3 Dollar Precision Frequency Standard

[paulie]

Playing around with my new $10 Ebay frequency counter I noticed it was a few counts off from my old 10mhz reference dug up out of the toolbox. Specs for the oscillator claimed 1ppm/yr and it was coming up a bit outside that. It wasn't obvious whether the TCXO in the counter was off or the one in the reference.

Come Saturday I took advantage of the local ham club shortwave receiver to do some zero beat tuning with WWV and turns out they were both off enough to explain the difference. Both dead on now. Another club member had also purchased one of these counters and used my reference to get his in tune. He wanted to build a cheap standard for himself so I flashed a chip and he was good to go. It occurred to me others might like to throw together something similar. If there's interest I can put up a file and diagram. Anyway here's some photos.

20mhz TCXO can be had on Ebay for a buck or so. This particular unit runs off a single AA cell. It can be alkaline or rechargeable lithium or NIxx. I prefer lithium AA because it can sit on the shelf for years and still have full charge. One of those $1 USB DC-DC converters boosts to the necessary 5v. At 50 cents for the AVR total cost was 3 dollars and change including case and push button.

Starts out at 10.000000mhz but several other useful frequencies are available by simply pushing the divide button:

Code: [Select]

; hits freq (FRQPB303 and later)
; 0 10.000000mhz
; 1  5.0000000
; 2  2.5000000
; 3  2.0000000
; 4  1.2500000
; 5  1.0000000
; 6  .50000000
; 7  .40000000
; 8  .25000000
; 9  .20000000
; 10 .12500000
; 11 .10000000
;   additional entries for FRQPB806:
; 12 80000.000hz
; 13 50000.000
; 14 40000.000
; 15 25000.000
; 16 20000.000
; 17 12500.000
; 18 10000.000
; 19 8000.0000
; 20 5000.0000
; 21 4000.0000
; 22 2500.0000
; 23 2000.0000
; 24 1250.0000
; 25 1000.0000
; 26 800.00000
; 27 500.00000
; 28 400.00000
; 29 250.00000
; 30 200.00000
; 31 152hz
;  after that starts back at 10mhz with original (FRQPB801) inc divisor once per hit (1-65536)

IIRC it eventually gets down to around 150 hz or so by pushing the button enough times. Holding it down does auto-increment but my new counter don't go down that far anyway.




Update: For those who might want to build one a hex code file is attached for Atmel atmega8. Fuseh=C0 Fusel=AF. Button input is pin 6 (PD4) and output is pin 15 (OC1A). Along with photo and info above this should be enough but I'll do a schematic and more details as time allows.




Update: Schematic added for those who like pretty pictures. Don't get much simpler than that.




Update: Another builder mentioned to me that the last couple entries in the table did not work out. I pulled out an old program that calculates divisions and sure enough he was right. Probably what happened was I pressed the button too quickly which caused it to skip entries.

Table is corrected and a copy of a program to calculate divisors and button pushes (DIVCALC.EXE) is attached. Rename w/o the .txt. To use click on it or type the name from command line and it will display results. Starting frequency and number of divisions can be entered as options. Default is 20mhz with 100 divisions. To create a text file of entries append the redirection feature to the CL (" divcalc >filename").




update: As promised a file for Mega328 (FRQPB303.HEX) is uploaded. Instead of having to hit the button a hundred times to get to 100khz only a few because each hit skips directly to next frequency in the above table. At the end it goes back to original method to allow many more divisions all the way down to 152hz.

Wiring is the same as Mega8 with addition of 20mhz clock out on PB0 pin. Not available on Mega8 because it has no CKOUT fuse option. To enable this hfuse=0xd0 lfuse=0xaf. If this feature is not used I think fuses can be left the same as Arduino. Arduino programs still work with the new oscillator. It's only necessary to tell the compiler to use 20mhz instead of 16mhz by adding an entry in BOARDS.TXT. Timing will be a lot more stable and accurate now. A friend tells me when not used as frequency reference he gets less than 2 seconds/week drift on his Pro Mini with an Arduino calendar clock program.

