hamradiotools

Nexus 1.0.2

Released 6 August 2026. The current release is 1.24.0. Download it · Every version

Added

  • SSTV transmit was sending no station identification of any kind. It now burns your callsign into every picture.

    This is the plain fact and it is worth stating plainly: an SSTV over is one continuous key-down of up to about five minutes carrying nothing but picture, and until now nothing in that transmission said who was sending it. There was no callsign drawn into the image, no CW ident after it, and the FSK ID that some stations append is something Nexus has only ever read, never transmitted. Your call now appears in the top-left corner of every transmitted image, white on a solid black plate, with no switch to turn it off — and Send is refused outright if you have not set a callsign in Settings ▸ Station. §97.119(b)(4) allows the call to ride in the image when the picture is the communication, and since the longest over Nexus can key is PD-290 at about 4:50, back-to-back images stay inside the ten-minute rule with room to spare. The end-of-communication ident is still yours to send if the QSO finishes on voice or you stop an image part-way. The plate is designed to survive the mode rather than merely be present in the file — sized as a fraction of picture width so it scales from Scottie's 320 pixels to PD-290's 800, white-on-black because that is the full tone range an SSTV mode carries, and drawn from a fixed bitmap font so the letters are identical on every machine. It is checked by decoding it back: the test suite encodes a picture carrying the plate, runs it through the real decoder and reads the callsign out of the resulting pixels, for all fifteen modes, clean and at 20 dB and 10 dB signal-to-noise. What that cannot prove is how another program's decoder renders it — for that, ask a station running MMSSTV or QSSTV what they see.

  • Send any picture to SSTV: it is resized, rotated upright and cropped for you.

    Before this you had to produce a file at the mode's exact pixel size yourself — 320×256 for Scottie and Martin, 320×240 for Robot, and four other sizes across the PD modes — or the transmit was refused. Drop a 4032×3024 phone photo now and it just works. Rotation is read from the file, so a picture taken in portrait no longer transmits on its side, which is the trap that would otherwise have bitten every iPhone user and nobody testing on a desktop. The picture is cropped from the middle to the mode's shape and scaled down in stages rather than in one jump, so fine detail — foliage, brickwork, fabric — softens instead of breaking into the crawling speckle that a single-step resize produces and that an SSTV transmission then faithfully sends for four minutes. Drag the preview to choose which part of the picture is sent; arrow keys nudge it, shift moves ten pixels, Home or a double-click re-centres. Only the axis with something to give will move, and when the picture already matches the mode it says so rather than offering a control that does nothing. Change the mode and the crop re-derives at the new shape, keeping your framing. JPEG, PNG, WebP, BMP and GIF are accepted (a GIF sends its first frame); a picture smaller than the mode is enlarged with a warning that stays on screen rather than being refused. iPhone HEIC photos are refused by name, with both of the fixes on the phone spelled out, because Nexus has no HEVC decoder and "could not load that image" would have sent you looking in the wrong place. The composer now also tells you the original size, the size it was resized to, and — before you key anything — exactly how long the rig will be transmitting.

Fixed

  • Nexus reopens at the size and the place you left it.

    The main window went back to its stock 1200×720, centred, on every single launch — resize it to suit your screen, quit, and the next start threw that away. It was most visible to operators running a manual UI scale on a 4K display, because the scale itself was being remembered: the app came back at 150% in a window sized for 100%, the wrong shape for the setting it had just restored. The size was in fact being lost at every scale; the scale mismatch only made it obvious. Nexus now records the window's size and position when you close it and restores both next time, per radio, so two rigs do not fight over one size. Maximized stays maximized, and un-maximizing afterwards gives you back the size you had before rather than the whole screen; closing while minimized keeps your last real size rather than the minimized one. Nothing is replayed blindly: a window saved on a monitor you have since unplugged, or saved bigger than the display you are sitting at today, opens centred and clamped to the screen you actually have instead of off-screen or overhanging — the main window is the whole app, so stranding it where you cannot reach it would leave no way back. The box is applied before the window is drawn, so there is no open-then-jump. (This entry was missing when 1.0.2 shipped — the fix was in the release, the note was not. The wording is Justin G0KSC's, from the parallel fix he sent as PR #11, written before either of us knew the other was on it.)

  • An FT8 QSO between two callsigns the message format cannot carry together never finished — the two stations traded the same transmission back and forth until an operator gave up.

