RTL Dongle - how to fix PPM drift?

[WayneH]

My dongles don't drift a fraction of that once warmed up (less than a minute in my cases). I don't believe I got lucky over a dozen times with having stable dongles, and others see drift more than 1 ppm.

I have four of them and to me it's a PITA dealing with how variable they can be. Both waiting for them heat up - sometimes I want to plug one in to grab something about a network and be done - and running them for hours.

As has been said, we're trying to do what they were not designed for, so at least some reasonably priced modifications would be nice. Otherwise the Airspy it is.

 

[Voyager]

From the video description: Dynamic frequency correction (in Hz) is displayed in spectrum window titles.

I see it now. There is so much going on, it's hard to tell where everything is. I was looking in the main window for it.

When the PPM value is set 10 below the correct value, the tuning eventually warps by 8 kHz to get back on center. Dongles drift slowly, so TRUNK88 corrects slowly.

Yes, too quick and you would run into problems with interfering signals throwing things off, or short reception nulls.

That generally works for me, but for some users, it doesn't. With a laptop or netbook, I could see where a dongle plugged in near the CPU fan exhaust holes could be affected when the CPU load goes ballistic (hello SDR#...)

So, is there any work on an SDR# plugin for AFC (which is what you've really added here). Wow. I bet I haven't used that term in a decade.

Some users have reported that wrapping the entire dongle exterior in foil greatly reduces drift. Not something I'd want to do or have had need to do.

Of course it does. The whole dongle is effectively a TCXO of a sort, so any additional 'insulation' will raise and stabilize the temperature at which it operates. Stick the thing in a piece of foam for better results.

Of course, I always wonder about that, as components and heat are mortal enemies.

Not really. TRUNK88 adjusts in 200 Hz steps, while at 800 MHz, a change of one PPM is over 800 Hz. Adjusting PPM will not center the demodulated audio - it will have a noticeable DC shift.

Agreed... shift to the tune of up to half a ppm.

What is the maximum compensation you've programmed in?

Also, is 200 Hz steps a set value, or is it 200 Hz steps at 400 MHz (or something like that)? IOW, is it truly Hz steps and not ppm calibrated steps?

When monitoring a single system, like a Moto TRS or a NEXEDGE TRS, I assume that all of the site transmitters are running off of the same reference oscillator, so not an issue. Also, nothing says that a fine tuning system can't maintain a per-channel correction factor to handle off-frequency transmitters. And I've rarely seen such transmitters, so not really much of a concern.

Much of my monitoring involves multiple systems (mostly non-trunked), so I don't have that benefit.

Thanks for the replies. Oh, and is this a feature of the current release version?

 

[WayneH]

When monitoring a single system, like a Moto TRS or a NEXEDGE TRS, I assume that all of the site transmitters are running off of the same reference oscillator, so not an issue.

Tuning in to Motorola Public Safety systems you have the greatest chance considering you'll encounter Simulcast otherwise a GPS reference is rare. Most of these private, commercial companies are going to slap the equipment in racks and hope the equipment made it out the door tuned to digital specs. I've bench checked a lot of radios (e.g., <100 per batch) in single ordered batches and I did encounter some - a handful at the most - that needed frequency correction.

I would never tolerate outside +/-100Hz of a freq used for tuning/testing when it comes to. Me, personally, I would adjust if it was >50Hz.

Anyhow, no argument here but it's not the equipment I'm trying to listen to that I'm going to blame for inaccuracies.

 

[Voyager]

I have four of them and to me it's a PITA dealing with how variable they can be. Both waiting for them heat up - sometimes I want to plug one in to grab something about a network and be done - and running them for hours.

As has been said, we're trying to do what they were not designed for, so at least some reasonably priced modifications would be nice. Otherwise the Airspy it is.

I know - keep track of 12. BTW, I mark mine with the ppm calibration on the USB connector with a sharpie so I know the CF for that particular dongle. Of the dozen or so I have, most are either in the 65 range or the 25 range. I try to keep similar ones in use on the same PCs so I don't have to fiddle with the CF. I recently sold one only because it was an oddball - having a CF that was not near one of the two aforementioned groups.

