Discussion highlights from the Podcast ...

·         Designed by Wes Hayward W7ZOI

·         Inspired by legendary HP8640 signal generator

·         Circuit Functions

◦     Overall – Tunable reasonable quality test oscillator

▪     Tunes 3.4 – 30 MHz in three bands

·         Main tuning and bandspread

▪     Good buffering and filtering for stability and signal purity

◦     Hartley oscillator with air-variable tuning capacitors

▪     Range 13.6-30 MHz sine wave output

▪     Divide by 2 and low pass filter for 6.8-15 MHz

▪     Divide by 4 and low pass filter for 3.4-7.5 MHz

◦     Voltage regulator aids frequency stability

◦     Buffer amps help maintain stability and low harmonics

·         Voltage regulator

◦     Integrated circuit 7806 gives good regulation

◦     SEPARATE VOLTAGE REGULATOR FOR OSCILLATOR TO KEEP OUT DIGITAL NOISE FROM DIGITAL CIRCUITS

◦     6v REGULATOR VS 5V TO GET MOST PERFORMANCE FROM Q1 AND Q2

◦     Ceramic bypass caps in and out for filtering and eliminate instability

·         Variable oscillator

◦     Hartley circuit for wide tuning range

◦     Air variable tuning caps for good stability

▪     365 PF TC1 (TUNING CAPACITOR 1) IS MAIN TUNING

▪     35 PF TC2 IS FINE TUNING

◦     JFET for minimal circuit loading

▪     2N4416 IS JUNCTION FET

▪     HAS HIGH GAIN AND LOW STRAY C UP TO VHF

▪     OTHER FETS – 2N3819, 2N5485, MPF102 WILL WORK BUT HAVE WIDER SPREAD OF CHARACTERISTICS SO MAY NEED SELECTION

◦     Diode on FET gate provides amplitude self-adjustment

▪     ALONG WITH C8 AND R4 RECTIFY SMALL AMOUNT OF OSCILLATOR SIGNAL AND APPLY NEGATIVE FEEDBACK AS OSCILLATOR OUTPUT TRIES TO CHANGE

◦     Iron core toroid for tuning inductor for good stability and low loss

◦     NPO caps in tuned circuit for stability and low loss

▪     C5, C6 C7 AND C8

◦     100 ohm resistor in drain circuit to prevent VHF instability

▪     SINCE JFET HAS GAIN UP TO VHF EVEN WIRING CAN ACT AS TUNED CIRCUIT.  10 OHM RESISTOR R5 “DE-Q'S” THE STRAY INDUCTANCE

·         Two-stage follower/buffer amp isolates oscillator

◦     JFET Q2 follower for low oscillator circuit loading

▪     ANOTHER JFET WITH HIGH INPUT IMPEDANCE

▪     DRAIN RESISTOR R6 DE-Q'S STRAY INDUCTANCE TO PREVENT VHF OSCILLATION

◦     Bipolar transistor Q3 provides high RF gain 2N3904 USED BUT 2N4401, 2N2222 ALSO GOOD CHOICES

▪     CONNECTED IN COMMON GATE AMPLIFIER CONFIGURATION WHICH HAS VERY LOW OUTPUT TO INPUT STRAY C FOR GOOD ISOLATION

▪     DRAIN RESISTOR R12 PREVENTS VHF OSCILLATION

◦     XXXXXXX OUT drain and resistor resistors minimize instability

◦     independent RC filters in DC SUPPLY TO each stage for clean signals

◦     Output buffer uses wideband xfmr on ferrite toroid core

▪     bifilar winding for ease of construction

▪     2:1 IMPEDANCE RATION ASSURES THAT Q3 COLLECTOR SEES HIGH IMPEDANCE FOR BEST GAIN

·         Highest 2:1 freq range feeds sine wave directly to output buffer. (13.6 TO  31 MHZ)

·         Other ranges gotten from digital divider chain

◦     Additional RC filtering and 7805 regulator in digital DC power assure clean signals

◦     Second tuning range from squaring amp and divide by 2 flip flop

◦     SQUARING AMPLIFIER IS BIPOLAR 2N3904 BIASED AS LINEAR AMP BUT OVERDRIVEN BY SINE WAVE INPUT TO PRODUCE SQUARE WAVE OUTPUT

