I looked a bit last night at the ax5043; I didn't realize (I should have remembered) that it doesn't just convert to baseband, it actually demodulates the signal. Are current designs just using that for FM demodulation and doing the AFSK modem in the TMS570? I don't see a way it could do both.
It does appear to do GMSK. The only concern there is if it can support the polynomial used by G3RUH for randomization, I think. I'd be surprised if it didn't. But I don't have the programmer's guide. And it doesn't matter, I guess, if it can't do FM and GMSK at the same time.
For AFSK, you can do a lot better than what a hardware decoder can do. See https://github.com/wb2osz/direwolf/blob/master/doc/A-Better-APRS-Packet-Demo... for details. direwolf can pull signals out of the noise in a way that a hardware decoder can't. The difference is significant. I've done some thing in my modem that can improve things even more.
There is a similar situation for 9600: https://github.com/wb2osz/direwolf/blob/master/doc/A-Better-APRS-Packet-Demo... But that's only the receive side, since this would only be transmitting it doesn't matter.
You could probably do the AFSK demodulation on a TMS570. Modulation of 9600 can probably be table driven, so that should be doable.
Note that there are far better modulation techniques than these. Almost anything being done now is using OFDM of some type. VARA is taking over in the packet world. All modern modulation for cell phones is OFDM. I think digital TV is, too. OFDM is certainly better for fading and multipath and since it's using low-baud subcarriers I'd guess it's better for doppler, too, but that's just a guess. It would affect the orthogonality (?) of the subcarriers though. Not sure.
You probably couldn't do OFDM on the TMS570. Certainly not 4 channels. You would probably need one of the TI chips that has a separate DSP.
On the ground side, anyone with a sound card modem and a reasonably modern PC could handle it. It would provide better performance, I'd guess singificantly better, than using AFSK and G3RUH. (It would be even better if you could get rid of putting it inside an FM carrier and directly modulate, but that's probably not a practical option.)
Also, on the satellite, if you converted to I and Q directly from RF, and you had a DSP or a fast enough processor, you could get rid of the AX5043s and do the FM and modem in the DSP. I remember seeing single chips that could do this, but I would have to hunt to find them.
Anyway, since we are just getting started, I wanted to point out that options are available that are better from a pure technical point of view than what is currently being proposed. I know there are other concerns like the availability of current working circuits, power budget, timeframe, etc.
-corey - AE5KM
Regarding what Golf does with the AX5043: The RT-IHU has to AX5043 chips. Under normal circumstances, one receives commands and the other transmits telemetry. So the 5043 is only doing one thing at a time.
On the receive side the uplink is FM AFSK, although not standard packet format. The 5043 demodulates, detects the tones, forms them into bytes and puts them in a queue. The software pulls the bytes out of the AX5043's queue and then does the unscrambling, forward error correction (Viterbi) and secure authentication.
On the transmit side, it is essentially the same. The processor encodes the forward error correction and line encoding (rather than scrambling) and stuffs bytes into the AX5043 transmit queue. The AX5043 knows how to do BPSK1200.
Rich can give far better details about the mechanism. I just call his routines and then pull the bytes out of or push them into the chip's queue.
73,
Burns Fisher, WB1FJ *AMSAT(R) Engineering -- Flight Software*
On Fri, Sep 23, 2022 at 3:06 PM Corey Minyard minyard@acm.org wrote:
I agree that Direwolf is probably a better performer. Can we get an audio stream out of an AX5043?
On 9/23/2022 2:06 PM, Corey Minyard wrote:
I'm running Direwolf as an APRS iGate on a Raspberry Pi and certainly have nothing against it. I guess my first thought, though, is that it runs on Linux, Windows, or Mac, so we'd have to fly something running one of those (presumably Linux). Alternatively, Direwolf is open source so I suppose we could try to port it to FreeRTOS or something similar. I don't think the 5043 gives you audio (but I could be wrong), so that's another thing we'd have to deal with.
So I guess I'm thinking that in the spirit of keeping it simple and reducing the amount of hardware and software development we need to do, Direwolf may be a bit too much for V1.
73,
Burns Fisher, WB1FJ *AMSAT(R) Engineering -- Flight Software*
On Tue, Sep 27, 2022 at 11:03 AM Bill Reed bill@brconnect.com wrote:
On Tue, Sep 27, 2022 at 10:03:27AM -0500, Bill Reed wrote:
I agree that Direwolf is probably a better performer. Can we get an audio stream out of an AX5043?
It's not so much that we use direwolf, which we can't do. It's if we are going to consider other options.
It sounds like, from Burn's description, that the AX5043 does FM demodulation then AFSK (and I assume HDLC). The packets from AFSK are then sent to the main processor where unscrambling (is this used to pseudo-ramdomize the data, or whiten it?) and FEC is done. That's not standard, as he said.
The FEC is going to go a long way to improving the performance. You could still do better with OFDM, but the current design is probably good enough.
-corey
Corey brings up a good point - what are we trying to accomplish with this board? I might have missed some early discussions about the goal, so forgive me if this has been discussed already.
Are we trying to solve pacsat only, or are we looking to design a more flexible board that might be able to do more than just the pacsat protocol? Is there an advantage to sticking with known off-the-shelf RF parts (i.e. the AX5043), or should we consider doing something more SDR-ish.
That might help us decide if the TMS570 is the "right" processor to get.
Rich
On Tue, Sep 27, 2022 at 3:42 PM Corey Minyard minyard@acm.org wrote:
I would like to see a faster uplink speed for than 1200 baud and possible higher operating frequencies.
1200 baud seems like old technology. The passes on LEO satellites do not offer the user a lot of time to connect and load messages. At one time there was the ability to load/pull parts of message on a pass and fill on a later pass. Thats work. Having a slower upload rate may mean the user will not use any passes but the overhead passes effectively. Higher speeds may mean you only plan to use the near overhead pass with a simpler antenna system or lower power.
Maybe current user base has a larger pool of equipment to support 1200 baud but with newer radios/sdr is it difficult to move them up to higher speeds? Maybe we should be looking at the user requirements and offer a solution. Higher seeds may put more strain on available memory. Higher speeds probably mean more power needed but possibly more speed may mean the transmitter is less active with fewer retries?
I like the idea of using the most advanced modulation which is most efficient if we can expect the user base to embrace the higher speed.
The operating schedule on a LOE seems to be limited due to battery and charge time needed. Speed communicating will impact this.
I have come to understand the falcon sat is so far south and low that there is very limited time to expect to work that system. The control operators are more taxed with limited/no access also. Higher speeds would help when the satellite is near the end of its life cycle.
Just some thoughts I have been having on this.
Bob
Yes, the unscrambling is for whitening. The FEC is a Viterbi algorithm which would be best if we could give it some guess as to the likelihood that a particular bit is correct. Normally it might be, for example, how close an amplitude shift keying signal was to the 1/0 line. Perhaps the RSSI would help on the 5043, except that in that case it would be the same over quite a long period--not much help.
73,
Burns Fisher, WB1FJ *AMSAT(R) Engineering -- Flight Software*
On Tue, Sep 27, 2022 at 3:42 PM Corey Minyard minyard@acm.org wrote:
participants (5)
-
Bill Reed -
Bob Stricklin -
Burns Fisher (AMSAT) -
Corey Minyard -
Rich Gopstein