Wow, these ARE detailed questions - someone is REALLY trying to
understand how this thing works....and deserves every bit of help
possible. While these questions may not be of interest to most, I'll
try to answer and expand in detail:
regards, Glen
>
>1. What is the significance of introducing the concepts of voltage
>source and current source in the analysis of this circuitry? I am more
>familiar with the concept of a voltage source...at least I think I know
>the "characteristics" of a voltage source.
Since Philips isn't very forthcoming about how the NE612 chip is
internally biased, I've had to make assumptions about DC biasing of the
chip. One assumption is that they use a DC current source to feed Q1 and
Q2 (I1 in the schematic). Current-source biasing is very common inside
integrated circuits since transistors can take less room and allow better
and more stable performance than resistors. You could replace I1 with a
resistor of about 660 ohms and get very similar results. This is a DC
source. V4 and V5 are both AC voltage sources.
>
>2. If V4 was based upon a 3MHZ voltage source, would the peak current
>through Q6 and minimal current flow through Q4 coincide with the timing
>of the peak current flow through Q2? I notice that the peak current
>flow through Q6 and minimal current flow through Q4 slowly change their
>relative positions over time with regard to the peak current flow of Q2.
>
Yes, with both V4 and V5 at the same frequency, peak current
would always be the same amplitude, and wouldn't shift with time. Let me
expand on your suggestion: what happens when you mix 3 MHz with 3 MHz?
You should get frequency components at sum & difference: at 0Hz and
6MHz. And indeed, if you make the one stipulation that V5 is 90 degrees
out of phase with V4, you'll see a 6 MHz sinewave coming out the top.
The peak/minimum current flow thru Q4 & Q6 changes with time
because the phase of 3MHz with-respect-to 4MHz changes with time. That's
mixing at work.

>3. "The mixer still performs its task if Q3, Q4, Q5 and Q6 don't act as
>switches. Output will be smaller, but all we really require of these
>four transistors is that they direct more current during their 125 ns
>"window" and less during the alternative 125 ns "window." (Quote from
>simulation) When I view the current traces for Q4 and Q6, is the
>minimum referred to in the quote from the analysis, when these traces
>"cross?" Likewise, the maximum referred to in the quote occurs when
>either Q4 or Q6 are at their respective peak values? I assume that it
>is no accident that the pattern of current flow for combined would
>reflect such a (summing strategy).
Perhaps its unfortunate that I've chosen signals levels part-way
between "switching" action and the more linear "multiplying" action.
Q4 and Q6 (Q3 and Q5 as well) "bottom out" at zero current on negative
peaks. With V4 smaller, you'd see Q4 and Q6 display more symmetrical
currents for +ve and -ve peaks - that's in the "multiplying" region.
As shown, they're into the "switching" region partly.
>
>4. The functioning of both Q4 and Q5 in their respective roles in the
>process is confusing when I consider the relationship of Q3/Q5 and Q4/Q6
>to each of "their" respective 3MHZ transistors. Even though Q2 is
>permitting maximum current flow through Q6, do I assume that at the same
>instant Q1 (minimal current--180 degrees out of phase with Q2) is
>permitting "some" current flow through Q4? (although the analysis
>reveals very small amounts of current flow in Q4).
Ok, you're getting into the heart of mixing action with this
question - you're close to seeing it all. But its hard to see the whole
thing working at once.
At the moment when Q6's current is maximum, you also see Q2 current
peaks as well. So all of Q2's current goes up thru Q6 rather than Q5.
Q5 is "off"
Over on the other side (at this same moment) Q1 is suppling a meager
current. Q3 is directing all of it up to R7, and Q4 is off.
It'd likely help you to see the Q3/Q5 currents laid over those Q4/Q6
currents. I had a look at this plot but it was too complex to follow,
so threw out the Q3/Q5 currents.
Keep in mind that I(q1) + I(q2) = 1.8ma DC (at every moment)
Also I(q3) + I(q4) + I(q5) + I(q6) = 1.8 ma DC as well...at every moment.
How these are distributed depends on the relative phase of the 3MHz and
4MHz signals. Its magical that you don't see any 3MHz or 4MHz stuff at
either output - the magic of double balancing.
>
>5. What is happening during the approximately .5 us, centered on 1.0 us
>with regard to current flow in Q4/Q6? It appears much different than
>the periods of time between 1.0 us.
There are moments (at 1us intervals) where everything is balanced.
Naturally, because the phase of 3MHz and 4MHz sources is constantly
changing these moments are fleeting.
But at the moment of 0us, 1.0us, 2.0us...etc, Iq3, Iq4, Iq5 and Iq6
are all equal, 1/4 of 1.8ma. each. And Q1/Q2 are balanced too: Iq1=Iq2=.9ma.
>
>6. Is the presence of 1 MHZ detected through the slowly changing
>amplitude of the current flows in of Q4/Q6? Likewise the slowly
>changing amplitudes (?) of the combined IC?
Well, I'd say that the combined (Iq4 + Iq6) current shows the
1MHz presence very clearly. A little harder to see in the separate
Iq4 or Iq6 waves, but its there.
>
>7. The schematic reveals the bandpass filter is connected to each of
>the outputs of the mixer. Maybe this is to come, but what are the
>advantages of using both of the outputs from this mixer? Are these
>advantages part of the rationale for the use of a doubly balanced mixer?
Pin 4 and pin 5 of the NE612 have very similar wavshapes. You
could use either one. But this chip suffers from having output signals
that are rather small - they need a lot of power amplification to get
up to 1.5 watts of 7 MHz out the antenna. Hey, its a "low-power mixer".
So we should squeeze as much output juice as we can.
Its easy to do: we simply take the output as a differential
signal, amplifying the difference in voltage between pin 5 and pin 4.
Ite fortunate that the 7MHz signal is out-of-phase between these two
pins. So we've doubled the output voltage - that's worthwhile.
That's the only advantage - more signal.

