Application Hints for the Hybrid Cascode IF Amplifier

18Nov07, update 21Nov, 2Dec, 6Dec07, 2Jan08.    

Modifications  (Some of these things are now included in the errata.)

1.  We have added two outputs to the board layout that are not detailed in the QST article.    The first output is a DC signal that is merely the AGC voltage.   A resistor (R37, value = 10K) is added to the board just to protect the AGC line from any loading that might appear on the line.   The intended purpose of this voltage is to drive an S-Meter.    A driver circuit is presented below.   The second added output is the 9 volt output from the regulator on the IF board.    This voltage is used as a reference for the S-Meter driver.   It also may be handy as a supply for an oscillator elsewhere in the design that might use the IF system.
2.   The input inductor, L1, is dropped from 7.2 to 6.8 uH.   A toroid can be wound, or an RF Choke can be dropped into this slot.   This L should have a Q of 40 or higher for best match and noise figure.
3.    The value of resistor R7 is decreased from 3.3K to 3.0K.   This is the impedance that fits with the 6.8 uH at 9 MHz.
4.    R15 changes from 3.3K to 3.0K.    If you want slightly higher gain, you can use 3.3K.
5.    Inductors L2, L3, and L4 are replaced with 47 uH RF Chokes.    This keeps the costs down and eliminates the need to wind some toroids.   There is one left, and the builder must wind this.   No, Roger won't do it for you!
6.    Add an 18 Ohm resistor, R38, in series with C12, which connects the emitters of Q7 and Q8.   For details, click here.
7.    External threshold control:     Jeff, WA7MLH, liked the behavior of the system when he adjusted R6 on the test bench, so he ended up with a front panel pot for the control.   This variation might be one to consider in some experimental receivers.


An S-Meter Driver for the Hybrid Cascode Hybrid IF Amplifier

Many receiver builders planning on using the Hybrid Cascode IF amplifier will want to incorporate an S-Meter in their design.   We have designed a circuit that should provide a good start for the design.  

Two curves are presented here for the closed loop AGC system response.     This is a curve of detected DC appearing on the "AGC line" (emitter of Q9) versus the signals applied to the input.   This particular curve was generated with a ferrite input transformer driving Q2.   (Note:  The curve shown was taken with an earlier version of the IF amplifier with slightly different biasing voltages for the manual gain control, so the voltages are different.   The maximum  AGC voltage is 5.5 in the curve.   The max is about 7.3 volts in the QST design. )

The red curve represents a relatively high AGC threshold.    This is the "ear-saver" kind of characteristic that I personally prefer for casual operation.  The receiver sounds as if there is no AGC in the system at all until the input to the AGC amplifier is -95 dBm.   The AGC loop then begins to reduce the IF gain as the signals become stronger.  R6 is short circuited for the red curve.    The blue curve is the one obtained with R6 (threshold) at 270 Ohms.     This is more in line with the kind of AGC that we find in many commercial radios where the noise within the radio is already enough to turn the AGC on.    The radio output never gets too loud and all signals sound the same.      It also yields an S-meter response that begins at low levels.    The builder/experimenter needs to make a choice and set R6 accordingly.  Some experimenters are using a pot for R6.

Looking at the curves, we see that a "no signal" input to an S-Meter driver circuit may be as high as 5.5 volts.   It will drop toward 2 volts as signals become very large.     This gives us something to use for design.  

A suitable circuit is shown below.   This is an updated version (6 Dec 07) of the original, designed for ease of adjustment and compatibility with about any version of the AGC that a builder might generate.  
 
Three connections are required to the outside world (plus ground.)    The first is  the 12 volt supply that is usually in the receiver.   This can go much higher.   The IF amplifier itself assumes that the supply never drops below 11 volts.    If a lower voltage is expected, redesign is in order, perhaps with a low dropout regulator.  We used a modest LM317L in the IF.    The second connection is to the 9 volt signal that is generated within the IF.    We used this as a reference in the driver circuit, but do not ask much more than a couple of mA of current from it.    The third input needed is, of course, the AGC signal itself.   This is the voltage predicted by the curve above.   The signal marked "Reference Voltage" should equal the voltage on the AGC bus with zero signal into the amplifier and with the manual gain control at maximum.   This is a little over 7 volts for the amplifiers we have built in the last couple of months.

U1 is nothing more than a high input impedance buffer.   It has a diode of downward offset in the output.    This merely compensates for the offset in the diode that follows the op-amp.    The circuit that follows is a diode network that allows the meter to attack quickly, but be damped so far as the recovery is concerned.     I believe that this general scheme can be attributed to Bill Sabin.  The various resistors are marked with regard to function.    The voltage on the damping capacitor drives U2, another non inverting buffer which is the meter driver.  

Op-amp U3 is a "perfect" rectifier.    It provides a stable "floating ground" to deflect the meter against, but it does not allow any deflection in the wrong direction.    The overall system is configured to drive a 1 mA meter movement, which is very common.    If a more sensitive meter is used, the meter deflection resistor (5.1K) can be increased accordingly.   For example, if a 100 uA movement is used, the resistor would become 51K.    If a less sensitive meter is to be used, emitter followers may be required in the output of U2 and U3.   They should be within the feedback loops.    The S-meter is adjusted to zero with the pot when the antenna is disconnected from the receiver.

I have NOT built this circuit.   However, it should provide a reasonable start for developing a circuit for a specific application.  