Note that these m8 and m328 programs also work with regular 16mhz crystal instead of TCXO. You get different frequencies but still useful like 8.000000mhz, 4.000000mhz, 2.000000mhz, 1.000000mhz, 0.500000mhz, etc.. Not everybody needs 1ppm precision.




update: Recently I needed fast access to frequencies in the 38khz/40khz range for some IR remote and ultrasonic module projects. Also working with audio circuits so the original program has been updated. The table above is edited to show the new hit counts and several more useful numbers have been added. Button pushes now skip quickly down into the audio range including 152hz. After that it restarts with one divisor per hit like the old version.

Also the button down delay is reduced so you can hit it faster. This speeds up auto increment too. A new file (FRQPB806.hex) is attached below.




...

[indy]

I was looking for a frequency reference and a time mark generator to allign my scope (somthing  like http://www.davidbridgen.com/calbox.htm )
Here I've found very good idea, and could be the basis for a nice calibration tool. Just add a narrow pulse output like the one used into calbox or into Tektronix TG 501 and 1 2 5 steps.
So yes I'd like to know some details.

[Lightages]

Yes, please publish the files!

[mtdoc]

Yes, please do!

[Thor-Arne]

Interesting, I'd like to know more. 

I've been wondering; Does increased frequency in OCXO increase the precision when divided down ?

[Skimask]

Nope.
1% off at 1Thz is still 1% off at 1Hz...

[Thor-Arne]

Ok, thanks.

So no point in going for a as high at possible base frequency and divide it down.

Nothing wrong by wanting a high max frequency.

[paulie]

True, no hit on accuracy. And multiplying harmonics like the method in the link indy posted above also no penalty. Unlike boosting with PLL which may not impact frequency but does introduce jitter. I do find the transistor multiplier circuit in that link interesting and my give the circuit a try myself.

ps. Hex file and a little more info added to first post. Thanks for showing interest. A few of the onlookers at last weekend ham meeting also want to make one now so there appears to be a real need for a time calibration unit. Maybe not rubidium clock or locked to GPS but don't cost hundreds of dollars either.

[Rerouter]

Thank you skimask, that was one of the only remaining nags in my head about a frequency counter project i have been doing on and off, using a big divider so i could use a reciprocal counter to measure ppm drift over time on a frequency,

[netdudeuk]

Good project but no TCXOs for a dollar or anywhere near that on eBay UK.

[Mechatrommer]

So no point in going for a as high at possible base frequency and divide it down.

there is, your tiny device will be capable of as high as possible frequency output. there is no pll output on that avr so you cant, say 1MHz base freq to produce multiplied 20MHz. i guess the fw is just adjusting its clk prescalers by button push. to push the limit further down, you can add dummy nops in your "led blinker" loop. and btw, the $3 avr is outputting square wave right? not pure sine?

[Thor-Arne]

I took a quick search for oscillator used in the OP's pic, ebay retuned only one hit.

Seller claims "I have no further specs available for this item as is is a proprietary part made by Raltron", so no datasheet I assume.

It's $4.99 for a lot 5 units, shipping to Noway is $29.34. Which is not going to happen. 
Shipping rates inside US should be acceptable.

[paulie]

Yeah... that's just one of the dozen or so deals I snagged since getting into this last week. It does take some shopping effort. Like favorite fishing hole, few will beat a path to your doorstep to clue you in on best bargains.  Generally to get good deals it may be necessary to buy more than one at a time and often takes some bidding or BO. Otherwise be prepared to spend a few dollars more.

Getting "pretty" numbers cheap is a little harder but if you just need stabilty and accuracy like for calibrating scope or counter then absolute frequency don't matter. With a little luck and open eyes sometimes high performance devices that might cost $10 or $20 over the counter can be had for pennies:



If you don't want to bother hunting for bargains then they are almost always available for a buck or two more:

http://www.ebay.com/itm/181032108148?_trksid=p2055119.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT
http://www.ebay.com/itm/191425644527?_trksid=p2055119.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT
http://www.ebay.com/itm/160999666268?_trksid=p2055119.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT

You can expect deals on Aliexpress to run a dollar or two below Ebay as a result of bandit Paypal fees. I rarely go that route myself though because of the excellent Ebay buyer protection.

ps. Also note that temperature compensated is not the only option here. If your environment is relatively stable around room temperature a regular crystal with trim cap will often suffice. Two of my fellow amateur radio buffs went that route and very satisfied. In my case temperature in the shop varies too much. For example in the last week recorded from 6C to 22C. I only heat when working so TC is more than a luxury. YMMV

[paulie]

The comment about it running on an AAA cell made it seem like it was a PCB assembly product of some kind...?  Most discrete oscillator through hole component units run on +5V or +3.3V or +2.5V or rarely +1.8V right?

