1 Like
If you are trying to run other USB devices, it won’t work as explained in the other thread.
Airspy needs to be the only USB device for the higher sample rates.
Just a single USB device, albeit pretty much saturating a USB 2.0 port. But people on here claim to be able to do 20/24 Mhz + MLAT on one non-server / non-desktop / weak sauce device like these PIs and Odroids. I want to see it for myself. I’m feeling pretty confident that this PI4 will crap out at 24 Mhz with MLAT just like the Droid did. If it doesn’t, I’ll be pleasantly suprised and will eat crow. The key here is 24 Mhz Airspy and MLAT. If this doesn’t work, I’ll go back to 12 Mhz Airspy + MLAT.
Mike
I believe it. But have you been running an Airspy for years @24 Mhz with MLAT? or even @20 Mhz? @24 Mhz is a new option for only the most recent airspy_adsb and hasn’t been available for more than a month or two I think. I have no problems runing an Airspy @12 Mhz with MLAT on any device thus far. It’s when pushing the limits that it’s failing. I’m trying to find one of these tiny computers that’ll do it without having clock/timing issues that cause MLAT to shut down @20/24 Mhz.
Mike
You’ve got to be very close to the protocol limit at 24MHz. While USB2 is nominally 480Mbps, that’s the raw signallng rate and the protocol/scheduling overhead is substantial. some quick research suggests that you can perhaps get 4096 bits of payload * 9 bulk INs/microframe * 8000 microframes/second = ~295Mbps. 24Msps * 12 bits/sample = 288Mbps which is uncomfortably close.
I’m going by what prog (Youseff) says. If he says a USB 2/3 port should be able to handle 24 Mhz (with bit packing enabled on the Airspy, which is key and which is default when one selects 24 Mhz), then I have to believe him. In fact, 20/24 Mhz works just fine on my Odroid. No temperature issues. No CPU throttle issues. The USB Port handles the throughput just fine @24 Mhz with bit packing on. The problem on the Odroid is that MLAT timing issues arose which would disable MLAT – I’d get the notorious “clock unstable”, although there were no signs or symptoms of any data loss.
Since everything works dandy with the Airspy @24Mhz (with bit packing) on the Odroid N2 and the only issue is MLAT not functioning properly, I wanted to have another platform to test on. So I got the PI4 and want to see if it will (a) handle the Airspy at 24 Mhz (with bit packing), with no significant errors/warnings in the logs AND be able to keep a stable clock for MLAT timing and thus allow MLAT to work.
Mike
I have been running at 20Mhz for several years. Well, MLAT worked until I added skyaware978 yesterday. It did the same thing to my RPI4 the day before, however, I now know that the RPI4 cannot handle more than one airspy.
I only have the version 1 and the mini. I just ordered the version 2 yesterday.
Using the dedicated USB port on the Odroid XU4 is your best bet.
I may be able to power up the Ordoid N2 to check it out on the weekend.
If mlat isn’t functioning, that implies you are (silently) dropping data…
If it is silently dropping data, why is it silently doing so? Why isn’t it logged somewhere. Nowhere in the logs is there ever any indication that data is being dropped, by any process – airspy_adsb doesn’t report it, piaware doesn’t report it, dump1090-fa doesn’t report it, mlat-client doesn’t report it. The only thing that ever gets reported is “unstable clock”.
Tests using airspy_rx with bit packing enabled showed 100% of the time that the USB interface itself was keeping up at 24 MHz.
I thought that other people whose setup drops data eventually ended up seeing a notation in some log that data was being dropped. And since I saw no info in the logs suggesting that, I figured it wasn’t dropping data.
It’s packets being dropped.
Was written before that airspy_adsb only detects “Lost samples” that it loses due to insufficient CPU.
(It’s probably not easy to find among all the posts though)
The samples on the USB bus don’t have a running number, so it’s hard to know if any of those get lost.
I’ve tested on an Odroid N2 i have access to and while i was able to make it work at 20 MHz with stable MLAT, 24 MHz was impossible.
I’ve tested on my Laptop with the R2, 24 MHz was possible with stable MLAT.
I’m pretty sure somewhere above or in the Airspy thread reported 24 MHz working on the RPi4, but i’m not sure.
Ok, thanks. So far, I’m fairing better on the RPI4 than on the Odroid N2. The RPI4 seems to be working fine with the Airspy @20 MHz and stable MLAT. I’d try 24 MHz, but I’m going to wait til we get peak traffic during the day to see if 20 MHz ends up dropping MLAT at that point. If it doesn’t, I’ll switch to 24 MHz and give it a try. (I really hate to mess up my stats - the only statistics I have are what FA provides since I don’t graph any of my own).
