RTL-SDR TCXO Upgrade

[bjcuizon]

Hi, I watched the video of the guy at analogzoo.com about making a 28.8MHz tcxo for the sdr dongle and I'm thinking about building something similar. http://www.analogzoo.com/2016/03/building-a-better-rtl-sdr-tcxo/
He basically uses a 19.2MHz 10ppm Crystal where he divides its frequency by two then filters out the third harmonic going out of the divide-by-two chip which is 28.8MHz. He also uses a normal variable capacitor and a negative temp coefficient cap both in parallel with the xtal.
My question is that can I just desolder the original, un-temp compensated 28.8MHz crystal from the sdr and just add the two variable caps in parallel?

*By the way my sdr dongle is the same as his.

Thanks

Schematic Pic

[Ebel0410]

Hi bjcuizon

imho, your best bet will be to buy a dongle which is already powered with a TCXO or CXO clock.
Modifying a RTL dongle is really tricky, smd are really small,  and I think this is not the best solution.
I bought a Nooelec NESDR Smart, not expensive, and I'am very satisfied with its clock performance (miles away from low priced dongles with a simple 28.8M crystal)
Save your time and money, buy another dongle.

(I bought a better piece of hardware last month, a SDRPLAY RS2, 12 bits DAC resolution,  very impressive)
Regards
Eric

[medical-nerd]

Hiya

There are 28.800MHz DIP oscillator modules and 14.400Mhz DIP TXCO modules (which can be passed through a frequency doubler) available on ebay. The signal can then passed through a filter made from the original crystal.
I aim to try both methods over the next few weeks.

And yes - I already have 2 x V3 dongles and a Thumbnet N3, but want to tinker with the standard dongles I also have. 

If you want a relatively cheap already converted dongle have a look at:

https://www.rtl-sdr.com/new-rtl-sdr-blog-units-now-available-in-store-hf-via-direct-sampling-software-switchable-bias-tee-less-noisespurs/

https://www.rtl-sdr.com/review-of-the-thumbnet-n3/



Cheers

[metrologist]

...
I bought a Nooelec NESDR Smart, not expensive, and I'am very satisfied with its clock performance (miles away from low priced dongles with a simple 28.8M crystal)
Save your time and money, buy another dongle.

(I bought a better piece of hardware last month, a SDRPLAY RS2, 12 bits DAC resolution,  very impressive)
Regards
Eric

Just bought the NESDR SMArt dongle. I have an older RTL.SDR with two antenna ports, goes down to 100 kHz. Just realized SDR# cannot connect to it on win10

[bjcuizon]

Hiya

There are 28.800MHz DIP oscillator modules and 14.400Mhz DIP TXCO modules (which can be passed through a frequency doubler) available on ebay. The signal can then passed through a filter made from the original crystal.
I aim to try both methods over the next few weeks.

Cheers

Thanks but the price for one 28.8MHz tcxo is more than a low cost dongle right?
Maybe I'll just buy another one with tcxo and so I can listen to atc and receive adsb at the same time 

...And if you are going to test both methods out, please let us know

[bjcuizon]

I just bought "that $20 silver dongle" from amazon aka RTL-SDR V3 with the TXCO (they guaranteed 1ppm stability) and other stuff as well. Its going to arrive next week. I guess it's better than buying the 28.8 Meg TXCO for the same price.
@medical-nerd Have you tried the methods you said? I'm interested to know your findings.
Thanks

[cdev]

They make good gifts for geeky friends.

But I would never spend more than a few bucks on one.

$7 is enough for the generic Newsky dongle..

Maybe $20 for the rtlsdr blog version 3 that has all the improvements.. maybe - but only IF you don't already have one.

I think they can be had for as little as $5 now in quantity. Perhaps even less if you buy 100 or more. Then you can have your company name printed on them. Thats incredibly cheap.

[bjcuizon]

They make good gifts for geeky friends.

I know, it really would. 

But I would never spend more than a few bucks on one. $20 for the rtlsdr blog version 3 that has all the improvements.. maybe - IF you don't already have one.