Feedback:   The first version of this circuit was posted on November 18th.   I got a note on Nov 21 from K4ZKU and Jeff said that he had breadboarded it and had it working with an earlier IF amplifier (probably from the Progressive Receiver) and that it worked fine.   Many thanks  Jeff.    The newer version shown above is actually a simpler and less interactive design.   The version shown above was first built by John, K5IRK.   Jeff, WA7MLH also has a version running now.   No problems reported with either.  Many thanks to both.

Shielding
Jeff Damm, WA7MLH, had some interesting observations after he dropped one of these IF amplifiers into one of his homebrew SSB transceivers.     He had built the S-Meter driver earlier and had it operating with the IF amplifier.   But he noted that the S-Meter was partially on when the board was in the transceiver.    He had no shielding on the IF, and some of the other modules were also open.   Jeff discovered that the problem was direct signal pick-up from part of his LO system.   As soon as the board was removed, but still electrically attached to the transceiver, the problem disappeared.      A shield is now planned for the transceiver.

Input Crystal Filter Considerations

The input to the hybrid cascode amplifier is shown below.
With no gain reduction, the current in the two transistors is about 3 mA with the biasing used in the QST design.   The input impedance seen looking into Q2 is very high, but with some shunt capacitance.    The net input impedance when looking into R7 is just the value of R7 paralleled with a few pF.   R7 was 3.3K in our design.    

The most convenient topology for many applications is one with a 50 Ohm input impedance.    (It is the most easily measured, which is a useful criterion in a homebrew receiver.)     C3 is merely a blocking capacitor and can be ignored.   The 3.3K is transformed down to 50 Ohms with a simple L-network.   The design formula for this is in virtually every Handbook or RF design manual that has appeared for nearly a century.    We did not merely present a reactance with the components, for the builder may wish to change the value of R7 to decrease or increase the overall gain.

Many classic IF systems drive a high impedance (a FET or IC) with the output of a crystal filter.    The filter must be properly terminated to achieve the desired design shape.     A common termination is 500 Ohms.     It is common to replace R7 above with a 510 Ohm resistor, dispense with L1 and C2, and drive C3 directly with the crystal filter.     This works fine.   However, you loose some gain with this procedure.   The higher termination resistance leads to higher gain.  One could design an L-network to transform from 500 Ohms up to 3300 Ohms.   This is shown below.
  The builder can design the network to suit his or her application.

Another alternative would be to use an L-network (or some other transforming circuit) to drop from 500 Ohms to 50 Ohms and to then go into a shielded IF amplifier box with coaxial cable to a higher impedance such as our 3.3K.   This is presented below.
   This is probably the scheme I would use when designing a new receiver or transceiver.    The Hybrid Cascode circuit was configured to allow experimental flexibility.

Output Level Programming, including a Noise Filter Interface

The output network is also configured for flexibility.    The present circuit is shown here:
  R3 is normally set to 51 Ohms while R4 is a short circuit.   C14 is again just a blocking capacitor and has no impact on performance.    This configuration then presents a 50 Ohm output impedance, for the impedance looking back into the Q8 collector is very high.    The 50 Ohm load, which might be a diode ring product detector, is then properly driven.    The impedance is proper on a wideband basis and the level is low enough for low distortion.    

The output from Q8 above is labeled with a "1 mA signal" current.   This is NOT the level present.   Rather, it is a hypothetical level we picked for some later calculations.    Let's examine this case with 1 mA.   This signal is presented to a net impedance of 25 Ohms, which is the parallel combination of the 50 Ohm load and R3.   With a 1 mA signal, Ohm's Law tells us that the signal voltage at the load is 25 mV.   This appears across the 50 Ohm load, so the power delivered to the load is 12.5 microwatts, or .0125 milliwatt, which converts to -19 dBm.

One might wish to drive other product detectors.    One could probably get reasonable performance with a NE602 or similar low current Gilbert cell by terminating the detector input with a 51 Ohm resistor.   On the other hand, if a transformer was used to drive the input of that cell, the drive might be excessive, leading to distortion.     I observed this with a transceiver built many years ago and only fixed it when a 20 dB pad was inserted ahead of the detector.   It all depends on the exact signal voltages.    

The Hybrid cascode was configured to allow the user to drop the levels.    Here's an example of such an application.
  We have decreased the resistor R3, which forces the signal voltage to a lower level.  R4 was changed to 100 Ohms.      The net load at the collector is now 150 Ohms in parallel with R3, which is again 25 Ohms.    The 1 mA signal will then generate the same 25 mV at the collector.     This will cause a signal current to flow into the 150 Ohm arm of 0.167 mA.    The voltage across the 50 Ohm load is then 8.333 mV.    The corresponding power is 1.39 microwatt, or .00139 mW.     This corresponds to a level of -28.6 dBm.    The circuit change has attenuated the output by 9.6 dB.    The output is no longer matched.     It would be possible to retain a matched output if the sum of R3 and R4 was 50 Ohms.

In some cases it is desired to place a crystal noise filter at the output of the IF amplifier, prior to the product detector.    If we were again using a 500 Ohm filter, the noise filter could be properly matched if R3 was changed to 510 Ohms.    This is shown below.
   This circuit uses a higher termination at the collector of Q8.  Hence, the gain will be higher by 10 dB.   We  will leave it to the reader to demonstrate this.