The RPI4 with heatsinks and a fan (perhaps the best fan) is maintaining about 52C with all ADSB stuff running and 41C with all of it off. Not sure what it will take as far as heat before an RPI throttles CPU.
Mike
Throttling starts at 80 C or somewhere around there.
Why not? 
It’s really simple: GitHub - wiedehopf/graphs1090: Graphs for readsb / dump1090-fa / dump1090 (based on dump1090-tools by mutability)
I’ll do that. Thanks
Mike,
I have a few RPI4s in my attic.
All have heatsinks on the CPU. I use heatsink adhesive not tape and low profile heatsinks 5-7mm to clear the POE module.
One has a GPS hat so it is not easy to use the normal fans. It doesn’t do much but the temp gets to around 50-60C during the day.
Another run runs an airspy mini at 20Mhz. It has a POE module with a fan and doesn’t get much above 40C. It is set to start the fan at 35 or 40C. I could probably increase this to 50C and still not throttle.
Jon,
Admittedly I have a history of throwing money at things. I’m too impatient to shop around. I bought one of those Canakit RPI4 bundles off of Amazon because it was an all-in-one solution whereby the only thing I had to come up with was the Airspy. I’m stating this because the fan that they included seems pretty abysmal. I had read somebody’s article online about the RPI4 and using a fan, and that article suggested that no heat sinks at all + the fan they were using provided better results than heat sinks + the fan they were using. Probably had something to do with the way they may have had the heat sinks attached. And, the heatsinks I installed from Canakit were tape, not adhesive (i know better - but I’m lazy and didn’t have any good stuff around). So far it hasn’t went above 53C, and it’ll never be in my attic but in my temperature controlled basement. Lucky me.
The fact that you have one running with a POE module and a fan and it doesn’t get much above 40C is impressive, and it tells me that the fan I have is probably garbage and likely that the heatsinks aren’t doing any good. I dont’ particularly like the case that came with the Canakit. It’s easy to put together, which is great – but just the way it is constructed makes air get drawn in through large gaps between the top and center of the case (which are both above the board). I suspect that any air that any fan (mounted on top/roof of case) is moving is just air that is being brought in between the large gaps – air that never actually passes across the board at all.
So I’ll probably search out a better case, a better fan, and will likely end up scraping the heat sinks off.
I’m happy with the RPI4 thus far though, since it’s handling the Airspy @20 MHz without dropping MLAT
Thanks for your input.
Mike
1 Like
Pi4 will handle an Airspy @20 with full MLAT, no sweat. I even got a Pi3 (not plus) to handle it, but overclocked at 1350 Mhz and running on the ragged edge - not recommended by any means. I think someone reported full MLAT @24 with the Pi4 on the Airspy adsb thread. If I had to go out on a limb, @jonhawkes2030 will most likely confirm as soon as he rips the wrapper of his new AirspyR2 
2 Likes
Switching from 20 MHz to 24 MHz, and MLAT is working on the PI4. Message rates are INsignificantly higher at this time. CPU temp perhaps 1C warmer. CPU usage just a smidge higher.
But we’ll have to see how things go when things are busier. There are still thunderstorms over portions of my coverage area, and plane activity in general is down. This is all with about 125 AC, 107 or so with positions. On a busy day, with my antenna setup, I’d expect to see about 150 planes or so. We’ll see if the PI4 can sustain MLAT during busy times.
2 Likes
That really doesn’t affect how much data goes over the USB or how much work the Airspy has to do.
Some more CPU usage, which is no problem with the RPi4 as you can see on the graphs.
I’d expect the Airspy usage of the critical thread be at around 50% or so.
Overall CPU usage is not a problem anyway, just that thread can’t be more than 75 percent of one CPU for reliable MLAT.
You mind sharing some of the graphs i’m curious
Hmm. Sure I can. Don’t know which ones you want. Hesitant to do a screenshot because of possible PNG size restriction or wide screen graphic making the thread ugly.
The graphs below are for what is now feeding this site: https://flightaware.com/adsb/stats/user/mtindor#stats-33250.
I started out at 12 MHz this morning, bumped it up to 20 MHz for an hour or two and then went to 24 Mhz.
1 Like
Looking like no problem at all.
Glad it’s working now.
Guess you are limited by terrain?
What is the Maximum Range I can Get?