    Nine of the sixteen combinations of callsign shapes could not complete a contact. The one most people meet is a home station working a DXpedition with a compound call such as PJ4/K1ABC; a /P station opposite a /R station, and either of those opposite any nonstandard call, are the rest. The 77-bit protocol has one message type that puts both callsigns on the air in full, and it has room for nothing else — no grid, no signal report. Its whole payload is a blank, RRR, RR73 or 73. Some pairs of callsigns force that type, because it is the only way to name both stations correctly, and once it is forced "I am calling you" and "here is your report" become the identical transmission. The auto-sequencer had no rule for that: each station read the other's report as a call, answered it with a report of its own, and then waited for a roger the other was equally waiting to receive. Nothing was malformed and nothing was mis-addressed — every transmission in the deadlock was a legal, correctly-addressed FT8 message to the right station — so nothing on screen said anything was wrong. The sequencer now separates the two the way the protocol intends, by what it last sent itself: if it is waiting for a roger, an over from the station it is working is that station's answer, and the exchange moves on. All sixteen combinations run to 73 in the usual five or six overs.

  • Those contacts also reach the log now, and the signal-report fields are left empty rather than filled with a number nobody sent.

    Seven of the sixteen combinations exchange no numeric report in either direction — the messages have no field for one, so neither station can send it and WSJT-X cannot either. Auto-log required a report to have been exchanged, as its check against logging a contact that never happened, and for these pairs that required the impossible: the QSO completed on the air and was then dropped, the Log QSO button no-opped, and a CQ run stopped dead because it waits for the contact to be claimed before it calls again. The check against a phantom contact is intact — it now asks whether the station you were working ever actually answered you and whether you ever transmitted to it, which is what it was always after — so double-clicking a stray RR73 still logs nothing.

  • A contact you never transmitted a single time could be written into your log.

    This is the one worth reading twice, because a log that invents a contact is worse than a log that drops one: it goes to LoTW and QRZ under your call, the other operator never worked you, and nothing on your screen would ever have told you. Click a station to work it, let the transmission sit queued without ever keying — transmit off, a band change, a rig that never went to transmit, or simply not getting to it — and if that station's RR73 to you was then decoded, the sequencer walked to the end of the exchange on their messages alone and the contact was logged. The Log QSO button had the same hole one step earlier: one over from the other station, no roger and no 73 from either side, and it would write the record. Both now require the same thing, in one place: that you put at least one transmission on the air in this QSO. That is the evidence the closing half of the sequence has always demanded, and the relaxation for the report-less callsign pairs above is what let the opening half go without it. Nothing that genuinely reached the air is refused — a real contact keys several times before it ends.

  • A /P or /R station could acknowledge a signal report nobody had sent it.

    The rule that lets the report-less callsign pairs finish reads a bare "you, me" transmission from the station you are working as their answer, because for those pairs the message format has nothing else to put in it. It was reaching too far: it applied to any callsign containing a slash, and a portable or rover call working an ordinary station carries its grid and its numbers perfectly well. A plain station calling you with no locator set sends exactly that bare form — so you rogered a report that was never on the air, and the other operator's own sequence then had nothing to go on. The rule is now limited to the pairs whose transmissions genuinely have no room for a report; everywhere else a repeated call is treated as "they have not copied me yet" and your report goes out again, which is what it means.

  • A contact whose closing 73 never arrived was lost, and a CQ run stopped there.

    In the callsign pairs above the roles of the exchange swap over: your roger is the transmission that ends the QSO, the other station takes it as the finish, logs you and signs off. Nexus was still treating that roger as the middle of the exchange, so it sat waiting for a confirmation the other operator had no reason to send, and if their parting 73 was lost to fading — which is precisely the transmission most likely to be lost, since nobody repeats it — the contact went unlogged, and a CQ run stopped calling because it waits for the contact to be claimed. The contact is now claimed off your own roger, once that roger has actually gone out, exactly as an ordinary FT8 contact is claimed off your RR73; the run carries straight on to the next caller. The other station has logged you either way, so this is the log agreeing with theirs rather than a contact invented out of nothing.

  • Two stations that called each other at the same instant never got past exchanging grids.

    Answer a station at the moment it answers you and both sides believed the other owed the first signal report, so both re-sent their grid, over after over, indefinitely — legal, correctly addressed transmissions that went nowhere. WSJT-X answers a grid addressed to it with a report whatever else it is doing, and Nexus does now too: whichever of you decodes first sends the report, the other rogers it, and the contact finishes in the ordinary five overs.

