Noisy noise generator

[Martin Czech]

:::Here is the schemo and two photos of the experimental board:
:::
:::http://www.geocities.com/m100de/noise.html
:::
:::What am i doing wrong ???

Now I'm leaving solid ground, since I can't fumble with
your circuit, just take the flameproof suit and take a look...


The gain of the opamp is about 270. That's quite a lot!
You observe 100mV of noise, this means only 370uV at the input.
Now, 370uV is not so much, all our environment is polluted with
50/60 Hz stuff and harmonics.

Problems where to be expected, since:

-high gain
-no enclosure
-large distances from resistors to opamps (I think so...)
-you filter with ground reference, but your diodes are floating
 with respect to ground. In turn, ground is the reference for your opamp.
 Any unsymmetry (components) will therefore weaken the filtering.
 A lot of components for a filter that is not very usefull, as you allready
 found out. 
 
I think the idea to throw out the dc by taking two diodes was
interesting, but this symmetrical approach fails due to the
mentionened reference problems. A bridge circuit with
difference amplifier would be better, but appears somehow
overkill to me.

I'd rather use a single diode to reference (i.e. ground) with an moderate
gain inverting amplifier maybe with a second amp stage behind, and an
integrator overall feedback loop to servo out any dc. This has the
advantage that really OUTPUT dc is canceled (due to servo loop) and
that you are free to choose integration time constants/ RC component
values, this could be made variable (pot, switch) and allows for very
long t=RC. Words don't serve well to describe circuits, I could draw a
sketch on paper and scan that.

I'd also use a good enclosure.


You know that a TL8x has 3MHz gain bandwidth product, which
means about only 11kHz bandwith for your circuit.

This all may sound indoctrinating, please keep in mind I'm
no native speaker (writer). 

m.c.

p.s.: 

does anybody know what the translation of "Oberlehrer"
oder "oberlehrerhaft" is?

m.c.