CMOS synthesizer idea

Another thought has crossed my mind. By using another programmable divider for the R counter and using  a fairly high reference frequency, I can synthesize all frequencies in the 80m band at 1kHz steps whilst still keeping the reference frequency above 5kHz.
For example, with a 12MHz reference, if i set the R counter to 1724 and the N counter to 507, i get (12000/1724)*507 = 3529.0023, i.e. 2.3Hz away from my target frequency of 3529kHz. Reference frequency is (12000/1724)= 6.96kHz 
Spending some time with a spreadsheet, It looks like I can get within 10Hz of any 1kHz step in the 80M band. If the 12MHz reference was increased I imagine the error would be even less, although the pulses might end up too narrow for the phase detector.

Hi all,

Having seen the Franklin oscillator mentioned in another post on this forum, I thought I would give this a try as the basis of my ceramic resonator based VXO/pullable reference oscillator. I found the Franklin oscillator gave a wider range, with the Pierce version I had been using, I struggled to get the oscillator much above 560kHz (with a 560kHz resonator), but with the Franklin i can pull it up to 562.5kHz and down to 551kHz. I tried using lower value capacitors from each leg of the resonator to ground with the Pierce oscillator, but it suddenly hooted spuriously at around 4.5MHz.

My reference oscillator is now based on 2 inverting buffers of a CD4069, with a 220k resistor across the input and output of the first inverter, and a 10pF feedback capacitor from the output of the second inverter to the input of the first. As there are 6 inverters on this chip, I have some spare to use as buffers and to make up the 'IS' gate of Harrys CD4040 diode counter. 

The 74HC193 based programmable divider worked well on breadboard but for some reason became 'glitchy' at certain divisor settings when I made a PCB. Like an idiot, I forgot to include pull-up resistors on the PCB even though the breadboard version worked OK without them. 

Having said all this, I have several MC145152 synthesizer chips in my junk box from scrapped PMR radios, and if you ignore the 'A' counter inputs it works as a single modulus synthesizer, so the variable reference frequency plus one of these chips may be all I need. But that would be cheating...

Regards,
John

Hi John,
This is a really interesting project.

I once read an interesting article in a local club magazine where the author built a cheap 1750 Hz "tone-burst" oscillator to open repeaters on VHF. The circuit required the oscillator to be running 24/7 but only the output gated with the 500ms timer. The resonator fundamental was 455kHz, which I thought was not a wise decision. There was no way I was going to put a 24/7 running 455kHz oscillator in my VHF transciever.

As regards changing the programable divider using a CD4040, I did that with my 1.2GHz synthesiser. I used several diode-matrices. Take a look at
http://85.226.187.247/conv/4040_divider_02.htm

br Harry

I haven't got much further with the locked oscillator as I have been working on the synthesizer blocks, but plan to look into it further having never had any luck with the 'huff and puff' VFO stabilising scheme. I have gone with the CD4040 with a HEF 4093 for the reference oscillator with the diode 'AND' gates to set the divisor. I am now using a 560kHz ceramic resonator (I avoided 455kHz in case I later use this as an IF) divided by 56 to get a 10kHz reference frequency. With one of the capacitors between the input side of the oscillator and ground replaced by a capacitor and varicap in series, playing about with capacitor values I get 560kHz -7/+9kHz, so this will fill in the gaps between the 10kHz steps at my target frequency of 3500kHz. Tuning voltage is 0-5V with cheap chinese 1SV149 varicaps. Interestingly, putting the variable capacitance on the output side of the oscillator gives only a very narrow pulling range.

The programmable divider is currently made of 3x74LS193 and will initially use DIP switches but may later use a PIC and a serial in/parallel out shift register. 

I wonder if it is possible to make the CD4040 diode counter DIP switch programmable by fashioning some more AND gates from diodes? the only issue I have with it as it stands is that the diodes are not referenced to ground. If I could control the division ratio from external logic I would sooner use this scheme than the 74LS193 as I only get 4 bits of division per chip. 

Note that on the improved CMOS synthesizer project page, pin 5 of the 4046 is not shown connected to ground and this will inhibit the VCO, the PCB does have pin 5 grounded however. This caused me some confusion until I checked the datasheet.

Hi John,
I like the idea of feeding a signal into an oscillator to get it to lock on to multiples of the oscillator. I hope you can post the results of you experiments. That would be really interesting to see.