  • The SSTV mode picker understated every mode's airtime by about a second.

    The pixel sizes were right, but the durations beside them were a hand-maintained copy that had drifted from what the encoder actually emits. They are now the exact figures, and a test compares the two tables so they cannot drift again — which matters more than it did, because the composer now uses that number to tell you how long the transmitter will be keyed.

Fixed

  • Connecting to a serial rig could put it straight into transmit — and hold it there.

    If your PTT Method is anything other than RTS or DTR on the CAT port — VOX and CAT keying, which is the default and what most operators run — Nexus started Hamlib's rigctld without telling it anything about the port's control lines. A serial port's driver raises RTS and DTR when the port is opened, and Hamlib only puts a line back down when it is the line it has been asked to key with, so on the default both stayed up for the whole session. On an interface wired to key the radio from RTS (or from DTR) that is the transmitter switched on by the act of connecting, with nothing in Nexus that would switch it off. Nexus now asks the daemon to hold those lines low when it opens the port, and closes the same gap in the two places it opens a serial port itself: serial PTT, which drove only the line it keys and left the other one up, and the native CI-V connection used by scope-capable Icoms, which keys nothing and so left both up. The CW and RTTY keylines already did this; the rule now lives in one place instead of being remembered at each. Who was exposed: a serial CAT connection through an interface that keys PTT or a CW line from RTS or DTR — the classic single-cable and homebrew interfaces, commercial CAT cables wired that way, and anything built to the RTS-is-PTT convention. On every platform, not just Linux: the earlier note here said Linux and probably macOS, and that was wrong about the platform most of you are on. Linux and macOS raise both pins on open. Windows raises them too, and while Hamlib lowers DTR itself as it configures the port, it leaves RTS to the driver — and RTS is the line the reports are about. A network/TCP rig has no control lines and was never affected. What to check: if you have ever seen the radio key on connect, drop to receive, or show TX with nothing sending, this was a candidate cause. What has NOT changed: if you key by RTS or DTR, Nexus never touches that line, so keying works exactly as before. Neither is RTS touched on a radio whose Hamlib backend uses it for hardware flow control (the FTDX10, FT-991, TS-2000 and TS-590 among about fifty others), nor is either line touched on a radio whose Hamlib backend keys the transmitter from that pin in its own right — the Yaesu FT-980 keys from RTS and the FT-757GXII from DTR, and Hamlib refuses to open either of them if it is also asked to hold that pin. Nexus asks the Hamlib you actually have, per radio, which lines it will accept before it starts the daemon, so this stays right when Hamlib changes: four popular radios gained the flow-control declaration between two Hamlib releases, and a list baked in here would have gone stale in the direction that costs you CAT. Anything it cannot establish, it leaves alone. If your interface takes its power from the DTR or RTS pin rather than from USB — a design from the RS-232 era — it will now see those pins low. What this is not: it is a real defect, found by reading the Hamlib source and confirmed against the Hamlib Nexus ships, and it matches a published report of the same behaviour with the same FT-847 cable under a different program. It is not established as the cause of any particular report of a silent radio, and it is not claimed as one — silence and a stuck transmitter are different symptoms. Nobody here has a serial rig: every Hamlib refusal above was reproduced by running the bundled rigctld, but no pin has been watched moving.

  • The Needed board hid US-spotted rows on exactly the pileups where a US spotter matters.

    An HF spot reaches the board when someone on your continent heard the DX — a JA station a Kansas skimmer copied says something about a path from your QTH; the same station heard only in Europe and Japan does not. Nexus keeps a list of the other stations that reported each DX, capped at eight so a busy pileup cannot grow it without limit, and it was keeping the eight that arrived first. So the moment a ninth reporter came in, the newest voice was the one thrown away — and on a pileup being re-spotted from Europe and Asia several times a minute, the single North-American skimmer was gone within seconds of arriving. The board then saw no local report and dropped a row you asked to keep. The cap now keeps the reports that answer the question the board is asking — the ones near you first, then the ones on your continent — so one US skimmer is enough no matter how loud the rest of the world is.

  • Changing digital mode could also make the Needed board's own-radio rows lie about their age.

    The same clock renumbering, one screen over. "Decoded by YOUR radio on this band" is the strongest row the board can show — it is your own receiver, not somebody else's report — and it is capped at two minutes old for exactly that reason. That cap was computed the unguarded way: a station decoded on FT4 at 17:00 and read after a switch to FT8 at 18:00 came out as zero seconds old, so an hour-old decode led the board as the freshest thing on the band. Both places that turn the heard-list's slot number into an age now go through the same conversion, and it refuses a number that belongs to a clock no longer running.