Notice the photo and reference in OP to "USB DC-DC boost converter". These are available from china for 20-30 cents or sometimes for a buck at Dollar Tree like the one in my photo. In fact the oscillator in mine runs off anything from 3.3v-5v so a bare lithium AA technically requires no converter. However it is nice to be able to pop in a 1.2v NIxx or 1.5v alkaline in a pinch.

Nothing says you have to use batteries anyway. 5v from a wall wart or USB plus works fine too. Personally I do like to be able to carry these projects around in my pocket and not depend on the "kindness of strangers" for power.

BTW if you expect to get that kind of help on these forums it's a good idea to include at least country in your avatar. Child molesters unlikely to locate your home with such general info.

[Vgkid]

If you start delving into part numbers, you can get some pretty schmick parts for not a lot of cash.

[Dago]

Good oscillators from a real distributor are also not that terribly expensive, like this +-0.28ppm 10 MHz oscillator http://www.digikey.com/product-detail/en/DV75C-010.0M/CW750CT-ND/3524566

[MK]

The most stable quartz cuts with the best specs are usually around the 3-30MHz range, above that they are running at overtones usually, and below that they are getting quite large, and more microphonic.

[paulie]

Due to some offline requests a diagram has been attached to first post:

[paulie]

So one week after the first WWV "tune-up" I checked again and drift for the Ebay counter varied between 0-1 count (0.1ppm) and reference 1-2 counts (0.2ppm). Interestingly another club member brought in his rubidium based standard and it was 2-4 counts off over 10 minutes. IDK if this is normal for those or a defect but at 1% the cost these cheap oscillators seem to do very well by comparison. I may do another checkup in a month or two.

[paulie]

btw, the $3 avr is outputting square wave right? not pure sine?

Yes, coming from AVR pin a VERY square wave. In fact because of Atmel output components sharper edges than you might see on any other MCU in its class. So lots of harmonics, probably way up into microwave region. As mentioned in the Ridgen Calbox blog this is good if you're doing things like calibrating scope probes, LA, etc.. Maybe not so great for distributing signals from GPSDO or Rubidium unit because in some few cases equipment being fed misbehaves w/o clean signals. Judging from that thread here asking about square vs sine not a big problem though.

Anyway in the process of playing around with transistor multiplier ckt like in the Calbox site and also a poor mans WWV receiver project I built several different 10mhz resonant LC circuits. My goal was to see how cheap one could get. I found by winding a few turns of thin wire around a paper form and tacking on a small ceramic total cost was possible under 10 cents. That includes using half an axial ferrite bead as tuning slug. I wondered what about hooking it up to output of the 10mhz frequency reference to suppress those nasty odd harmonics. Lo and behold, result was clean fundamental sine wave. At least as far as I could tell w/o a lab grade spectrum analyzer. Looks great on the scope and very little picked up on HF and VHF reciever.

For those who might be curious here's a pic.

[Alex Eisenhut]

Did someone say three dollars?

[netdudeuk]

Could this be easily made to run with an AT Tiny85 ?

Or a 328P ?

[paulie]

328 may not be favorable because it is not code compatible with m8 and provides no actual advantages. Tiny85 however is code compatible with m8. So much so that I have many hex files that run on either chip un-modified. In my case m8 was chosen because it was available at lower cost than any other AVR including the Tiny series. Even bottom of the line Tiny13.

Mechatrommer hints that an MCU with PLL would allow multiples of the input clock to be generated. The Tiny85 does have a PLL capable of 64x but unfortunately only works with internal "calibrated" RC clock which is useless for a precision frequency source. Maybe STM or some other processor has ability to use external clock with PLL. That would be nice.