Yeah, V3 seems the best as they really thought about the customers and its applications.
I guess it will stay for long before V4 as most improvements are done in this version...And PCB space is getting smaller.
Have a look at this:

[metrologist]

This is the one I got.

http://www.radioforeveryone.com/p/officially-called-nooelec-nesdr-smart.html

[bjcuizon]

My RTL-SDR arrived. I finally got it! I dont know whats the difference between amazon's  standard global shipping and priority shipping which costs a bit more. Anyway I just chose the standard one and it still got delivered in 3 business days to NZ. Thats 7 days less to what they assured me (9-12 days). 

[cdev]

I would go for an OCXO..  I think thats worth the effort.. a TCXO, much less so. I have not had any problem getting the accuracy specified for the RTLSDR TCXOs without them - maybe Ive just been lucky.

[bjcuizon]

Yeah, that's why I bought it. I'm purely interested in its direct sampling ...And then..maybe listen to HF
What's interesting is that when I went back to amazon's website for the dongle I just bought, it just said "Currently Unavailable". So I guess I just got lucky right there to grab the last piece in time.

[cdev]

You need to protect the input from ESD.

[bjcuizon]

You need to protect the input from ESD.

Thanks for pointing that out . That was one of my questions about the dongle..if the dongle had esd protection when it was directly sampled.
In the vhf/uhf range though...wel, they said they put an esd rated diode in there.

[cdev]

You should look for the schematic of how its done on the BA5SBA "500 KHz to 1700 MHz" RTLSDR box, because they do a good job there of protecting the input and extracting the signals you want.

Basically you want an LC network band pass and impedance matching network that also puts the input at DC ground potential. There is a differential balanced input o pins 4 and 5 and you want them to be the secondary of an 1:4 RF transformer so twice as many turns - also center tapped, with the center grounded through a capacitor. The input needs to be low pass filtered which can be minimally done by an L and two Cs but you can make a better one with a little thought put into it, knowing what signals you want to go where. I am just waking up and my notes are on my other computer. The discussion about the BA5SBA board is all in Chinese on the HelloCQ web site. (here is his user page http://www.hellocq.net/forum/u.php?uid=120 or https://shop107578303.world.taobao.com/?spm=a1z10.3.0.0.cuLaul ) You need Google Translate.

You should make a decent filter, its easy and definitely worth it. Give me a bit of time, I'll add more info here.  You can probably find the schematic for his board - which has an image of a dolphin jumping on it - online - I found it on the hellocq site. But that design could also be improved a lot.

You can use the rtl_power app to quantitatively compare the effects of your changes  see http://kmkeen.com/rtl-power

But I would just copy it as a start. It works pretty well. But I would experiment on the RF transformer, it seems the transformer could be much better if you were willing to invest in calculating an optimized RF transformer.

This one was suggested..

https://www.mouser.com/ProductDetail/Fair-Rite/2873002402

You could also optimize the L and C values more for for VLF than HF.

(I actually made a long post here yesterday but then thought it was hijacking your thread, so I deleted it..)

[metrologist]

I modded my NESDR with a port and considering gluing on the circuit described above if I don't integrate it inside the device. Here is also a pic of the "dolphin" circuit.

[cdev]

Metrologist,

What is/does the connector (go to)?

The two pins 4+5 of the RTL chip are a balanced differential input and quite sensitive to ESD damage.

Be careful about ESD!

You can use a transformer to isolate the chip inputs from the antenna.

See the pictures below and the schematics posted higher up in this thread.

This approach should work with all of them, see how it was done above with a generic straight from the factory, Newsky TV28Tv2 dongle which as far as this mod is concerned is electrically identical to the ones sold by rtlsdrblog, cosycave, etc.

(although I think the other midsized one works better in this context because it has better grounding for layout-related reasons, all four corners of the board being grounded, which is important to reducing HF noise. Also it uses the better crystal out of the box, and you don't have to pay extra for it.)

How you connect it can make a very big difference in the number of spurs you receive on lower frequencies.

You really need good bypassing, lots more bypass caps. The voltage regulation has to get major upgrades over what they come with for the dongles to work acceptably doing direct sampling at lower frequencies.

[metrologist]

The port connects to the Q input. This was in the link I posted earlier here (http://www.radioforeveryone.com/p/officially-called-nooelec-nesdr-smart.html)

So I gather that the two unpopulated pads are the Q+ and Q-? And I should use them both with a transformer. Can I measure the input impedance directly with a VNA?

