OK, here's what is happening. A recent unbudgeted equipment failure has pretty much wiped out the chance of buying a SStran AMT3000. So now I am looking for junk box ways to build a transmitter that will be frequency stable. I'm looking at the Panaxis AM-100 from the diagram and explanation in the library here, and coupling it to an antenna just like what is shown on the SStran web site. Looking to make this a functional proof of concept, and get through about a year with it. Now here is the problem... I want to run at 1640Khz, and finding a cyrstal doesn't seem to be easy. I also don't like the non-adjustability factor. So what I want to do is replace the oscillator portion with some kind of single chip (or low parts count) digitally controlled oscillator, but I'm having problems finding it. A PLL and VCO set up would be great, but if they were easy to understand and build, everyone would be using them. Most of the low part count info that I can find is for much higher frequencies, seems no one wants to make things for down in the AM band.
Any way I choose the AM-100 because I probably have most of the components to buld it, except for the oscillator. I do have 3.57xxxxx crystals coming out my ears though.
If I have to, I can use a wider bandwidth divide by N chip and use higher frequency crystals, but finding those might not even be easy. The crystals I would need (for 1640Khz) would be (in Khz):
3280
4920
6560
8200
9840
11480
13120
14760
Any help would be welcome.
Had another idea... WHile I would like to reinvent the AM-100 with a frequency agile oscillator, this might not be feasable. So would it make sense to use a lower frequency crystal (or ceramic resonator) and using a multiply by N chip to get it up to 1640 Khz?
Think I'm getting closer
http://www.dckits.com/LTC1799f.pdf
Sorry, it seems the linking is broken, you'll have to cut and paste.
Greg,
It sounds as if you have a dilemma. I checked the data sheet from your link and it appears that this IC will not provide adequate frequency vs. temp. stability to be useful. If you can, base anything you do on crystal control.
Not to discourage you but it doesn't take many orders for parts to add up to the cost of the SSTRAN kit considering minimum orders and shipping. Some IC companies will provide free engineering samples, but I doubt crystal companies will since you are probably looking for a non standard crystal.
Maybe you can extend your search and find a systhesiser chip which uses the 3.58 MHz. crystal.
If it will help, I have a 13.110 MHz. crystal in junk which would give 13.110/8=1639 kHz. This might at least let you get your circuit working before you invest in the exact crystal. I am willing to mail this to you gratis if you want. Perhaps we can PM off board.
Neil
Thanks for the offer Neil, I might take you up on that crystal. I found a couple of crystals at Allied that will get me 1638.x including one that will work with the mentioned divide by N in the AM-100 plans (4.xxx Mhz), and it was only about $2.00.
Now to make for more choices, I was looking over the Maxim site, and found some goodies that have a .5% error which is what they say you get from a crystal. These are serial programmed devices, and the first is in a TO-92 case so it is easy to work with. I'll be contacting Maxim to see what they can do for me. Here are the URLS (again no linky)
http://www.maxim-ic.com/quick_view2.cfm/qv_pk/2616
http://www.maxim-ic.com/quick_view2.cfm/qv_pk/3359
http://www.maxim-ic.com/quick_view2.cfm/qv_pk/3491
The DS 1077 and DS1085 would be nice in a microcontroller based transmitter because of the separate data pins. Note that the DS1085 has a minimum of 8.2Mhz on one output which means that either the internal divide by N could be used, or an external div-N device (at least for me). I'll probably check on the DS1065 since it is easier to solder.
One thing I need to check on is what happens to the error when you start dividing the frequency... It should also be reduced by the same amount as the division because the error is at the oscillator not the output, so these might make a very tightly controlled transmitter if you are dividing by 8 or more (0.0625 %). Either that or I'm thinking backwards and the error gets worse after division.
It's looking like I can put one of these together for about $15 plus shipping and should have a power supply kicking around to use with it. The antenna will still need some scrounging, but we might have everything on campus to maintain the buildings (it is plumbing stuff after all).
Greg,
It appears that you are doing your homework and maybe you will come up with something really nice to share here.
Regarding crystal accuracy, the frequency error in Hz. is divided down through a divider but the percent error is not. If the error is .05% that will be the error after division.
Also, be clear about frequency error and frequency stability. Usually the frequency error on a communications crystal can be "netted" to zero with an adjustable cap. but we are also interested in the long term stability of the oscillator so we don't get unacceptable drift with time and temperature. The pros use a crystal oven usually at 75C to hold the crystal temperature constant yet even then circuit changes will pull the frequency. You won't need this for part 15 AM though.
A side note: I periodically check the frequency of my SSTRAN on AM and my FM-25A on FM using a NITSC traceable counter. The SSTRAN is always within +/- 10 Hz. and the FM-25A within +/- 200 Hz. Not bad at all! I have not had to adjust either one after the initial setup.
Neil
Since both of those are PLL controlled, I would have expected the Ramsey to be better, but I guess it is a factor of at least 10 and maybe 100 higher in frequency.
Greg,
Percentagewise, the drift I cited for the Ramsey FM-25A is (200/90900000) x 100% = .00022% which is better than commercial radio standards (.005% for two way) and is well within broadcast FM standards. 200 Hz. may sound like a lot, but considering the carrier is around 91 MHz. it is negligible.
On AM the situation is different since a drift exceeding 20 Hz. will be audible if co-channel signals are present. Percentage wise, at 1640 kHz. this is (20/1640000) x 100% = .0012%, much greater than the percentage drift I have seen on either unit.
It is also worthy to note that I was able to net the FM carrier to within +/- 1 Hz. (my counter resolution) initially and a 200 Hz. carrier error in the FM band is insignificant.
Neil
I exchanged a couple emails with Greg Zimmer at Linear Technologies, and he said that the LTC6907 would be a better fit because of the tighter accuracy:
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1010,C1096,P12954,D9059
So that's what I'll probably go with.
I started looking at the Maxim chips, and none of them really seem to hit the mark. I could get 1638 and 1642, but only really close on a few settings (in their frequency worksheet). So I was kind of bothered by that. And the TO-92 package will not hit my frequency, so it's out.
So now I really just need to see what I can salvage, and what I need to buy.
I found a crystal for 1640 Khz (4920 Khz divide by 3)) that I can use, and all the trimmer caps too at http://www.surplussales.com/Crystals/Crystals-3.html
I'm still going to buy the LTC6907 and all the stuff to make it work because I really want to see if it will work. The crystal is just a backup.
So I'll be ordering the rest of the stuff that I need this weekend. I have most of the resistors in stock, so I just need the silicon devices and ceramic caps.