BTW I was able to double and even quadruple oscillator frequencies using 74hct86 XOR gates with a couple resistors. Unfortunately that does increase complexity/build-time and bust the $3 barrier. Also maybe tough to fit in that cheapo case/battery holder.

[netdudeuk]

Does the attached hex file run directly on the Tiny85 ?

Thanks.

[paulie]

It's been a while since I played with that code but IIRC this version does not run on ATTiny. I do have a version that definitely does run on both and could maybe dig it up if there's interest.

It was a little different regarding divider selection. Instead of one hit you had to enter four groups of button pushes to define the divider. This has the advantage of skipping ahead to virtually any frequency without needing to push up to 65536 times. I went with the one push version here because it's simpler and only the first few frequencies as seen in that table are compatible with my new counter. It also sends the numbers out serial at 57kbaud. Maybe this version does that too, but I'm not positive.

[paulie]

Did someone say three dollars?

LOL, almost missed this one.

I am constantly amused at the expensive and complicated Rube Goldberg setups that some hobbyists and even professionals cook up. Partly a result of sino-phobia which is pandemic in these forums and partly a strange obligation to just throw money around (Lamborghini Syndrome). Sometimes reinventing the wheel makes sense. For example education or entertainment (building is fun) but unfortunately that is not always the case.

There's only one thing better than solving a problem with a $3 Ebay trinket and that's solving it with a $1 trinket.

[paulie]

A friend noticed couple entries in the page one table were off so I fixed that and uploaded the program used to generate divisions along with details on use. If there are any questions don't hesitate.

[paulie]

Exactly one month after last calibration I thought another checkup might be in order. This time only the frequency standard was tested against WWV and was about 2 Hz off. Interestingly 2 under whereas last time it was 1-2 over so this may be indication those Raltron TCXO are pretty good for long term drift. Room temperature is about 10C higher now but I don't know if that has much effect.

Other news... on request from a couple local pals I updated the code to run on Mega328 and button pushes are easier. Instead of dividing once per hit it skips directly through the table frequencies on page one. If anybody here wants to build this version let me know and I'll put up files. Wiring diagram is the same as for the Mega8.

[netdudeuk]

Great thanks. Please will you upload the 328 stuff.

I ordered an ATMEGA8 a while ago and am still waiting.  I have several 328s.

I've also got a couple of these 10MHz Vectons so I'm raring to go.

Thanks

[Flump]

could this be done with an arduino ?as i have a spare uno and i have a 10mhz ocxo
would love to have 5mhz and 1mhz at the push of a button

[netdudeuk]

The Uno uses a 328 so that's now supported but I expect that you'd need to lose the boot loader and then it wouldn't really be an Arduino any more.  However, those chips are about a £1 from China and you could probably use the Arduino to program it.
 

[paulie]

No need to over-write the bootloader. Hex files can be downloaded with avrdude using serial interface just like any Arduino program. In fact one of the fellows here is using a Pro Mini board and he can still use it for loading Arduino programs when not being used as a frequency reference. It is only necessary to tell the compiler a 20mhz clock is used instead of the original 16mhz. This is done by way of a new entry in boards.txt file.

Hardware modifications to that tiny module were the only tricky part and that wasn't too difficult really. Just a little fine soldering work to reroute the clock to the TCXO which was HUGE compared to the original SMD ceramic resonator. The clock accuracy is like a million times more accurate and precise now. He tells me when running a calendar clock ino file it loses less than two seconds a week.

Modifying a Uno should be much easier with it's full size crystal. Too bad Arduino designers didn't think to bring out the B6/B7 pins which would eliminate the need for any modifications at all. One of the few gripes I have with using these boards whether it be Arduino or non-Arduino assembler, BASIC, or C.

Anyway I will dig up the m328 file and post when I get a minute. Putting out some work related fires ATM.

[paulie]

Mega328 file uploaded to first post along with implementation details. Dozens of individuals have worked with the Mega8 version but only a couple guys besides myself with the new one so not as well tested. If anybody finds bugs or even if it works OK it would be appreciated to post your experience.