I did not see your images before, so now I see what is under the glue...

It seems to me my older "dolphin" board with the filtering and transformer work about the same as my crude direct injection. Neither one has really picked up much of anything below HF other than than the broadcast band; however, I did catch a few CW beeps and sideband mumbling on 80m with the NESDR.

[cdev]

The thing you need is to systematically notch out or filter out all out of band interference. The way the RTLSDRs are designed (for video) makes that necessary.

You need to analyze the interference and tackle it bit by bit.  Another approach -maybe a better approach, actually, is using something like a magnetic loop antenna which simply only "works" on the target frequency.

For lower frequencies use longer wires or more loops around the circle, and more capacitance. That is actually by far the simplest approach.

Most antennas work badly if at all on really low frequencies.

A loop is much better than most of them. Another antenna that works well on low frequencies is the PA0RDT Mini whip type ("e-field probe") antenna.

It's my understanding that one can use most computer's sound cards as an input to receive VLF directly, but I have never done that. 

[medical-nerd]

Hiya

I have had 2 of the original version Wellbrook ALA1530 loops for over 10 years, one inside and one outside. They completely transformed my shortwave reception compared to an untuned longwire antenna. I have also used antenna tuners with these successfully but it is not really required.
The first version was £125 for the outside aluminium version and are available sometimes on ebay but quickly get snapped up. I would recommend reading the reviews to give an indication as to how useful loops can be.
There are very good examples of tuned homebrew loops in the on-line literature that can be built for minimum cost with a vintage variable capacitor.

Cheers

[metrologist]

Thanks for reminding me, I have a loop antenna that I built into a x-tal radio. I also found this site useful and further down is a link to a site detailing the Wellbrook, including a complete teardown, x-rays of the amplifier module, and schematic. http://www.g8jnj.net/activeantennas.htm

[cdev]

Its so extremely easy to make a loop. Definitely the best HF antenna for somebody with limited space.

Vacuum variable capacitors would definitely be the way to go. Surplus military units can be had for between $60 and $100 and seem worth it when you consider the importance of having a low resistance for the loop.

The bushings on more conventional variable caps can eat up a lot of the gain by reducing the Q.

[metrologist]

I have some crusty old silver plated open air variable caps - will have to do...

The antenna is something like 15 turns next to 3 or 5 turns

[cdev]

Most variable caps will do fine.. Depending on what size cap you have, what frequency you want to receive, and how many loops your string, most loops of wire from 18 to 20 inches or so up to a few meters in diameter - when put in series with your capacitor should be able to resonate somewhere in HF - Connect a small pickup loop to a length of coax, put it in the plane of your big loop, and experiment by tuning the capacitor up and down. Keep in mind that the peak may be very narrow so it might be easy to miss it. Set your SDR to view as wide of a range as possible, (my computer does fine with 2.8 Mhz sample rate) and then set the SDR to some frequency near where you expect it to resonate..

tune in an area where there are normally a lot of signals, and tune up and down with the cap. Look for something that looks like a wave of strong signals.

The area of strong signals may be as little as 30 or 40 kHz wide or it may be much wider. The better a connection you have electrically- the lower the resistance in the system, the sharper it may be.

Its sometimes worth it to clean your variable caps to get the best connection.

[metrologist]

Antenna is 12 next to 3 turns, and dual gang AVC, with 400 pf on the main coil and 200 pf coupling the coil. It tunes the MW band nicely and you can see the hump sweep just beyond the ends of the band when tuning.

It does nothing everywhere else. I tried tapping it at various spots, opening it up, putting the coils in series, and adding another 1000pf AVC in various ways. All I could accomplish was MW station images. I'll try to build something else and filter out the MW band.

[cdev]

You should take a photo of your setup - maybe there is some basic thing that will be obvious from a photo.

If you're talking about direct sampling, it is what it is.

Its not a $90 or more HF-optimized SDR, and it will never be. Its a inexpensive DVB-T dongle with wires connecting it to an HF antenna, a use it was never made for. They are TV cards.

As far as I know, of at least a half dozen manufacturers whose names have come up at various times in this context, only Mirics has ever mentioned using their (still mostly intended for TV use) digital tuner chips for SDR.

Still..

Its a very very simple mod as they go and the parts are so cheap as to be affordable for almost anybody.

