I only downloaded it yesterday myself! 
STRSVR seems quite easy at first sight. Almost too easy! I need someone to test it’s effectiveness! Hopefully I can get this AGI-3 working with an ESP32Xbee. It isn’t mine so I’m being more cautious than usual!
Level shifter is on it’s way!
I’m using STRSVR, taking a serial stream directly from the F9P board’s usb port, converting it to 1004 and 1012 then sending it to Snip as a pushed in stream. Snip is also taking the output of my first ESP32, which adds an empty 1008 and casts both as two seperate mount points.
Then I’ve got the second ESP32 sending f9p messages plus the 1008 to RTK2Go (WEBBPARTNERS).
One way or another the AGI-3 is going to work too!
It’s all interesting stuff. I’m probably doing some of it wrong but I’m learning in a way that I’ll remember it!!
Is 1008 not included in the stream that has 1004 and 1012? I assume AGI-3 needs it. I actually have all 1006, 1007 and 1008 on the stream for AGI-4. Not all three are needed, cannot remember if 1008 is enough.
AGI-4 works quite well without the additional messages if GLONASS was shut off (obviously not as well as if both constellations were active). It even finds RTK fix without 1008 but more or less only randomly and usually takes almost forever to move from float to fix. If I remember right, it is about the RTCM 3 interpretation by Topcon. The receiver needs to know if it is a Topcon base or some other brand (that did not make the mistake Topcon did).
AGI-3 could be different, have no experience.
Well, it’s very easy to add the 1008 if required. literally just add 1008 to the list of messages required in STRSVR. I’ve got no experience of this receiver as it isn’t mine but am determined to get it working.
My limited experience doesn’t go beyond an F9P.
Isn’t a 1008 just an antenna descriptor?
Isn’t it “an antenna descriptor”, not “just an antenna descriptor”? 
You tell me why 1008 is needed by many of the major brand GNSS receivers? I doubt the very detailed antenna model (available for all users from a public database) is really needed at cm level accuracy. I could guess why even a dummy 1008 can make a difference but others at the from must know this better, no need for guessing.
I’ve not looked at what STRSVR populates it’s 1008 with.Might be interesting.
Someone here said he inserted an empty 1008 into the stream. I assume Trimble was just looking if it can find a Topcon antenna description or not. An empty messages goes to the “or not” category. Surprisingly clever from Trimble eg. compared to their SBAS satellite number handling. 
Not sure if Topcon does the same.
Yes, I’ve got STRSVR adding a 1008 too now and it’s empty, exactly the same as the one my ESP32 is adding.
This project stalled as the system in question had issues. The issues turned out to be a cheap android type radio in the tractor causing severe interference!!
The system now has an AGI-3 with an internal GSM modem.
It will receive my corrections but takes an age to connect and is extremely unstable. I’ve tried STRSVR dome version converting to 1006,1075 and 1085 messages and also 1004,1006,1012. Same effect.
If I relay an euref base I have access to and strip out all but 1006,1075 and 1085 it connects immediately and is very stable.
Just wondering what I might have wrong.
Don’t know what is going on but I doubt you have done anything wrong.
Have you checked if the Euref stream has the same GPS L2 code as your private base (was it F9P?)? I don’t know how it changes things and I don’t understand why e.g. RTCM 1004 needs to state which L1/L2 code is used. I thought 1004 “just” informs the carrier phase and the modulation data could be anything, the receiver just needs to “remove the data modulation” before correlating with the carrier.
Someone here seemed to know the GPS code stuff quite well and might be able to explain why e.g. SNIP RTCM decoder says F9P refers to L2X for L2 code while for my commercial base SNIP reports L2W for the L2 code?
Even if the AGI-3 did not understand the L2 information, it should be able to find a fix with the L1 data only since your private base is so close. No idea if it takes more time, never used a single frequency receiver.
Off topic but I tried our Fendt VarioGuide RTK receiver (the Fendt variant of the Topcon AGI-4) and it did not find any fix from 1075 (and/or 1085) messages. Looks like AGCO has no purchased the update to support MSM messages. Neither do they have updates for Galileo support. The first brochures of AGI-4 refer to coming Galileo support but I asked at the TFF forum and looks like AGI-4 with Topcon cloths neither has Galileo implemented as of today. What a shame.
With STRSVR I converted F9P data to RTCM 1004, 1006, 1012 and 1033 and I got an RTK fix from my F9P base. It took time, needed good satellite visibility and was quite unstable. A bit odd when F9P as a rover seems to keep RTK fix much better than the AGI-4 when receiving corrections from my commercial base.
This is basically what I have found. Very unstable.
The L1 and 2 codes vary. I don’t understand their relevance at all.
[EDIT]
The euref base is sending L2W codes, mine L2L.
Did you manage to get anything working in the end @Alan.Webb ?
I had a problem today trying to set up a Topcon AGI-3/X30 with an F9P base.
Entered all the ntrip details, unit starts receiving data, but no RTK.
Reboot
Instant RTK fix. Works great.
5 minutes later the unit stops applying the corrections and falls out of RTK. It’s still receiving messages though.
Reboot doesn’t fix the issue.