  • A 6 m spot's local reports were being thrown away in favour of ones the 6 m gate cannot use.

    Nexus keeps up to eight of the other stations that reported each spot, and the entry above taught that cap to keep the reports nearest you first. It measured "near" once, at the widest radius any band uses — 800 km, which is the 2 m figure. On 6 m the board asks a tighter question: it wants two reports from inside 250 km, because that is the size of a sporadic-E patch. From EN52, sixteen skimmers publish a grid inside 800 km and only five inside 250 km, so the eight slots filled with reports that ranked as "near" while being no use to 6 m — and pushed out the ones that were. A genuinely corroborated 6 m opening then read as a single report and was dropped. The ranking now asks each band's own question at that band's own scale.

  • A short-skip sporadic-E opening on 6 m did not raise an alert.

    The regional opening gate — the one that fires on a band-wide surge among the receivers around you — gained a distance test below, so that a busy evening of local 6 m FT8 could not pass for an opening. The test asks for a path past 500 km, but the near-receiver half of it was measuring 700 km. Sporadic-E on 6 m starts at about 500 km, and a burst that puts a dozen stations 540–580 km out through three receivers near you is a real opening and the one a 6 m operator most wants to hear about; it satisfied neither half and the gate stayed shut. The near-receiver test now measures at 500 km, where the physics puts the floor. Nothing else moved: the receive-only sentinel — the alert that fires when you are not even on the band — still needs two independent receivers past 700 km, and the anti-superstation rule it exists for is unchanged. One near receiver is enough for the distance test alone because that test is not the anti-superstation check: a big station on a hill can inflate a count of stations, but it cannot make a path longer than it is, and the gate separately demands three distinct local receivers before it believes anything.

  • Changing digital mode could make a dead 6 m band alert again.

    The opening detector times your own decodes by the slot clock, and that clock is renumbered whenever the transmit period changes — 15 seconds on FT8, 7.5 on FT4, 4 on FT1. Stations already in your heard list survive a mode change carrying their old numbering, and read against the new clock they could compute as "decoded this instant". An hour-old decode became the freshest evidence on the band, which is the same stale-evidence fault that made 6 m alert on nothing and stay latched. Entries numbered under a clock that no longer exists are now dropped rather than guessed at; anything still on the air is re-decoded within a slot or two.

  • 6 m could still open on activity that never left the neighbourhood.

    The tightening that removed distance-blind evidence from the opening gate missed one path: the regional gate, which fires on a band-wide surge near you and asked four questions — how many stations, how many local receivers, how many two-way contacts, how band-specific — without ever asking how far anything went. A busy evening of local 6 m FT8 among a dozen neighbours answers all four the same way a sporadic-E opening does. That gate now also requires at least one path past 500 km, either one of yours or one a nearby receiver copied. HF is untouched: an F2 opening's paths are continent-scale by definition, so a distance test there would only add noise.

  • The openings log under-counted the stations in a VHF opening.

    Confining the 6 m/2 m open gate to the last ten minutes — so a ninety-minute-old decode can no longer prop a gate open — was also applied to the station census the display and the openings log show. A forty-minute Es episode with thirty distinct stations, never more than a handful in any one ten-minute window, was journalled as about a quarter of the stations it actually carried. The gate keeps its ten-minute window; the counts are the whole episode again.

  • Purging the logbook left the LoTW and eQSL sync positions behind, so the next sync brought back a sliver of your confirmations.

    An operator purged a 26,000-QSO log, re-synced, and his awards card read "3% — 816 of 26,007 QSOs confirmed": DXCC 131 against LoTW's 249, the Challenge 277 against 1,202, WAZ 31 of 40, and 23 DXCC credits against the 248 ARRL had granted him. Every worked total was right and every confirmed total was wrong. Each service is synced incrementally — Nexus remembers the date of your last confirmation and asks the service only for what has been matched since — and purging the log did not clear that date. So the next "Sync LoTW now" asked for the last few days of matches against a log that had just been emptied, and brought back a few hundred confirmations instead of a whole history. That date gates the confirmation pull and nothing else, which is why the contact totals looked right while every award did not. Nothing but changing your LoTW username ever reset it, so the rest stayed out of reach permanently. A purge now clears both positions. The purge dialog says so before you confirm, because the sync that follows is a full confirmation-history download and takes considerably longer than a routine one — it is the pull that brings your confirmations back, not a hang.