[paulie]

Due to personal need for more convenient access to useful frequencies I  modified the original program to include several additional "quick skip" entries. Details in first post.

[netdudeuk]

Paulie, I've come to build this project today and notice from the first post that I recently bought the same frequency counter as you.

I'm really pleased with it (although did get caught out by the fact that the red and black are the wrong way round on the keyed source connector) but the display is far too bright.

Do you (or anyone else) know which of the three variable pots is used to control the display brightness ?

I'm guessing the one in the diagram but would prefer not to just give it a go.

[jdraughn]

It would increase the cost drastically, but a rotary encoder with push button switch could make a awesome little upgrade.

[netdudeuk]

Well, 99p posted for the encoder isn't too bad and even less for the switch.

Let him know when you've written the extra code 

[mtdoc]

I bought a set of these today.  This looks like a fun project and it will give me an excuse to learn how to use AVRdude which I've been meaning to do for a while.

[netdudeuk]

I finally got round to building one of these today using an M8, the Vectron 10MHz TCXO and the frqpb806.hex firmware.

I connected it to my DS1054Z and a frequency counter like that used by the OP.

Starting off at 5 MHz, the scope shows that and so does the frequency counter, fluctuating by only 1 Hz.

Stepping down, when I get to 625 KHz on the scope, the frequency counter actually shows the second harmonic (1.875 MHz).  Stepping down again and again continues to show the second harmonics on the frequency counter.

Why do I start to see the second harmonics on the frequency counter ?

Any way round this ?

Thanks

[mtdoc]

Got this working on the breadboard last night.  Had to remove the crystal on the Arduino Pro Mini and solder a wire to access the XTAL1 pin which is not broken out on the Pro Mini.  Now I just need to put it in a box. I'm still considering whether to use a battery or a power jack. One nice thing about the Pro Mini is it will take unregulated power input.

Thanks to the OP for a fun project 

[mtdoc]

Put this in a box last night. Now I have a cheap sanity check frequency standard. Yeah! Thanks Paulie for the idea.

[neslekkim]

I bought a set of these today.  This looks like a fun project and it will give me an excuse to learn how to use AVRdude which I've been meaning to do for a while.

Hm, those tcxo's seems cheap, but the seller uses GSP , $29.34 for shipping..

[neslekkim]

Do anyone have a link to this frequency counter?, found some on ebay, but most are pic based?

[netdudeuk]

There are loads of them on eBay. However, there seems to be a problem with my 2.4GHz version where once the output of the TCXO gets down to a certain frequency, the counter always shows the second harmonic and not the actual frequency, even though my scope shows the correct value.

I've posted about this earlier in this thread but no one has suggested possible causes or solutions for this.  I'd still welcome input from those more experienced, please.

[neslekkim]

yes, there are loads, mostly pic based, and I wonder if the claims are unrealistic, for instance:

up to 1.1ghz:
http://www.ebay.com/itm/181662710870
up to 2.4ghz:
http://www.ebay.com/itm/181806842808

etc..

[netdudeuk]

That's the one that I have. Is there really no one on one on the EEVBLOG forum who can answer the harmonics question? 

[netdudeuk]

Hi guys

Still not having much joy with this project.

I have this TCXO -

http://www.ebay.co.uk/itm/Vectron-10MHz-TCXO-SMD-5V-2-5ppm-30C-75C-T1115-10M000000-Qty-2-/160999666268?pt=LH_DefaultDomain_0&hash=item257c53f25c

It is running from a DP832 at 5v with the output going only to my 1054Z.

According to the scope, the frequency constantly fluctuates and is always somewhere between 9.61 MHz and 10.1 MHz.

Any idea why the output signal seems to be so far out of spec ?

Thanks

[TerminalJack505]

Hi guys

Still not having much joy with this project.

I have this TCXO -

http://www.ebay.co.uk/itm/Vectron-10MHz-TCXO-SMD-5V-2-5ppm-30C-75C-T1115-10M000000-Qty-2-/160999666268?pt=LH_DefaultDomain_0&hash=item257c53f25c

It is running from a DP832 at 5v with the output going only to my 1054Z.