....

Likewise with a loop antenna. A loop is as good as it gets in a limited space but its optimized for a single frequency at any particular moment, its resonant frequency. And a given diameter loop will perform best in a limited range.

Still, hams use them all the time to work DX all around the world, even on low power. Ships at sea use them because they are compact and work.

Somebody can build a loop out of everyday household items, even the capacitor.

Loops work great for direct sampling.

From what ive read I actually think that the direct sampling SDR situation certainly can be optimized quite a bit more than Ive seen it done so far.

I bet another 6 db or more of S/N ratio would likely be easy to get by better matching the input to the RTL chip.


---------

It appears the RTLSDR blog dongle has connections for i2c also brought out. So its essentially an RTL2832 breakout board.

Has anybody here used its i2c bus to talk to the RTL outside of the usual way via the RTL chip and the osmocom driver?

[bjcuizon]

By the way, I have one question (well not related the OP  )...Is the input impedance of the V3 dongle 50 or 75 ?
I read that the generic dongles have a 75 input but then they, the rtl-sdr blog admin, said that the mismatch loss between 75 dongle and 50 \$\Omega\$antenna will be less (~0.2dB).
But then I thought that they might have placed a matching circuit for it in the V3. So does anyone know?
Thanks.

[medical-nerd]

Hiya

The input impedance of the V3 dongle is 50ohm.

https://www.rtl-sdr.com/wp-content/uploads/2017/06/RTL-SDR-Blog-V3-Datasheet.pdf

(after reading this datasheet you will be unable to resist buying one!!!)

Cheers

[cdev]

The VHF/UHF input of the DVB-T dongles is nominally 75 - some may be 50 ohms, depending on if they had the factory make them from scratch using a different design.

The difference between 50 and 75 ohms on VHF input is fairly minor.

I was actually talking about the pins on the chip's impedance.

Its not the same thing.

As you can see here..  https://www.youtube.com/results?search_query=%22direct+sampling%22+RTL2832

Direct sampling is not the same at all as using a dongle on VHF and UHF.

For years now, people have been modifying their cheap RTLSDR dongles to use them for HF. Why not, since they cost so ,little and considering the price they work really well.

Pins 4 and 5 of the RTL2832 chip's input are used for HF when direct sampling

their input impedance is specified at something much higher than 75 or 50 ohms, plus its a differential pair. 

Normally that input is not something consumers ever deal with because its not used on dongles with an R820T tuner chip.

But, its fairly easy and for another few dollars in parts (if that) you can set up a filter to filter out the various unwanted signals, so you can receive HF.

It seems that a number of companies that sell dongles - for example, RTLSDR blog and CosyCave, a UK company,  now sell dongles that make it significantly easier to do that by providing better access to the pads.

Most people shouldn't though unless they know what they are doing because otherwise, say if you just connect up wires with no protection for the inputs, the risk of frying your dongle with a bit of static electricity is substantial.

That said direct sampling isnt hard to do.

You dont need to buy a special dongle either, any of them make a good starting point.

This video shows what can be done direct sampling with a generic Newsky dongle - using a home made filter and RF amp daughter board setup.

[bjcuizon]

Hiya

The input impedance of the V3 dongle is 50ohm.

https://www.rtl-sdr.com/wp-content/uploads/2017/06/RTL-SDR-Blog-V3-Datasheet.pdf

(after reading this datasheet you will be unable to resist buying one!!!)

Cheers

Thanks for clarifying that @medical-nerd..and thanks for the datasheet link.

By the way, @cdev I asked the seller if they put esd protection on the direct sampling front end and they said they did. It's also mentioned on their datasheet that they have already put a "real ESD rated diode" and still kept the BAV99 diode just for clipping the strong signals.

The BAV99 diode which is used on most RTL-SDR dongles is not a true ESD rated diode. We have added a
real ESD rated diode for better protection. The BAV99 remains in the circuit as it works a strong signal
clipper, which prevents damage to the R820T2 from overly strong signals. Please remember that not even
this will save your radio from a lightning strike or huge ESD impulse, and any permanently outdoor
mounted antenna system must have its own lightning and ESD protection.

[cdev]

The way thats worded its not at all clear he's referring to HF.

