Hello,
Some time ago, Rod Coleman suggested using a CCS to obtain a clean fixed bias by generating a voltage drop on the grid resistor - the bias power supply and its capacitors are isolated by the constant current sink and are no longer in the signal path.
Great idea! I tried it in my R-C coupled GM70 SE amplifier. The schematic is below.
The bias PSU is well filtered by a MOSfet follower buffered voltage reference, and I tried two different constant current sinks: a cascoded, LED biased bipolar junction transistor CCS, and a 10M45S-DN2540 Mosfet cascode (which works down to 0.5mA current. A DN2540-DN2540 cascode couldn't go below 5.5mA in my tests).
The result - well, this CCS bias definitely sounds better to me than the classic fixed bias! I was very happy, until I checked the signal on the scope...
The negative going swing has its top soft-clipped!!!!!
This only occurs after 18v peak swing with the BJT CCS and after 40v peak with the depletion mode MOSfet CCS. I hadn't heard the distortion because I only listened at a reasonable power level - I have sensitive speakers and neighbours...
Please see the oscillograms below. It's not a hard clipping, the waveform distortion appears at a certain level and remains quite the same up to the maximum swing (120v peak with D3a).
I can't figure this one out - what is causing this distortion and ruining a promising bias topology? (note: everything works perfectly with the classic fixed bias. The load is a 68K resistor as shown in the schematic.)
Some time ago, Rod Coleman suggested using a CCS to obtain a clean fixed bias by generating a voltage drop on the grid resistor - the bias power supply and its capacitors are isolated by the constant current sink and are no longer in the signal path.
Great idea! I tried it in my R-C coupled GM70 SE amplifier. The schematic is below.
The bias PSU is well filtered by a MOSfet follower buffered voltage reference, and I tried two different constant current sinks: a cascoded, LED biased bipolar junction transistor CCS, and a 10M45S-DN2540 Mosfet cascode (which works down to 0.5mA current. A DN2540-DN2540 cascode couldn't go below 5.5mA in my tests).
The result - well, this CCS bias definitely sounds better to me than the classic fixed bias! I was very happy, until I checked the signal on the scope...
The negative going swing has its top soft-clipped!!!!!
Please see the oscillograms below. It's not a hard clipping, the waveform distortion appears at a certain level and remains quite the same up to the maximum swing (120v peak with D3a).
I can't figure this one out - what is causing this distortion and ruining a promising bias topology? (note: everything works perfectly with the classic fixed bias. The load is a 68K resistor as shown in the schematic.)