According to the scope, the frequency constantly fluctuates and is always somewhere between 9.61 MHz and 10.1 MHz.

Any idea why the output signal seems to be so far out of spec ?

Thanks

Are you using the scope's hardware counter?  Or the software counter?

Sounds like you might be using the software counter.

[netdudeuk]

Ah ha !  So I was thanks.

Enabling the counter in Measure > Counter gives me a very steady 10.0000MHz.

Why would the frequency counter at the bottom of the screen be so flaky though ?

Also, as you're on a winning streak so far, maybe you would also be able to answer my questions up above regarding me using a frequency counter with this circuit and only seeing the 2nd harmonic frequency ? 

Thanks

[TerminalJack505]

From what I understand, the frequency measurement at the bottom of the screen bases its measurement on the captured waveforms.  This is done in software so, as a tradeoff between speed and accuracy, they cut some corners in their calculation.  (They likely don't account for partial cycles.)

So far as your other problem goes, I can't say.  It might be helpful if you posted the waveform.  Maybe there's something about the waveform that is confusing the frequency counter.

[Howardlong]

The counter is derived in hardware directly from the trigger input, whereas the other measurement is derived from the quantised information that's shown on the display, and all the inaccuracies that brings. If you stretch the waveform out while still running, it becomes more accurate, but it will never be anywhere near as good as the hardware counter.

[netdudeuk]

It might be helpful if you posted the waveform.  Maybe there's something about the waveform that is confusing the frequency counter.

As suggested.  The .TXT file cab be renamed to .CSV and then loads straight into Excel.

This is where 625KHz is shown on the scope and the frequency counter shows 1.87MHz.

Maybe someone can see the issue ?

Thanks

[uChip]

Hi,
This looked like a fun project, so I thought I would build one of these frequency standards for myself.  But I'm not very good at the dead bug style assembly shown in the thread.  Plus, I find laying out circuit boards to be sort of therapeutic.  I ordered up a few of the TCXOs and started a layout.

Never one to leave well enough alone, I thought I would add a small display so that you can see what frequency you have selected.  If you are going to hook this to a frequency counter the display is kind of superfluous, but I don't even have a separate frequency counter yet, so the display helps.

Also, someone asked for a rotary encoder for input instead of a button.  I thought that would be easier too so I added that as well.  It makes selecting the frequency a little easier.  The encoder also has a button, so I use that to switch between the table values and the numerical divisor.  Since the encoder makes selecting the frequencies a little easier, I added a few more common (even) values to the table.  And since in divisor mode spinning the knob all the way to 65000 is a pain, I made it so when you switch to divisor mode it jumps to whatever divisor the table mode was last pointing at.

A lot of times you will want to use this with a battery, but batteries have a habit of going dead, so I added a battery monitor that shows when the battery is getting low.

Finally, since there is a microcontroller in there anyway, I decided to make it remotely controllable as well.  There's an FTDI port that you can drive from a USB converter to send commands to set frequency and change modes.  The commands are all strings so you can drive it from any terminal emulator.

Here is a photo of the thing in operation,


a close up of the board,


and a screen shot of the output.


And since sharing is what the forum is all about, you can find all the design files, source files and documentation here: https://github.com/uChip/FrequencyReference .  I've shared out the board on OSH Park so you can purchase without resubmitting the design if you like.  Details are on github.

Did I get carried away?  Oh yeah, big time.  But it was fun and I learned a couple of things.  I totally blew the 3 USD budget, but it's still less than 22 USD total.

Thanks Paulie!
  - Chip

[fatboy77]

Hi paulie, I want to say to you many thanks for sharing this project!  My question is i saw somewhere you could build some boxes for sale and i am interested for one of them. Please let me know if it's possible.  Thank you, have a nice evening/day!

[splin]

With luck you'll get a helpful response - but note that you are posting to a 5 year old thread...

[WattsThat]

Paulie was banned...

Based on uChip’s design above, you could build you own. Two downsides that I see, one is that you’ve got to buy 3 boards minimum when you use oshPark and two, the 4 digit hp led bubble displays are now unobtainium.