Building my first oscillator

[W4AMV]

Just to be sure, upload your current schematic and where you are breaking the connection between the buffer and the 3866. Your buffer output alone looks quite good. So it would appear that it is incapable of properly driving the 3866. Either it cannot source or sink the required base current. As an experiment, you should investigate what impedance your buffer can drive with little or no distortion. AC couple via a 0.1 uF block and then decrease the load R from say 10 K downward.

To your questions,  Use you scope, I mean with the scope set for DC coupling observe the DC+AC signal and see if the voltage swings are such that the 3866 would create distortion with its current DC bias settings. Adjustments could be made first and easy, by placing the 3866 on its own supply voltage and adjust. If you have enough supplies you could place one on the collector and one on the base bias network.

On the second, to much drive from the BUFFER. And that may be a good piece of the issue. Try coupling the buffer to the 3866 base via a small C, say 500 - 1k ohm of series reactance at 3 MHz. 1 k is 53 pF, try a 47 pF series C.

On the third, YES. But for now that IS NOT the issue.

[W4AMV]

incidentally, I might add, when we say lightly couple with the series C... If you look at the pdf I sent, you will see that as we make that series C small, the equivalent parallel R transformed from the series R INCREASEs. So you see, the load on the buffer can be "lightened" by either an increase in the series load R, or by a decrease in the series cap connecting that series R (the 3866 base) to that of the source of the buffer. 

[ssbothwell SWL]

i tried swapping out different load resistors onto the buffer output and noticed that around 1kohms the waveform started to get deformed. below 1000ohms the voltage dropped significantly the deformation gets pretty bad.

heres some images to show you what i mean:
10k: http://i.imgur.com/jyqwm.jpg
5k: http://i.imgur.com/ZONNa.jpg
1k: http://i.imgur.com/CrHgU.jpg
50: http://i.imgur.com/KWtxm.jpg


i got this brilliant idea to replace the 2n3866 voltage divider circuit with 15k pots. tweaking things around quite a bit i was able to get this one sweet spot where the waveform looked nice: http://i.imgur.com/n3yHV.jpg

to get a nice sinewave output i had it set like this:
base-ground : 330ohm
base-V+ : 467ohm
emitter-ground : 15ohm

here is the current schematic: http://i.imgur.com/9J430.png

with these voltage divider values i actually get the same result in ltspice as i do on the circuit board. well except that the spice model reports 5mA base current but in real life i am getting about 1mA

[W4AMV]

The MPF102 is not a great FET. That is to say, its gm is quote low and therefore the output Z of the FET buffer is not very low. The output Z of a source follower is ~ 1/gm and so the outcome of your tests with the R loads makes sense. The Zout certainly quite high, say 1500 ohms or so. Very well. Now, your 3866 output eLL network is OFF target. You are essentially loading the collector with 50 ohms! A better value with 12 volts collector would be in the neighborhood of 200-500 ohms. Go design an eLL match now to transform 50 ohms to say 350 ohms at 3 MHz for your 3866 output network.  See how that plays.

[WU2D]

I think that you are loading the FET a bit too much and this may be spoiling the waveform. Try adding 5 more parts; a simple emitter follower between the source and the base of the amplifier. Mike WU2D

[ssbothwell SWL]

oh boy the forum is back up.

i tried to calculate the values for an l-network using the formulas W4AMV supplied combined with some demonstrations from section 20-10 of the arrl handbook.

using a 14v power supply, Rs=50 and Rl=350 i got:

Q = 2.44 = sqrroot(350/50 - 1)
Xl = 122r = 50 * 2.44
Xc = 143r = 350/2.44
L = 6.47uH = 122/2pi3
C = 370pF = 1/2pifXc

i made the inductor using 5 turns on an ft37-43 toroid. this gave me 8.75uH but its class as i could get without going under according to this calculator: http://toroids.info/FT37-43.php

i had to use a combination of silver mica and ceramic capacitors to get to 370pF.

heres what the waveform looks like with this l-network:
http://i.imgur.com/AL9YA.jpg

it looks a lot better but it still doesnt look right. its more triangular now.

did i do the calculations right for this l-network?

i have a bunch of 22222 transistors so i can try adding an emitter follower between the 2n3866 and the buffer stage.

[W4AMV]

Hi. First your scope shows a reasonable Pout, 1.27 V rms or 15 dBm. More than ok for your mixer.