Where is the HF input,

I saw connections for other signals but not for the HF signals, where are they connected, also what do they do for low pass filtering and impedance matching? Or do they just break out the inputs.

If so, (if the HF leads specifically are not protected) given that they are fairly similar in the crucial respects I personally would rather risk one of my cheap-er dongles connecting to it for direct sampling.

One thing about using TVS diodes.. The voltage at the HF input may exceed 0.7 volts fairly consistently in high RF environments.

That causes all sorts of broadband garbage. Thats a good argument for preselection, especially, but if thats too complicated, then putting an AM broadcast band trap on any receiver you use for serious HF listening.

the same goes for below AM broadcast band.. There you need a low pass filter set at around 550 KHz..

At this point the main thing I want to use it for is stuff thats less than 1 Mhz. Thats where my usual HF receiver's coverage stops.

[medical-nerd]

Hiya

I agree it is not clear if the direct sampling mode is protected from ESD damage, however unlike the cheaper generic dongles all inputs to this dongle (HF and above) use the SMA input connector.

From the rtl-sdr site:

'HF support via direct sampling. Connect an HF antenna directly to the SMA connector and tune from 500 kHz – 24 MHz with the direct sampling mod. (No hardware modding or soldering required)'

'Improved ESD protection on the radio front end. The BAV99 diode which is used on most dongles is not a true ESD rated diode. We have added a real ESD rated diode for better protection. The BAV99 remains in the circuit as a strong signal clipper, to prevent damage to the R820T2 from overly strong signals.'

'Improved front end circuit. The standard matching circuit on the RTL-SDR was designed for DVB-T use, and tends to attenuate signals above ~1 GHz. The new matching circuit has less attenuation above 1 GHz and similar performance below. We used very high quality, high SRF, high Q inductors in this circuit.'

 'Added several access pads on the PCB. Access pads for the unused GPIO pins, CLK in/out, 3.3V, GND and I2C pins have been added. The CLK input/output is disconnected by default (see change 6). Access pads for the I branch have also been added as some users and industrial customers are using these in special projects. These pads are only for advanced users who need them for special projects. Take care as these pins are not ESD protected.'

 'Added an experimental HF direct sampling circuit, which is diplexed out from the SMA connector. This has little to no effect on VHF/UHF operation, but allows us to make use of the Q branch on the RTL2832U chip for direct sampling, which allows us to receive from about 500 kHz to about 24 MHz. (Below 500 kHz is unavailable due to attenuation from the bias tee circuit). We used a ~10dB 50 Ohm preamp as a buffer and to overcome losses in the transformer and filter. We also added a strong 24 MHz low pass filter, and added an impedance matching transformer coil to ensure good direct sampling performance.'

So the access pads to clock and control signals are not ESD protected. The HF front end passes through a buffer which possibly provided some protection to the RTL2832U chip.

It is not specified if the input to the HF section is taken off before or after the 2 protection diodes.


Cheers

[cdev]

That sounds promising!

[medical-nerd]

Hiya

There is a description of an enthusiast using the V3 dongle for HF here:

http://www.udxf.nl/The-RTL-SDR-V3.pdf

and the rtl-sdr site also describes a modification to the V3 dongle involving removing the bias-t inductor (L13 on the picture) to improve LF reception.

I can't find a circuit diagram or a detailed block diagram for the V3 but it seems that a BGA2867 mmic is used somewhere in the input path - the lower left part of the board picture.


Cheers

[cdev]

It would likely be fairly straightforward to replace the "bias tee" setup on the dongles with a much better one. (For cost reasons I am sure the setup they come with is fairly minimal)

Minicircuits has a dedicated biastee that works well all across that range for around $6 in quantity. That would likely be the ticket if it can be inserted somewhere into the signal path in lieu of the other parts.

[bjcuizon]

Hi, I have another question here....
(I'm like hijacking my own thread, hehe and I don't feel like making a new thread for it as it's just a short question.)
So I'm planning on building a quarter wave antenna out of solid core wire (probably 12AWG) but then I coudn't find anything like solid wire on websites of local harware stores or depots here in New Zealand. Most of the house wiring here is done in stranded core wires but from where I came from, they do it in solid...so I thought it would be the same.
Does any one know where to purchase such things. Or does anybody have some ideas on building it with different materials? Thanks, your help will be much appreciated.