You wrote: "With the diode counter based on the CD4060 or 74HC393, etc. i find it useful to temporarily feed the output into a flip flop to halve the frequency and make the mark/space ratio 1:1 ..."
I also found that, but with the single programmable divider I suggested in the article to take the output from the MSD in the divider chain to get the same thing. The M/S is not 50/50 but it is still quite good for clocking subsequent stages.

At the moment I have loadsa projects on the table. Unfortunately we are still trying to get the house tidy after all the house-moving we have done this year. That, covid and all the travel problems it has cause has made 2020 into a really bad year. Hope 2021 will be a goodyear - tyred of 2020  

Anyway, I wish you success with your synthesizer projects. I like "simple" and available resonators sounds like a good way to go.

Best regards from Harry - SM0VPO

Thanks for the reply Harry. The synthesizer projects on your homepages have provided much inspiration for experimenting; whilst i do have a stock of devices such as the MC145152, the lack of reference frequency choice is a problem for me and there is more of an opportunity to learn and experiment by making the blocks up from simple logic devices. 

I have carried out some experiments with ceramic resonators to see what would give me the best combination of frequency stability and pulling range, but from what I have in my junk box the 455 and 560kHz units seem to be the best. Anything above 1MHz just does not seem stable enough despite trying a variety of oscillator configurations. My tunable ceramic resonator circuit currently uses a CD4001 with the usual capacitor at each end of the ceramic resonator to gnd replaced with a pair of 1SV149 varicap diodes.   

With the diode counter based on the CD4060 or 74HC393, etc. i find it useful to temporarily feed the output into a flip flop to halve the frequency and make the mark/space ratio 1:1, otherwise the needle-like output pulses are difficult to spot on an oscilloscope and do not clock my frequency counter reliably. 

Another idea I have is to divide a crystal based oscillator down to say 100Hz then lightly couple this into an LC oscillator with a low value capacitor in an attempt to injection lock the LC oscillator to, say, 100Hz steps, although I suspect this may give me unwanted FM on the output of the LC oscillator. A quick lash up on a breadboard did appear promising.

Hello John,
This sounds like fun and it is great that you are sharing your ideas with us. I agree that the math for loop filters is somewhat daunting, but I have done a lot of experimenting to produce a "rule of thumb".

I use two filters for a modulated VCO, the first is always a low-pass at 20% of the lowest modulating frequency. This RC is bypassed by a couple of back-to-back diodes that speed up the frequency changes if they are large. This RC filter is also damped to prevent the PLL from "hunting".

The second filter is usually 1 or 2 low-pass RC filters that filter out everything above the modulating frequency. This is needed to remove the reference frequency.

Simple RC filters are simple. I like simple :-)

If you are not modulating then the first filter can be very short so that you always get a good short lock time.

I have created two articles on this, with practical circuit blocks:

http://85.226.112.143/conv/syn-info.htm
http://85.226.112.143/conv/synth_10.htm

In your application you are not using modulation so the filter only needs to be good at the reference frequency.

As regards the reference frequency, I found you can use an oscillator and a CD4040 to divide by anything you want to make a reference frequency from just about ANY crystal or resonator.

See figure 4 in http://85.226.112.143/conv/4040_divider_02.htm
There is a PCB for that :-)

In my experience frequency stability is not really an issue with resonators unless you have a really high N-counter. At normal room temperatures the temperature drift is negligible. N=350 is quite low.

But please continue to share your ideas, and I hope that the above links can stimulate further ideas.

Very best regards from Harry - SM0VPO

Hi all, 

I have been experimenting with synthesizers recently but am frightened of loop filters, bode plots, poles and zeros etc. On idea I have is to use a 450kHz ceramic resonator which can be pulled a couple of kHz for the basis of the reference frequency. That way I can have continuous tuning and still use a high reference frequency. With the ceramic resonator running at 450 khz and divided by 45 to get a 10khz reference frequency, with n=350 i get to 3500kHz. Pulling the ceramic resonator to 451.285 gets me to 3510kHz with n still at 350. With N set to 351 i then get 3510 to 3520 khz, etc.  I realise this wont be as stable as a crystal oscillator based reference frequency as the drift of the ceramic resonator will be multiplied by 3500/450 = 7.78 at the low edge of the band and 3800/450 = 8.4 at the top edge of the band, but may still be an acceptable compromise. 

Shifting the reference frequency to say 20kHz and increasing the pulling range to fill in the gaps between steps may be an option as I could pull a 455kHz ceramic resonator from 447.614 to 455.097 without too much trouble. 

By the way, the 74HC393 works well as a programmable divider based on Harrys' CD4040 diode based counter scheme. 

Regards,
John