◦     DIGITAL DIVIDER 74HC74 TYPE DEVICE WORKS WELL UP TO 30 MHZ AND DRAWS LESS POWER THAN OTHER 74XX74 TTL TYPES

◦     FIRST DIVIDE BY 2 CIRCUIT PROVIDES SQUARE WAVE OUTPUT OF APPROX 6.8 TO 15 MHZ

◦     SQUARE WAVE OUTPUT IS CLEANED UP TO SINE WAVE BY LOW PASS FILTER

▪     SQUARE WAVE CONSISTS MAINLY OF FUNDAMENTAL FREQUENCY AND ODD HARMONICS (3, 5, 5 ETC.)

◦     Third tuning range additional divide by 2 flip flop (total of divide by 4) FOR 3.4 TO 7 MHZ TUNING RANGE

▪     Square wave output OF DIVIDERS cleaned up by low ripple steep cutoff LPF

·         .07 Db CHEBYCHEV DESIGN

◦     Low ripple to keep output constant across frequency

◦     steep rolloff removes 3^rd and higher harmonics of square wave

◦     Low pass filters use iron core toroid inductors and NPO capacitors for low loss

▪     MAY NEED TO CHECK INDUCTANCE AND CAPACITANCE OF EACH COMPONENT AND TWEAK INDUCTORS TO GET BEST AMPLITUDE STABILITY ACROSS EACH BAND AND MINIMAL  HARMONIC CONTENT

◦     Low pass use resistive input attenuation to preserve passband response

·         output buffer uses resistive feedback for flat frequency response

·         Also bifilar wound broadband transformer on ferrite toroid

◦     2:1 TURNS RATIO (4:1 IMPEDANCE RATIO) KEEPS COLLECTOR CIRCUIT IMPEDANCE HIGH FOR BEST GAIN

◦     Once again RC decoupling in DC power feed

·         Resistive attenuator on output for isolation and to minimize SWR mismatch.

·         OVERALL OUTPUT LEVEL IS +7 DBM RMS OR 0.5v INTO 50 OHM LOAD

Wes Hayward, w7zoi, 26feb12, update 28/29Mar12, 9apr12.   (Click to download a PDF the article.)

Radio Frequency signal sources are like power supplies: we can always use one more of them. This signal source is based upon the Hewlett-Packard HP-8640B, which is one of the most popular generators available. The generator is no longer available new, having been replaced by more up to date synthesized instruments. But the 8640 is still available and popular on the surplus market. The HP-8640 uses a single variable oscillator operating at VHF. That source is then frequency divided to generate the various bands. The phase noise and stability get better as the division ratio increases. This divider chain basis is used in our design, although with significant simplification.

This signal source consists of a Hartley oscillator operating from 13.6 to 32 MHz. It is necessary to have a tuning range of at least one octave to really take full advantage of this divider topology. This range is available directly, or is divided by 2 or 4, providing output all the way down to 3.4 MHz. Each of the two divided ranges is filtered with a 5th order low pass filter. Only one such filter was used for each band. Each was designed for a cutoff frequency just above the top of the range. The passband ripple was varied to achieve convenient, off the shelf capacitor values in the filters. The harmonic suppression is reasonable, but is much worse than the instrument's namesake. The original HP-8640B has a stellar harmonic suppression of over 60 dB, which resulted from the use of 3 filters for each octave. The excellent suppression is atypically good for a signal generator. Our 8640-Jr signal source is shown in the schematic of Fig 1.

Note that we don't call this a signal generator. A signal generator is a well shielded instrument that can be used to measure the sensitivity (MDS) of a serious detectors of only modest sensitivity such as oscilloscopes. Not only is shielding good, but a signal generator is immune to interactions from other generators that might be attached. A modern signal generator has a well defined, low output impedance, usually 50 Ohms. Frequency stability is, of course, good. Our little offering does not fit all of these criterion. It is, however, reasonably stable. Moreover, the oscillator is well isolated from external influences, allowing it to be used with another similar unit for the evaluation of IMD of mixers or amplifier in the lab.