Next, the series reactance to transform 50 to 350 ohms, a reasonable choice, is 122 ohms. You have that and that would be accomplished with a series C of 435 pF at 3 MHz. Lets assume the transform Q is large enough (i.e. GT 1), and so L can be obtained directly from the 122 ohm value, L at 3 MHz, is 6.47 uH.

So you are close. Check your math. I choose 3 MHz for Fo. You may be closer at 2.5 MHz. In any case, the SERIES C is 435 pF and the SHUNT L to the 3866 collector is 6.5 uH. If you build a low pass filter next, YOU ARE DONE. The current waveform is JUST FINE. However, if you are a purist, build a 3 section Butterworth filter with a cutoff frequency of say 4 MHz, you will get 18 dB/oct attenuation of the harmonics, it will look super.

Of-course, without going into the details, I hate to break the news, but a better waveform to drive the mixer is a SQUARE wave! Not to worry, what you have is just peachy. I would not waste time with another bipolar stage, but simply reduce the series C between the buffer source node and the base node of the 3866. Again, the real issue is the gm of the MPF102 is not very high, yes a bipolar would be a better follower and in combination with the FET (a BIFET darlington) wonderful. However, not worth the effort.

[ssbothwell SWL]

very exciting! i have built low pass filters in the past and i have all the parts i need to make a low pass for the 80m.

currently the oscillator range is 2.5-3mhz. would it be possible to increase the range to 500khz - 4mhz? i would like to be able to listen to broadcast  band radio as well as 160m and 80m bands if possible.

i know that adjusting the range is a matter of geting the right size coil and balance of fixed and variable capacitance in series and parallel. however so far my experience has been either or a short range of frequencies with fine grained tuning or really wide range of frequencies with really course tuning. are there any tricks using multiple tuning capacitors to get course and fine grained tuning over a wide range of frequnencies?

edit: also i have a question regarding the stability of the oscillator. the reading on my oscilloscope jumps around a little bit. right now i see it fluctuating between 3.049 and 3.067MHz. i'm not sure if this fluctuation is caused by the oscillator being unstable or by the scope not being able to detect the frequencies very precisely. how can i check this? would a dedicated frequency counter be more accurate than using my oscilloscope?

[W4AMV]

ok, 2.5 to 3 MHz, is a nice tuning range. More is possible, much-much more is also possible. However, as a general rule, as you increase the tune range, achieving frequency stability become more difficult. There are solutions, but that is a new thread! I would encourage you to finish this oscillator and you need to check its stability with a frequency counter. Ideally one with a stable time base, say 5 ppm would be nice. Your goal should be drift less than 500 Hz upon turn on and say after 10 minutes it would be nice to see less than 50 Hz. Trust me, that is not easy. You are going to have to package this unit with care. Isolate it from fluctuations in temp and mechanical vibration. At this point, get a freq counter, or if you have a short wave RX, tune it in on you Rx, see if it stays put. What FUN. Place it on your desk, lightly rap the desk with a pencil, can you hear the tapping in the Rx?!

[ssbothwell SWL]

i'm super excited to get this receiver operational. i adjusted the tank circuit to drop the range down to 1.3-1.8MHz so that i can use it for broadcast band listening and i recalculated the l-network for the lower frequency band. i'm gonna turn this into a sweet little am radio.

can i use any old diodes for the mixer or must they be schottky diodes? what specs should i look for in the diodes?

does it matter that the amplfication is not linear across the frequency settings? i get 1.5VRMS at 1.3MHz and 1VRMS at 1.8MHz

[W4AMV]

i'm super excited to get this receiver operational. i adjusted the tank circuit to drop the range down to 1.3-1.8MHz so that i can use it for broadcast band listening and i recalculated the l-network for the lower frequency band. i'm gonna turn this into a sweet little am radio.

can i use any old diodes for the mixer or must they be schottky diodes? what specs should i look for in the diodes?

does it matter that the amplfication is not linear across the frequency settings? i get 1.5VRMS at 1.3MHz and 1VRMS at 1.8MHz

The diode barrier voltage or knee voltage as low as possible. So, for the Schottky 0.3 V vs. Silicon at 0.7. You can use Silicon, say 1n914 would be a nice choice and add adjustable bias to the ring diode to slightly overcome the barrier or turn on voltage. Then your LO voltage swing will do the rest. Try to match the diodes for forward currrent vs applied diode voltage. Not essential but nice. This type of mixer operation is so-called starved LO. If you need a circuit I can pencil sketch. However, it is not unlike the one you sent several posts prior, but you need to introduce an adjustable DC bias through the diode ring. NOTE. You could do FET or BIPOLAR active mixers as well. However, the ring diode balanced mixer is a classic. You should see conversion gain of -6 to -10 dB. Yes, this mixer is lossy.

