>Here are my questions:
>1) How can we calculate the DC voltages of this stage. Opamps should be
>easy, right ;-)
Ok, it all starts with the +8v feeding in thru R4(510k)...that's
the biasing reference voltage going into U4 pin 3. Since U4's bias current
is so small, there'll be a very small D.C. voltage drop across R4, and
pin 3 will end up with almost +8v too.
Now at 0Hz (D.C. voltage), U4 has INCREDIBLE gain. That is, if
there's ANY voltage difference between the two input pins (pin 2 and
pin 3), the output D.C. voltage (at pin 1) will saturate:
if pin 2 is ABOVE pin 3, the output will saturate down near zero volts...
if pin 2 is BELOW pin 3, the output will saturate up near +12v.

Notice that there's a DC path from the output (pin 1) back to the "-"
input (pin 2) through R7(510K). Once again, the D.C. current drawn
into pin 2 through this resistor is tiny (insignificant). So to keep
pin 2 = pin 3, the output will have to sit very close to +8v.
So all three pins sit close to +8v.
If you were to drag pin 3 down to +6v, the other two pins would follow.
Ain't negative feedback wonderful?

>2) How do we calculate the gain of this stage. The manual says 30db.
>What is the voltage gain?
This is AC gain (DC gain is zero since C20 and C21 block DC
voltages). AC gain is set mostly by the ratio of R7(510K) to R2(10K).
This gives a gain of about 50. Actually, source impedance is a little
higher than 10k, since we should add U3's output resistance of 1500
ohms (reducing gain a little).
That's where the 30 dB comes from. But you can consider that gain is
actually twice that, since BOTH output signals from U3 pin 4 and U3
pin 5 are used to drive U4 (each have the same AC amplitude). This is
a differential amplifier configuration.

>3) Why is C21 smaller than C20?
To find the answer, you should look at the input impedance
going into each input to U4 (this gets tricky):
Input Z looking into R2 is 5K, not 10k as you might expect. Consider
that when the input side of R2 is going UP, its output side is going
DOWN by about the same amount. Remember, pin 2's voltage tracks pin
3's voltage...and pin 3 is driven from U3 with a signal of similar
amplitude, but opposite polarity.
Input Z looking into R3 is very very high, since each end of
R3 has the same AC voltage on it....except for the effect of C22(150pf)
and R4(510K). R4 actually sets the upper limit on input Z here.

So C20 sees 1500 ohms on one side (from U3) and 5K on the other
side (from R2). This works out to be a high-pass filter, letting thru
everything above about 350 Hz.
But C21 can be a LOT smaller because impedance levels approach
510K. C21(0.01uf) just happens to be a handy value that is used elsewhere.

>4) What is the purpose of C19? Is this a terminator of some kind?

Remember that U3 is a double-balanced mixer. It gives about
equal output at two frequencies:
L.O. + I.F. ( 4.0008 Mhz + 4.0Mhz = 8.0008 MHz)
L.O. - I.F. ( 4.0008 MHz - 4.0Mhz = 800 Hz)
We want the 800 Hz. stuff, but we DON'T want the 8MHz. stuff
since it could cause U4 trouble (wasn't made to handle R.F.).
So C19 "terminates" the R.F., leaving audio alone.

C19 and C20 and C21 were all ceramic capacitors in my kit.
These are prone to generating "microphonics". And since they're
ahead of a LOT of audio gain, they are the prime suspects for
replacement if you have microphonic trouble.
Check for microphonics: once audio is coming out the headphones,
tap the board and listen for "boink" aftereffects. If you hear any,
try tapping or scraping along these three caps to see if its worse.
If you can run a fingernail along the face of these caps, and hear it
in your headphones, then you might consider replacing them with
mylar, or polyester or any plastic film-type caps of the same value.
Should help reduce microphonics a lot.

>Just a question. Will Q1 turn on without C24 installed?
>Is C24 just to add a delay in turn on?

Yes, you've got it completely right. R8 will
drage Q1's gate up to +8v, which turns the FET into
a low-value resistor.

In a previous post, you mentioned about C24 "charging"
Its R8 that does the charging, via the +8v available
at U4a's output at pin 1.
Time constant is 0.1seconds (1MEGohm x 0.1uf)

When the FET acutally turns ON depends on its pinch-off
voltage (different for many kinds of FETs)...may not
be exactly 0.1seconds.

I think a neat experiment would be to short out R9
temporarily, to see how really LOUD the sidetone actually
is (blow your ears off).
And you'd hear the T/R thump too, although the
sidetone may cover it up.