OCD log – Part I
Hundred mV of DC offset at the amplifier output is perfectly safe for the loudspeakers and nothing to be bothered with. There is no need to do anything. But….I don’t like to have that while solution is simple and without negative effects.
LuDEF DC offset stability is mostly dependent on T101(BC546) temperature and will inevitably travel some 200 mV form cold to stable hot state. At stable temperature, it will vary very little. It is pretty OK now but summer is approaching. So, I’ve made DC servo circuit which keeps DC at +- several mV.
Attention was payed to minimize any effect it could have on amplifier sound. Servo output is reduced with output resistor by 30 dB at its low pass frequency of 0.16 Hz. That results with servo output attenuation of 70 dB at 16 Hz, 90 dB at 160 Hz, and 110 dB at 1600 Hz. Proper “audio grade” opamp with JFET input and ultra-low distortion was used anyway.
Circuit has enable/disable jumper so it can be easily checked if there is any audible effect and if one belongs to those supernatural beings that can hear details at -100 dB below signal. I can’t. 🙂
If someone wants to build this as well, I’ll attach gerber files.
Hundred mV of DC offset at the amplifier output is perfectly safe for the loudspeakers and nothing to be bothered with. There is no need to do anything. But….I don’t like to have that while solution is simple and without negative effects.
LuDEF DC offset stability is mostly dependent on T101(BC546) temperature and will inevitably travel some 200 mV form cold to stable hot state. At stable temperature, it will vary very little. It is pretty OK now but summer is approaching. So, I’ve made DC servo circuit which keeps DC at +- several mV.
Attention was payed to minimize any effect it could have on amplifier sound. Servo output is reduced with output resistor by 30 dB at its low pass frequency of 0.16 Hz. That results with servo output attenuation of 70 dB at 16 Hz, 90 dB at 160 Hz, and 110 dB at 1600 Hz. Proper “audio grade” opamp with JFET input and ultra-low distortion was used anyway.
Circuit has enable/disable jumper so it can be easily checked if there is any audible effect and if one belongs to those supernatural beings that can hear details at -100 dB below signal. I can’t. 🙂
If someone wants to build this as well, I’ll attach gerber files.
Theoretically you could but most preamplifiers need positive signal for negative feedback. This circuit produces negative feedback DC signal so it could be used only if preamplifier has appropriate point for adding negative DC signal that will directly affect output.
of course it isn't audible
now just show where you did inject servo out ( I see where, but then, I know amp circuit more than well 🙂 )
Scratch those numbers, this DC servo is much more crazy. Actual, in circuit measurement, revealed that low pass frequency is not 0.16 Hz but around 0.002 Hz! Actual attenuation of AC signal at 20 Hz is already 110 dB. How is that possible?
For all bumpkins like me 😉, that casually assume that low pass frequency is determined by R2 C5 constant, it is revealing to check textbooks after observing actual circuit. Low pass frequency of inverting operational amplifier LPF is determined by C5 and resistor in parallel with it. In parallel with C5 is only many megaohms of operational amplifier impedance between input and output.
C5 was reduced to 150 nF, resulting with low pass at 0,01 Hz. That provides AC attenuation of about 90 dB at 10 Hz and 130 dB at 1 kHz and enough servo speed to mostly catch fast changes at power up and Iq ramp up. For DC drift due thermal changes, even first version speed is enough.
DC offset varies +- 1 mV around small fixed offset determined by opamp input DC offset and ground potential difference between DC servo and output grounding points. In my case, it is fixed – 5 mV.
For all bumpkins like me 😉, that casually assume that low pass frequency is determined by R2 C5 constant, it is revealing to check textbooks after observing actual circuit. Low pass frequency of inverting operational amplifier LPF is determined by C5 and resistor in parallel with it. In parallel with C5 is only many megaohms of operational amplifier impedance between input and output.
C5 was reduced to 150 nF, resulting with low pass at 0,01 Hz. That provides AC attenuation of about 90 dB at 10 Hz and 130 dB at 1 kHz and enough servo speed to mostly catch fast changes at power up and Iq ramp up. For DC drift due thermal changes, even first version speed is enough.
DC offset varies +- 1 mV around small fixed offset determined by opamp input DC offset and ground potential difference between DC servo and output grounding points. In my case, it is fixed – 5 mV.
later revision of same amp (which I tried only as proto) with CM based bias circuit ( practically borrowed from Babelfish XA252) is having greater stability of Iq and DC Offset, both in time/temp and rails change domains
though , I'm really finding present iteration more than good enough - I did it deliberately ( to sane level of heatsink temperature) but also accidentally (to insane level of heatsink temperature) - used amp ( well, all of recent ones) with T-Bed fan off .......... nothing to worry about, when speaking of DC offset and Iq change ........... I think that's enough of trying winter-summer conditions in real time
so, most likely not reverting to CM arrangement, that being reserved for Plethora line
though, your investigations are more than useful - fun and progress are often going together
though , I'm really finding present iteration more than good enough - I did it deliberately ( to sane level of heatsink temperature) but also accidentally (to insane level of heatsink temperature) - used amp ( well, all of recent ones) with T-Bed fan off .......... nothing to worry about, when speaking of DC offset and Iq change ........... I think that's enough of trying winter-summer conditions in real time
so, most likely not reverting to CM arrangement, that being reserved for Plethora line
though, your investigations are more than useful - fun and progress are often going together