[ssbothwell SWL]

would i need a bias on the diodes if i am using schottkys or is that just for silicon diodes?

the only diodes i have currently are 1n270, 1n4001, and 1n4004. i might as well go with schottky diodes if i have to go buy new ones anyways.

[W4AMV]

The diodes you mention are a bit to large for this application (to much capacitance), however at these low frequencies they may work just fine. You can try them out first in LTSpice as you should build the mixer in Spice to see its function. You can remove the bias stuff I added in the view 1 of this schematic. I ran the mixer and in view 2 you see the time waveforms. The inputs are 1.8 MHz, LO, 1 MHz RF, 800 kHz is the IF. The IF port signal shows the sum, the difference, and the and distortion all removed via a LPF (not added here). You can take a FFT snap shot of the IF signal see panel 3. The 800 kHz is the most prominent. Note how the xmfrs are built in Spice. I used 1n914 diodes.

[ssbothwell SWL]

i ended up finding some really cheap schottky diodes at the electronics shop: http://www.fairchildsemi.com/ds/1N/1N5818.pdf
mine are the 1n5817 version.

i threw together a diode mixer on my breadboard. my triflar transformers have 11 windings and i used 104 capacitors in place of .4uF ones.

with an input into the mixer stage of 1.62V RMS at 1.6MHz i get 414mV RMS 1.36MHz at the IF output. is that the amount of voltage drop i should expect from the mixer stage?

is there anything i can do to make sure the mixer stage is working properly?

is the next step to add another amplifier and then an envelope detector?


edit:

1.62VRMS = 52.4mW = -12.8dB
.414VRMS = 3.4mW = -24.7dB

that makes it a -12dB gain from the mixer right? is that a little much attenuation? you said it would be around -6 to -10dB.

[W4AMV]

hi, I am  a little confused. There should be 3 signals involved, the LO, RF, and IF. The conversion loss is the ratio of the RF level to the IF level. Can you clarify? Is the 1.62 V the RF signal?

[ssbothwell SWL]

i was under the impression that the RF input is just via a wire antenna? is that not correct?

when testing the circuit should i use a rf function generator into the RF Input?

[W4AMV]

yes.

[ssbothwell SWL]

ahh okay. so i'm not entirely sure what to be looking for here but here is a bunch of numbers:

LO Frequency: 1.5Mhz
VRMS at IF output with no RF Input: 340mV RMS

RF Input (function generator) Frequency: 2MHz
RF Input Voltage: 1VRMS

when i connect the RF Input to the mixer circuit the IF output voltage becomes 1VRMS and the wave looks like this:http://i.imgur.com/SO0v0.jpg

is my RF input voltage way too high? what should the if output from the mixer look like?

[W4AMV]

Yes, back off the RF input and watch the IF port output. With no RF input, at the IF you are seeing the LO input blow by. The mixer is not perfectly balanced. If it were that LO signal would be null. Now a LPF would be handy, even an RC network might be fine. You need to suppress as much of the LO bleeding to the IF so you can see the IF with some clarity on your O'scope. The ratio of the IF level to the RF level will be the conversion gain (loss).

[ssbothwell SWL]

ahh okay. i just re-read arrl handbook section on DBMs. At this point my mixer is not balanced.

the arrl handbook says that "diode and transformer uniformity results in equal LO potentials at the center taps of T1 and T2. The LO potential at T1's secondary center tap is zero (ground); therefore the LO potential at the IF port is zero."

the handbook doesn't offer solutions to correct the balance.

are you saying that i can use a low pass filter somewhere to correct the balance? where is it supposed to be used? i'm not sure i understand what you are saying to do with the LPF.

[W4AMV]

Hi. Add the LPF at the IF port of the mixer. Lets assume the ports of the mixer are 50 ohms. That is, our signal generator as well as antenna etc... are all 50 ohms. Then we just need to add a 50 ohm LPF at the IF port with the main purpose to pass the IF signal and reject the LO and RF signal. A simple Butterworth filter would be a good start. A 3-section LPF Butterworth has element values of 50, 100 and 50  ohms, Cshunt-Lseries, Cshunt.  Start there add more sections if required.