May I suggest that it might help to view the four transistors
q3, q4, q5, q6 as a bridge or ring. They're actually arranged in a
similar fashion to the diode-ring balanced mixers. But they're rarely
drawn that way.
The double-balancing trick is very similar in both circuits:
With the diode ring, out-of-phase input signals are made with the
multi-winding input transformer. In the gilbert-cell, its done with
the Q1/Q2 differential amplifier.
Then the mixing action is done by four devices in a ring:
be they 4 diodes or 4 transistors. Perhaps the diode ring collects
output signals in a more orderly manner (a single-ended output).
Only two of these ring devices are "on" (dominant) at any moment.

Great questions. Keep at it - its beautiful when you can see how the
whole thing works.


I guess it's time for my 2 cents worth. I have worked for a
leading manufacturer of mixer and have been exposed to there care and
feeding.

1. Double balanced diode mixers.

There are many falvors of these depending on the need and your bank
account. The difference in their performance is mosly a function of the
diodes used and the local oscillator drive power necessary to have them
function correctly. The most common variety uses +7 dbm ( that is 5
milliwatts) of local oscillator (LO) drive. The Miniciruits SBL-1 is
such a mixer.
The next LO drive level is usually +17 dbm (50 Milliwatts) or +10 dbm
(10 milliwatts).....these guys cost about 3 or 4 times the +7
units.....Next comes the +23 to +27 dbm (200 to 500 milliwatt....yes
that's a 1/2 watt) of local oscillator drive. These guys cost in the
ball park from about 50 to 100 bucks each in small quantities.

The difference in these mixers is their ability to handle signals
without generating spurious responses. A company called Watkins Johnson
makes many mixers and if you contact them they can provide excellent
articles on mixer selection,

For average ham use say in a QRP rig's receiver, the SBL-1 mixer
will provide very good performance.....If you spend about 10 or12 bucks
more the SBL-1H using a +17 dbm (50 milliwatt) local oscillator drive
mixer really work fine.