Photos of the rf source are shown below.

Examination of Fig 1 and of Fig 7.27 of EMRFD will show considerable similarity. In spite of this, this experiment provided a few interesting details which we will outline here.

The signal is extracted from the VFO with a source follower while it was pulled from the EMRFD signal source with a single turn link through the tank coil. The link is the preferred method, for the harmonic output is lower by over 20 dB. While the output amplifier contributes some of the harmonic distortion, the dominant contributor is the follower.

The secret of this design, although it is certainly not a secret, is the common base buffer. This was originally inserted in the EMRFD RF source when we tried to use the instrument to examine a crystal filter. The VFO would try to lock to the crystals when the output was reflected from the crystal filter and back through the output stage to the VFO. The extra buffer with it's excellent reverse isolation completely eliminated this difficulty. The rf source had not been usable for IMD measurements until the common base buffer was added.

The VFO in the '-Jr originally used a 2.2 pF gate capacitor. The particular JFET we were using and perhaps the gate diode combine to compromise the starting gain for the oscillator. The circuit would change output level when at the lower end of the tuning range. The changes were often in the form of sharp steps in output level as frequency was changed. Changing the gate capacitor from 2.2 to 4.3 pF completely eliminated this problem.

Note the method of extracting signals from the divider chips. In each case, a large capacitor provides DC blocking so the following resistors do not alter the DC conditions on the CMOS chip. A voltage divider then drops the level to that needed to drive the output amplifier. The 680 Ohm resistor provides a high Z load that the divider can drive while the 51 Ohm resistor provides the proper resistance to terminate the low pass filter. When first built, the 8640-Jr used a classic surplus variable capacitor from a WW-II BC434 Command Receiver. These feature a built in gear reduction drive (about 60:1) and a dial mechanism, all in a package that is surprisingly easy to mount and use. But there seemed to be problems with excess noise as the circuit was tuned. The capacitor was changed to another from the junk box, but the same problems were there. Eventually, they were traced to the inadequate starting gain mentioned above. But this had not been confirmed until the capacitor had been swapped out for the 365 pF AM broadcast band capacitor. The combination of two capacitors works well, although the band spread is extreme at the low end of the tuning range while almost inadequate at the high end.

It was recently (28Mar12) brought to my attention that an enterprising purveyor of radio kits has decided to assemble a sack of parts for this design and offer it to the amateur radio community. He says that he will use polyvaricon variable capacitors instead of the air variables that I used. This could lead to problems. See my Q measurement results on these parts at http://w7zoi.net/2faces/twofaces.html.

This design uses an output amplifier biased to about 20 mA emitter current. The similar stage in the EMRFD circuit, Fig. ,7.27 had a current of 35 mA. This means that the output stage of the EMRFD circuit could be driven harder. This means that either the output could be higher, or that a larger output pad could be employed. Both would be an advantage.

There is no adjustment of output level. Some sort of variable attenuator would be handy such as the pot adjusted circuit of Fig. 7.22. Alternatively, a PIN diode attenuator would do the job.

Spectral analysis of the output is interesting. The harmonic attenuation is poor as mentioned above and depends upon the band selected as well as the position within the band. The high band, even when neither digitally derived output is in use, still has a spur at half the output frequency, but it is at about -55 dBc. This is some of the digitally derived lower frequency leaking through. Experimentation and design refinement would probably improve this problem.

The schematic diagram is labeled with an output power of +7 dBm. This is approximate. The following data was measured, showing a slight variation from the nominal value:

       High band             +9.2 dBm at 13.6 MHz             +7.6 dBm at 32.6 MHz

       Mid band              +4.8 dBm at 16.3 MHz             +8.0 dBm at 6.8 MHz

       Low band             +6.8 dBm at 8.17 MHz             +8.2 dBm at 3.4 MHz

The low pass filters are purposefully designed for a cutoff that is close to the top of the respective bands. Hence, it is important to actually measure the inductance can

capacitance value, or otherwise confirm the proper operation of these filters. If the filters are suspect, they can by bypassed during construction and debugging.

Back to CWTD Home