I have build the Starless circuit and got it working. I did not select the semiconductors, did not use the differential feedback resistor and used 1% metal film resistors. As expected DC offset was firmly clamped at +13V with a plus-minus 18V supply. I was able to trim that down to plus-minus 500mV by adjusting the output emitter resistor in the negative current mirror with a 15kHz trimmer. I will look tomorrow if the feedback resistor helps.
There is of cause always the cap or servo solution.
There is of cause always the cap or servo solution.
I build the starless in it´s most basic form. I even used a pimped B1 buffer for simplicity.
I did some experiments with servos but ultimately decided on the cap solution. Used in this way it does not do harm i think and it is simpler and less expensive. I used an 120nF Evox MMK just because i have them. 10nF is enough in this position so use what you like.
Interesting enough the buffer worked without bleeder resistor so the little leakage the cap has seems to be enough to ignite the Fet. more later...
I did some experiments with servos but ultimately decided on the cap solution. Used in this way it does not do harm i think and it is simpler and less expensive. I used an 120nF Evox MMK just because i have them. 10nF is enough in this position so use what you like.
Interesting enough the buffer worked without bleeder resistor so the little leakage the cap has seems to be enough to ignite the Fet. more later...
As i said the circuit had initial offset that was way too high. This can be trimmed with the output emitter resistors in the mirror. 30kOhm added gave around plus-minus 500mV so this is good enough. The rest is taken care by the cap. See the circuit as build. I will build now the other channel and listen.
I forgot to tell you how the circuit measured. The square wave looks the same over a wide dynamic range so this circuit is not easy to overload. RIAA precision is plus-minus 0.2dB with the values shown. I used 1% metal film resistors and surplus styroflex caps of unknown origin. The circuit tracks the RIAA curve until 360kHz - 3dB, the null in my anti RIAA so it is wide band. I decided to lower the gain. It should be around 60dB theoretically. I measured 57dB with my Sennheiser tube meter. This is enough in my system. I have a rather loud main tube amp and high sensitivity speakers so usually use no line stage. The tube amp has an input volume control.
I forgot to tell you how the circuit measured. The square wave looks the same over a wide dynamic range so this circuit is not easy to overload. RIAA precision is plus-minus 0.2dB with the values shown. I used 1% metal film resistors and surplus styroflex caps of unknown origin. The circuit tracks the RIAA curve until 360kHz - 3dB, the null in my anti RIAA so it is wide band. I decided to lower the gain. It should be around 60dB theoretically. I measured 57dB with my Sennheiser tube meter. This is enough in my system. I have a rather loud main tube amp and high sensitivity speakers so usually use no line stage. The tube amp has an input volume control.
I used the ( enhanced ) Wilson because i wanted to have a basic starting point. This can be improved on and i may. Yes, this circuit does not work with a floating supply.
My plan is not to develop this any further. It will be interesting to listen to this because the circuit produces nearly only second harmonic but a bit more then i would wish. I expect a sound that is very similar to tubes without NFB what is not a bad thing. The next circuit i will show is based on your CEN. It should have less second. This circuit will hopefully then further developed to it´s optimum.
My plan is not to develop this any further. It will be interesting to listen to this because the circuit produces nearly only second harmonic but a bit more then i would wish. I expect a sound that is very similar to tubes without NFB what is not a bad thing. The next circuit i will show is based on your CEN. It should have less second. This circuit will hopefully then further developed to it´s optimum.
Sure, i will give it a good listen.
The next circuit i will explore is called The Naked Truth. It has a lot in common with the Starless but is parallel symmetric for some second cancelation. Here you find the helper transistors and the differential feedback resistors. I have already tried this in other circuits i have shown a long time ago. The helpers make the sound more dynamic so i already know that it is beneficial to include them. This modification can also be done on the Starless. For the moment i keep it as is to compare what makes what. I may later rebuild the Starless. There will also be a BJT version of the Naked Truth.
The next circuit i will explore is called The Naked Truth. It has a lot in common with the Starless but is parallel symmetric for some second cancelation. Here you find the helper transistors and the differential feedback resistors. I have already tried this in other circuits i have shown a long time ago. The helpers make the sound more dynamic so i already know that it is beneficial to include them. This modification can also be done on the Starless. For the moment i keep it as is to compare what makes what. I may later rebuild the Starless. There will also be a BJT version of the Naked Truth.
If one wants really good 2nd cancellation in CEN, one might pick a J74 with 0.8mA higher Idss and then add a degeneration R of 5R1.
This will help to match the transconductance of the P-FET to the N-FET.
See my F5X thread and the earlier F5 / F5 balanced threads for detail.
Or go balanced.
Patrick
This will help to match the transconductance of the P-FET to the N-FET.
See my F5X thread and the earlier F5 / F5 balanced threads for detail.
Or go balanced.
Patrick
Here is the circuit that i have being talking about. The mirror with cascode. It should be lower noise and distortion plus better DC stability because the bases are clamped by Leds.
It is even not very complicated. See also the EUVL trick of distortion cancelation. One more small improvement and we are there what i really would like to build.
It is even not very complicated. See also the EUVL trick of distortion cancelation. One more small improvement and we are there what i really would like to build.
Last edited:
R2 is not needed. C8 shunts the signal to ground. I included this resistor by default because the input capacitance of the buffer is not linear. Usually i also put a 150pF to ground. At very high frequencies there is a null and then there is stray inductance so the impedance goes up again. This RC filter makes sure that the input signal is bandwidth limited and no signal goes into this problem zone. As i said, in this case it is not necessary.