Anyone got any ideas?
No I didn’t unfortunately. We got it working with another base on rtk2go with a staggering 150Km baseline!! Way better than it was on SBAS. I see STRSVR has been updated so I might have another go when I’ve got the 724 up and running on CAN BUS.
I just managed to obtain RTK fix with my AGCO branded Topcon X30 + AGI-4 using a correction stream from my F9P-based base station.
Had the same problem initially that RTK messages were coming through, but no RTK fix. Boot and instant RTK fix.
Will need to check next time whether it drops out of RTK in 5 min or not. I’m using STRSVR to convert the F9P data + add the following RTCM 3.0 messages: 1004(1),1006(31),1008(31),1012(1),1033(31). Some of them are empty and for some reason STRSVR loses most of the Glonass satellites from the initial MSM message data 
If I forward unconverted and converted entries to SNIP, I can see that the converted data has less of both: GPS and Glonass satellites. But RTK fix works and is obtained within a couple of minutes.
Edit: It is possible STRSVR does this reduction in satellite numbers for a reason - some older receivers refuse to obtain RTK fix if the L1 correction data has more than 12 satellites.
Did you set the TADJ parameter? What about the antenna type, ADVNULLANTENNA?
I didn’t do anything about TADJ, but did set ADVNULLANTENNA. What do you think should be done with TADJ?
I have been able to connect to a Finnish reference network DGPS (NTRIP) stream that broadcasts just messages 1006, 1071 and 1081, so for DGPS via NTRIP the message 1008 is not needed.
I also tried with my older Fieldbee L1 base (with older SkytraQ chip), relaying that stream to RTK2go and with that STRSVR does not lose any of the GPS or Glonass satellites. However, with even 20 satellites in view and close to the base, the receiver fails to move from RTK Float to RTK Fix and the Topcon/AGCO (Valtra) system complains about position accuracy and refuses to autosteer.
In the end I decided to keep the correction stream that originates from a F9P, but convert it into two separate mountpoints: L1 only and L1/L2. This way the L1 stream provides correction with max 12 satellites, but IT WORKS.
Could be different with “genuine” Topcon AGI-4 that has better software updates and might accept more satellites.
The TADJ=1 or 1.5 seems to be popular, Stefal’s rtkbase is using that, also here:
Don’t know if it is needed and if I remember right, RTKLIB was supposed to take care of the related issue on the current release.
I have the Fendt variant of the TOPCON AGI-4 receiver and it does not work without RTCM 1008 (I’m not using the F9P base). I could not run it with MSM messages either, perhaps different AGI-4 software?
Which AGI-4 SW did you have on the AGCO (I guess Valtra/MF/Challenger variant) AGI-4? I have that too but couldn’t get the latest SW last summer because it is not loaded to the official diagnosis tool. Did not help telling the SW release info to the local service and their support guys (got it from the release info from the thefarmingforum.co.uk).
I’m not familiar with MSM messages, does 1071 enable RTK accuracy? I thought 1074 and 1077 were more popular (but don’t know the difference).
Okay, so I made some corrections to my above messages to remove misinformation.
So now I have a fully functional AGCO (Valtra) AG3000 receiver (AGI-4 hardware) + AGCO version of Topcon X30 screen. With fast RTK Fix from F9P, even with just 6 satellites. AND the autosteer actually engages!
The F9P base is providing full set of RTCM messages to one mountpoint on an NTRIP server. Then - on a totally remote PC - I relay that mountpoint to STRSVR and use it to convert the original RTCM messages to just 1004,1006,1008,1012 and 1033. Message 1008 has ADVNULLANTENNA. In 1006 I have the base location that set in STRSVR in the same “options” menu as ADVNULLANTENNA. This overrides any previous base location info. Message 1033 is empty and I’m still not sure if it needed. Does no harm.
This in my view is very convenient. I do not need complicated and limited message injection systems at the base location, but can adjust the messages with STRSVR anywhere with a Windows tablet - including in the cab. It just has to be on 24/7 or at least when working.
I just tried this F9P again as a base for the Topcon AGI-4 (the MF/Valtra variant and the Fendt variant).
I had the same messages as Juurikko above plus 1019 and 1020. I was using the Stefal Raspberry Pi base, it also runs RTKLIB on the Raspberry (not sure about the RTKLIB version). Had the ADVNULLANTENNA and the TADJ=1 setting (just because Stefal has it on the default settings).
Our Valtra was doing fine, much better than at my previous test about a year ago. The Fendt was mostly fine too first but later on struggled a bit more than expected when close to buildings (where RTK is often lost even with my commercial base).
I wonder why it worked better this time. Had a look at the RTCM1004 message content (from the SNIP decoder). All GPS satellites had L2X now for L2. Previously some 30% did not use L2. I tried to google about GPS modernization status but did only find ancient documents. Would anyone know if they now all have the new L2 signal that F9P supports?
I also observed some odd behaviour. The RTCM1012 message included G32, good SNR and pseudo range values. RTCM1004 on the other hand reported 0 SNR for G2 (which actually was good from my other base). I wonder if this is an RTKLIB issue, SNIP issue or something else?