[ssbothwell SWL]

looking at the normalized value table for a 3 section butterworth i see g1 = 1 ; g2 = 2; g3 = 1
where do you get 50,100,50 from?

also, what should the cutoff point be set at for this filter? would 1mhz be sufficient?

with the normalized values from arrl handbook and assuming a 1mhz cutoff i get these values for my filter:

C1,2 = 3184pF
L = 15.9uH

i checked my numbers against this online calculator: http://www.pronine.ca/butlf1.htm
i seem to have done everything right but when i tested the circuit in LTSpice it isnt working right. i fed it the filter with a 1.5MHz sinewave at 1V and at the load resistor i still have a 200mV-Pk sinewave.

http://i.imgur.com/c0RB1.png

is there a way to feed the filter with noise so that when i look at the fft view i see the actual cutoff point?


EDIT:

i put the filter together on breadboard got fairly similar results. i had to round up to 7turns on a ft-37-43 torroid which gives 17.15uH and the closest i could match the capacitors was 3166pF using several mica caps in parallel. it starts to attenuate at about 1Mhz but there is still small amounts of signal until 2.5Mhz when it fully cuts off. Using a 1VRMS input i get 180mVRMS output at 1.5MHz.

is there another filter type that has a sharper cutoff? i tried a 3 section chebyshev with L,C values of 2000pF and 7.7uH but the cutoff was even less sharp. i thought chebyshev was supposed to have a sharper cutoff?

i put the filter onto the IF output and i now have a 55mV 1.5MHz output. http://i.imgur.com/YvlbY.jpg

will i see the intermediate frequency once i have completely filtered out the LO?

[W4AMV]

Hi. Because the LO and RF are so close together it will be difficult to build a simple filter, i.e. Low Pass that will provide enough attenuation. A higher Q narrow bandwidth bandpass which is tunable would be one choice. Of course it would be nice to achieve better balance in the mixer. That requires care, matched diodes, etc... I would do the best you can, add a demodulator and recognize that the audio once stripped from the carrier will be easier to filter.

[ssbothwell SWL]

i was having a really hard time understanding the big picture of how a superhet receiver works but i think i am starting to get it. can you tell me if what i say here is accurate?

1. rf signal is picked up by an antenna, bandpass filtered for the desired frequency and fed into an rf amplifier.
2. the filter is variable via a ganged capacitor which is also controlling a local oscillator tuned typically 455KHz higher than the bandpass filter.
3. the LO signal and the RF signal are fed into a double balanced mixer which produces a new intermediate frequency (455Khz in this example) which is the difference between the LO and RF signals. this IF signal has the same modulation as the original RF signal.
4. the IF signal is always the same frequency (455khz for example) because the difference between the RF signal and the LO is always the same due to them operating on a ganged capacitor.
5. carefully tuned filters and amplifiers can be applied to the IF signal because it is never changes.
6. an envelope follower is used to finally demodulate the IF signal and produce an audio frequency signal.

i didn't mention imaging or any other details but is this essentially how a superhet works?


EDIT:

i did some more post mixer filtering attempts. this time i made a 3section butterworth lowpass at 455KHz. i tuned my LO to 1800KHz and the RF signal to 1345KHz. the output from the filter is a 1800KHz signal at 58mVRMS. perhaps my mixer is not setup correctly and is not producing an IF frequency?

EDIT2: well i just went through the wiring of my mixer really carefully and everything seems to be in order. it is just like this schematic: http://www.qrp.pops.net/images/2008/2008-larger/big_dbm.gif

[W4AMV]

yes, you have it correct. I suspect the mixer is working fine, but having difficulty seeing the IF. Try increasing the LO level if possible. Clearly, with the LO removed and the RF only present, you should see no IF. Now as you add the RF to the mixer you should see at the IF port, at the very least, the LO signal modifying the RF. That composite signal will have both the SUM and DIFFERENCE frequency present. What would be helpful, and this is a good experiment, set the LO frequency AT THE SAME or NEARLY the SAME as the RF Frequency. The DIFFERENCE frequency is DC. You are now seeing the action of the mixer operating as three different functional circuits.
1. RF MIXER
2. PRODUCT DETECTOR
3. PHASE DETECTOR

As you pass the LO frequency pass the RF frequency you will see the IF cross from DC, a beat note, to a higher frequency.

A 455 kHz ceramic filter is inexpensive and building a simple IF amplifier maybe your next best project. For now though you should be able to get the mixer functional and understand its operation. If I have time I will post a BB diode ring mixer and its waveforms at your frequency and what you should expect to see.