[metrologist]

Is this for the upper bands and a small antenna? Maybe you could find small copper tube, like what is used to supply refrigerator ice cube makers? If for HF, maybe you can still use stranded, but it will likely stretch. You may also look for copper clad steel wire.

[bjcuizon]

Is this for the upper bands and a small antenna? Maybe you could find small copper tube, like what is used to supply refrigerator ice cube makers? If for HF, maybe you can still use stranded, but it will likely stretch. You may also look for copper clad steel wire.

Ooops ...Sorry I forgot to mention the band.
It's for VHF Airband 

And I'm also probably making another smaller one for ADS-B reception.

[metrologist]

I have recycled analog TV antenna's for VHF using the aluminum tubes and combination of rivets and screw clamps. Otherwise, higher bands work well using small brass or copper tubing.

[cdev]

For ADSB (1090 MHz) a quarter wave whip is 68.75mm. You can make a tiny little ground plane. That works surprisingly well if its up a bit.

Or you can make a gain antenna. Try making a collinear antenna out of many short segments of coaxial cable and seal it into a piece of thin PVC pipe. (see https://www.balarad.net/ for example )

That can give you very high gain at the horizon. You can also make another variant of UHF collinear antenna that looks a lot like a VHF ground plane with short coiled stub sections along its length.

ADSB enthusiast web sites have dozens of designs. One thing that's cool about ADSB is that you can data mine your results to generate pretty pictures of your antennas coverage and compare them to one another.

[bjcuizon]

I have recycled analog TV antenna's for VHF using the aluminum tubes and combination of rivets and screw clamps. Otherwise, higher bands work well using small brass or copper tubing.

To no avail, there was no solid core wire in the local bunnings here in NZ. So I had to find a substitute.
I found some of those galvanized wire for fencing applications, (like the material used on those diamond shaped fences surrounding the airport), and made a dipole out of it. They were already cut to 300mm so I just had to connect 2 of them for a 600mm long radial which is 1/4wave of 124Megs (roughly the centre of the vhf air band).
I also decided not to solder the galvanized wire but instead use screw terminals so that I could just disassemble it and store it easily.
I mounted the whole thing on a wall and I connected some el-cheapo coax to it.

The results turned out to be good even if it was mounted inside I get a -10dB peak (on sdrsharp) when the gain is 28dB. (And I'm 5km from the airport)

As for ADS-B, I'll probably make a 1/4wave gnd plane or a dipole using some paper clips. ...and wire some rg6 to it.

[cdev]

Are aircraft horizontal polarization?

I have recycled analog TV antenna's for VHF using the aluminum tubes and combination of rivets and screw clamps. Otherwise, higher bands work well using small brass or copper tubing.

To no avail, there was no solid core wire in the local bunnings here in NZ. So I had to find a substitute.
I found some of those galvanized wire for fencing applications, (like the material used on those diamond shaped fences surrounding the airport), and made a dipole out of it. They were already cut to 300mm so I just had to connect 2 of them for a 600mm long radial which is 1/4wave of 124Megs (roughly the centre of the vhf air band).
I also decided not to solder the galvanized wire but instead use screw terminals so that I could just disassemble it and store it easily.
I mounted the whole thing on a wall and I connected some el-cheapo coax to it.

The results turned out to be good even if it was mounted inside I get a -10dB peak (on sdrsharp) when the gain is 28dB. (And I'm 5km from the airport)

[bjcuizon]

Oops...sorry, the pics are tilted (default setting of my tablet camera)
Just tilt your head to the left

By the way, that's the best spot I can get with very less noise. If I move it 20cm to the right the noise will significantly increase.

[cdev]

You need a balun!

Maybe you already have one if you have any 300 ohm to 75 ohm matching transformers?

Oops...sorry, the pics are tilted (default setting of my tablet camera)
Just tilt your head to the left

By the way, that's the best spot I can get with very less noise. If I move it 20cm to the right the noise will significantly increase.

[bjcuizon]

Yeah, I'll probably get one of those tv ant baluns on the hardware store. Those are 300 to 75 ohm, right?
Thanks.

[cdev]

They will work although the impedance of a dipole is 72 ohms so a 1:1 would be a better match at the resonant frequency. You could build a switchable 1:1 and 1:4 balun.. Use the elecraft modem design..