before (weight difference between SMPS PSU and two 80VA toroids).
Off course, one can choose the topology freely: OpAmps, discrete, BJTs, MOSFETs, etc... - power can be up/down scaled too if you wish - the main point is not to discard good single- secondary transformer just because one thinks that it makes coupling caps mandatory
To achieve that, I used an OpAmp (LTP can be used too if one wants discrete solution) to split single rail into dual-rail PSU.
The 24V AC gives about 32V DC after rectification and filtering and we turn it into +/-16V DC by using a voltage divider (R3,R13) and applying its middle node to OpAmp's input so the feedback loop keeps amp's output at the same potential (which we declare to be a ground so the output sits on zero volts potential and the 32V supply voltage turns into +/-16V maintained by OpAmp).
One thing worth mentioning is that PSRR is really good, the amp is totally silent on 93dB/1W speakers (the output noise is less than 100uV and I can't measure lower than that at this moment) with only 2 x 10,000uF filtration per channel.
I chose inverting topology because my preamp is inverting too, but non-inverting mode can be used with the same result.
I tested different OpAmps (AD797, AD825, AD829, LT1122, OPA637, to name some) but the ADA4898-1 sounds best to me. Of course, each of them needs different compensation scheme to stay stable (AD825 and LT1122 are stable without any compensation - not even C10 is needed - and they sound very good too).
The gain is 20dB (10V/V, set by R7/R2 ratio). Zin is determined by the value of R2 and the feedback loop provides very low Zout (we have 3 digit damping factor here so speaker choice is not critical).
The output power is about 20W_peak at 8 Ohms load and about 36W_peak at 4 Ohms load which i consider more than plenty for home use.
The bias current of the output stage is set by voltage drop at D1-D4 diode string - 4 diodes give about 350 mA through MOSFETs (2.1V of total Vgs). If 3 diodes are used it will be about 200mA and 5 diodes will generate a bit over 0.5A bias current so you can choose in accordance with your preferences and available heatsinking, but more than 0.5A will not bring much since we here enjoy the square law output characteristic of the MOSFETs which enables them to stay "longer" in the class A regime.
Output waveforms look pretty much the same as they do here (post #1)
At the end, it sounds so good that I decided to move it into this small case (post #28) which is now a bit heavier than before (weight difference between SMPS PSU and two 80VA toroids).
Off course, one can choose the topology freely: OpAmps, discrete, BJTs, MOSFETs, etc... - power can be up/down scaled too if you wish - the main point is not to discard good single- secondary transformer just because one thinks that it makes coupling caps mandatory