Time for a reality check....these mixers must be used properly for
you to get the stated performance.....that is...all 3 ports (RF, LO, and
IF) of the mixer want to see 50 ohm terminations.....The IF port
especially is the most important. The LO port is easy....if you have
some extra drive available padded it with a resistive pad and then into
the LO port. The RF port can be driven by a diplrxer which passes the
frequencies you want and terminates all others in 50 ohms. The IF port
can be terminated in a 50 ohm input impedance amplifier or a diplexer
like the RF port.....I prefer a diplexer most of the time because it
terminates the IF port in 50 ohms at all frequencies and dumpss the
image freqs into the 50 ohm load.

I realize that many QRP rigs are offen powered from batteries and
that receiver drain is a concern.....but if conditions are tough its
nice to have a good mixer in the receiver front end.



>Can anyone please explain the details of the oscillator circuit connected to
>Pins 6/7 of mixer U5?

>
Answer:
There's a transistor inside U5: base comes out at pin 6 and emitter
comes out at pin 7. All the biasing is inside the chip. So you've got a
Colpitts oscillator, similar to the VFO (Q2).
In case you don't see it, here are the analogous parts:
C5 is analogous to C29
C4 is analogous to C28
These are the two critical parts to this Colpitts configuration, providing
the positive-feedback path required by any oscillator.
There's no single part that is analogous to the crystal. It is the
resonating element, containing both inductive equivalent and capacitive
equivalent of a resonant circuit.
But what may be confusing is that 22uH choke. Let's take a closer
look at its purpose. The short explanation is that its there to pull the
crystal to a lower resonating frequency.

A little background: we need about an 800Hz offset between RX and
TX. That's the "sweet spot" CW note that most folks like to hear. This offset
is done by having U3's 4MHz. BFO crystal run at a frequency that is 800 Hz.
higher than U5's 4MHz. crystal.
The 22uH choke pulls Y5's resonating frequency lower by adding
some inductive reactance to the crystal's own. Now 22uH is a lotta inductance.
If it only pulls the crystal by 800 Hz. then Y5's own internal inductance
must be absolutely HUGE. And it is. Inductive reactance inside the crystal
is a fraction of a HENRY. That's astronomically big compared to Q2's resonating
inductance of 2.5 MICROhenry. So you can see that pulling a crystal's frequency
is really difficult.
Now C28 and C29 also affect U5's oscillating frequency. You'll notice
that C29(150pf) is much larger than U3's equivalent: C18(47pf). This was done
to keep U5's oscillating amplitude low, in an effort to reduce U5's
spurious frequency output at pins 4, 5. Having C29 large also helps lower
the resonating frequency of Y5. But not enough. Adding 22uH to the crystal
pulls Y5 down some more.

If you've got a working rig, with working RX and TX, you can try
this experiment to explore the resonating frequency of U5's 4MHz. oscillator.
Since some may be following the construction sequence on QRP-L (and not
have U3 wired yet) I'll come back to this experiment later on....

Short out the 22uH choke (RFC2). This will raise the oscillating
frequency of U5, much closer to the oscillating frequency of U3. You should
either hear no sidetone, or a very low-frequency sidetone. This experiment
demonstrates that the rig's TX/RX offset is set by the frequency difference
between U3 and U5 oscillators.
Glen VE3DNL leinwebe@mcmaster.ca


One small detail on 602 type mixers... The biasing internal to the chip is
sufficient for most applications. Occasionally you may find a resistor
from the output (emitter) pin to ground.

This is done to increase the bias current by lowering the effective emitter
resistance by parralleling the internal resistance with the external
resistor. (R.effective = R.internal || R.external) A second effect is to
extend the oscillator range or increase oscillator output, so you may also
see this technique used when someone is using the mixer chip close to the
specified freq. limits or in a tx mixer application.