And AS... if only I knew were mine was. iMall says it shipped on the 2nd, AS says it shipped on the 5th (maybe I have those two reversed), and one tracking site says it shipped on the 8th while two other tracking sites say it has yet to ship. Nothing so far shows it hitting US customs yet.

If you buy one, select DHL!!!

This partially explains my recent 'bad mood'. I can't wait to see if it lives up to the reports others are posting. Keep your MTV - I want my AS. (you have to be at least 40 years old to get that joke)

 

[SCPD]

Andrew described how to measure the relative error of a known transmitter to your RTL stick. Analog receivers use this for AFT. Older Uniden receivers support a "WI" command to see the "window voltage" as an indicator of how out-of-tune receiver or transmitter actually are to each other.

Here's some gory details on what to do with that measurement ...
---------------------------------------------------------------------------------------
As a practical matter, the warp correction on an RTL stick can be done in two ways.

1. software translation - include the correction in the calculation.
2. set a 20 bit register in the RTL2832U.

If you use all 20 bits, the resolution is 0.0625 ppm - that about 53 hertz resolution at 851 Mhz.

That assumes the RTL stick is actually tuned on frequency.

The RTL2832U uses an SDM PLL (sigma delta modulation phase locked loop). The PLL is controlled by three registers - quotient, remainder and multiplier. The integer math is not capable of resolving to the nearest hertz. It's more like 168 hertz. I'm going by memory here so that may not be accurate but it more-or-less lines up with the 200 hertz value quoted by Andrew. The good news is this truncation error can be estimated.

When you "calibrate" your RTL stick on a given frequency - you've actually compensated for (a) gross clock error and (b) truncation in the SDM registers. That same correction value may have an additional 168 hertz error when applied to a different center frequency. This of course assumes the reference transmitter is on-frequency.

Anyone who's an absolutely obsessive stickler for having your RTL stick on frequency can apply technique #1 above to cure the truncation error. In most cases, this small error produces no measurable consequence for listening or decoding.

 

[Voyager]

Me, personally, I would adjust if it was >50Hz.

Same here, but we don't always maintain the equipment we monitor.

Heck, two of the NOAA transmitters in range here are about 400 Hz off each other, and I would expect them to be locked to a NBS standard. I like to use them for calibration because it's easier than waiting for a FM Broadcast station moment of silence.

 

[SCPD]

To make it completely useful for multi-frequency use, it would have to track the frequency error of multiple transmitters. (which IS possible, and could be done)

And is already being done. Go take a look at Denny's SDRTrunk. He has independent AFT for each channel.

 

[Voyager]

And is already being done. Go take a look at Denny's SDRTrunk. He has independent AFT for each channel.

OK. I have that installed... may need to update though. So many programs out there... (most of them really nice)

 

[slicerwizard]

So, is there any work on an SDR# plugin for AFC (which is what you've really added here).

I have no idea. I don't do SDR# plugins.

What is the maximum compensation you've programmed in?

No limit that I can recall. It will follow a signal as far as it has to.

Also, is 200 Hz steps a set value, or is it 200 Hz steps at 400 MHz (or something like that)? IOW, is it truly Hz steps and not ppm calibrated steps?

200 Hz everywhere.

Oh, and is this a feature of the current release version?

Auto correction was added in version 4.89G (16 months ago)

BTW, some folks use TRUNK88 to monitor single (non-trunked) frequencies by manually selecting a frequency as the CC frequency in TRUNK88 and then using the CC monitoring feature (Shift F9 - unmute control channel) to send the centered/demodulated audio to another program with VAC/VBC. I know some users who do this to keep Unitrunker on a control channel. There seem to be a lot of single channel digital (P25, NXDN, DMR) PS systems showing up and TRUNK88 can be used to keep a dongle centered on their RF channel.

 

[natedawg1604]

I just noticed that rtl_test -p generates a continuous report of "current PPM" and "cumulative PPM". After running rtl_test -p for about ten minutes on my RTL 2832U, the program shows a consistent "cumulative PPM" of 34.

Can anyone advise how rtl_test calculates "current" and "cumulative" PPM and whether it is relatively accurate?

 

[slicerwizard]

It counts the number of raw I/Q pairs delivered over a period of time, which is affected by the dongle's reference oscillator.