Aleph X - rootcause of absolute DC offset drift

[uli]

And I can measure much more drift on Vgs for the upper FETs (biased by the BJT) than the lower ones.



Patrick

The problem is, that the (negative) temperature coefficient of the Aleph current source is defined solely by the tempco of the BJT (~ -2mV/K) in it. The tempco of the "lower" output stage is influenced by a) the tempco of the used Zener (or other reference), b) the tempco of the CCS Mosfet and c) the tempco of the outputstage.

To bring this into synchronicity is kinda task

Uli

 

[kropf]

An active offset regulator will work fine. If I ever get tired of
adjusting the pot in production, I'll put one in.

I have tried a servo to adjust the current through the differential pair and it
works very nicely. I haven't installed it in my main amps to see if it does
anything to the sound. The reason I investigated this was to run a low-V
(series transformer primaries) and high-V mode (parallel transformer
primaries). Trouble is, I had to adjust the current source with every
switch.

Filtering the audio bandwidth from the servo output is quite easy since
summing the balanced output leaves mainly the common mode DC
absolute offset.

Keep in mind that the value of resistors from output to the
Sources of the input diff pair are very dependent on the specific
amplifier - 4.7K is a typical figure, but the overall performance
varies with the values, so you have to test them not only for
efficacy in controlling offset, but also distortion vs amplitude and
frequency and CMRR.

Does anyone have a theory on why the resistors from the output to the
source degrade the performance? Some ideas:

(a) A DC problem: The resistors to the output make the current source
for the differential pair less ideal and more like a resistor current source.
This could affect the X feedback.

(b) An AC problem: The output stages are not perfectly balanced, so
some of the audio signal is fed back into the differential pair. This
increases distortion.

If (b) is the main problem, the audio frequencies could be filtered passively
rather than with a servo. Tie tne outputs together with, say, 220 ohm
(5W) resistors and then shunt to the negative rail with a 220uF cap. Take
a 1k resistor from this point to the diff pair. This will clamp the absolute
offset pretty hard, but attenuate a 60Hz common mode signal
by about 20dB (higher frequencies, much more).

Comments?

Jeremy

 

[macka]

I mess with this for a while.

Per Nelson's guidance R1, R4, R44 and R45 help control absolute DC offset at the outputs by sinking current to ground when the outputs sit either above or below 0V potential.

I installed 50 watt 30 ohm power resisters.

What is interesting is that the McMillian resisters if chosen carefully for a particular scale of the design work well in terms of early control of the offset. I left it at 11K ohm in consideration of the overall effect being largely heat related on the input pair.

The initial absolute DC offset at the outputs does effect the relative offset and I saw 100mv at start up then went down to 10 mv in an hour. This is because the absolute offset drives more current through one side of the amp fets and less though the other so one side gets hotter then the other.

Finally I installed a 1/2 bias switch which is like standby mode and pre heats the heatsinks to around 125 watts dissipation.

The beauty of this arrangement is that at inital turn on there is only half the bias current and a relatively low absolute offset of about 5-6 volts. After about 15 minutes I flick the toggle switch to full bias and with the amp pre heated the absolute offset swings nicely from 5 volts down to under 1 volt over a period of 30 minute. The relative offset remains very stable but I use Munforf inout capacitors to minimise effects of the upstream signal.

Switching the bias is seemless and it saves on power (bills) and stressing the amp during warm up from cold.

Macka

 

[kropf]

Finally I installed a 1/2 bias switch which is like standby mode and pre heats the heatsinks to around 125 watts dissipation.

The beauty of this arrangement is that at inital turn on there is only half the bias current and a relatively low absolute offset of about 5-6 volts. After about 15 minutes I flick the toggle switch to full bias and with the amp pre heated the absolute offset swings nicely from 5 volts down to under 1 volt over a period of 30 minute. The relative offset remains very stable but I use Munforf inout capacitors to minimise effects of the upstream signal.

Switching the bias is seemless and it saves on power (bills) and stressing the amp during warm up from cold.
Macka

1/2 current or 1/2 voltage, that is the question? If the switching the
current keeps me from having to adjust the amp, it looks like a good idea
to me. Where did you install the switch for 1/2 current? There would be
several ways to do it.

Jeremy

 

[gl]

Hi Ian,

The list that moe29 just posted is the definitive one posted by Nelson. It's one of the real pearls in the ocean of stuff published about the Aleph-X. I notice that you said that you use 30 ohm 50W resistors as output to ground loading resistors. I really think you should change these to 100 ohms. I suspect (but don't absolutely know) that the sound of your amps will improve. JMHO.

Graeme

 

[macka]

I found that 30 ohms gave more control of the offset.

To my way of thinking its a 60 ohm resister across the load with a split T junction centre tapped to earth

I just did what worked for me.

Greys original circuit used 30 ohms without complaint and I think its comes down to the specific circuit.