  • Importing a confirmation report over the contacts it describes changed nothing.

    A LoTW, Club Log or eQSL download restates contacts you have already logged — that is what a confirmation report is — and the confirmation, the granted award credit and the STATE and COUNTRY ride on those restated rows. Import treated a row matching a logged contact as nothing but a duplicate and dropped the whole row, so importing a download over the log it describes skipped every row as a dupe and repaired nothing. "Already in the log" means do not log it twice; it never meant ignore what the row says about the one you have. An import now merges those rows into the contacts they describe, using the same one-way merge the Sync confirmations button uses — it only ever adds, so no import can un-confirm or un-credit a contact, and a QSL you have merely requested still counts for nothing. The import toast reports updates separately from the dupe count, because an import that adds no contacts at all can be the one that repairs every award total you have. This is the hand-repair path: if your confirmations are behind, download your report from LoTW (or Club Log, or eQSL) and import it, and it lands on the contacts you already hold instead of being thrown away.

  • Logging a contact re-read your whole logbook off the disk.

    Two copies of Nexus can share one log.adi, so before rewriting that file Nexus checks whether the other copy has written to it since — a cheap comparison of the file's timestamp and size against what Nexus last saw. Writing a contact onto the end moved both, and Nexus never updated its own note of them, so the check missed every time and re-parsed the entire file instead: measured at 60 ms on a 26,007-QSO, 3.7 MB log, on every contact you log and on every contact WSJT-X hands over in companion mode. Nexus now notes what it wrote — but only when it can account for every byte in the file, so a contact written by the other copy still forces the re-read that keeps it from being overwritten.

  • A confirmation marked V read as unconfirmed.

    ADIF spells a confirmation you hold two ways — Y, and V for one an award credit has been granted against — and Club Log and DXKeeper both write V. Nexus read only Y, so importing a master log from either of them silently dropped the confirmations on exactly your best contacts, the credited ones. Both are now read; N, R (requested, not received) and I still confirm nothing. A value padded with spaces by a sloppy export is no longer read as a refusal. This is separate from the two LoTW faults above and never touched a LoTW download: LoTW's own report only ever writes Y or N.

  • Light theme: the "NEW ONE — Work it" pounce banner had no readable text.

    When a needed station appeared, the banner slid down from the top of the screen with its NEW chip and its Work it button — and nothing between them. The callsign, the country, the frequency, the mode and the band were all being drawn in the light theme's near-black ink on a near-black background, so the one thing the banner exists to tell you was the one thing you could not read. The chip and the button paint their own colours, which is why they still showed and the data did not. Dark mode was never affected. The self-update prompt at the bottom right had the same fault from the same cause and is fixed with it.

  • Light theme: the "needed" colours were the dark theme's colours everywhere they appeared.

    The eleven colours that say what is worth working — magenta for an all-time new one, violet for a new zone, orange for a new band-slot, and the rest — had light-theme versions written for them, and none of them were ever used. The light theme rendered the dark theme's pastels instead: colours picked to glow against a near-black panel, shown on white. That affected the need chips on roster/station cards and in the Satellites pass list, the tags and row tints in the decode feed, the POTA/SOTA/DXpedition badges on the band strip and band map, and the mode cells on the Needed board — where the FT8, FT4 and RTTY cells were pale fills carrying near-white text, effectively blank. All of it now uses the light inks. You will see these colours change in light mode; they are the same eleven hues, taken dark enough to read on a white panel. Three of them (new US state, wanted, new mode) were re-tuned a shade darker than originally written, because this is the first time they have ever been rendered and measured, and they fell short of the 4.5:1 readability floor. The world map is not part of this: it carries its own copy of the palette and still draws the dark colours in light mode.

  • Light and dark: the ATNO, new-band and SOTA badges had unreadable labels.

    The white letters on the magenta "NEW ONE" decode tag (3.4:1), on the orange band-slot tag (2.2:1) and on the violet SOTA badge (2.5:1) all sat below the 4.5:1 readability floor in dark mode. Every badge that fills itself with a needed-colour now takes its letter colour from the theme, so it stays legible in both. You will see those three labels flip from white to near-black in dark mode.

  • The pounce banner's bottom corners are rounded again — the rule that rounded them named a design token that does not…

    The pounce banner's bottom corners are rounded again — the rule that rounded them named a design token that does not exist, so it had been doing nothing and the banner was shipping square-cornered.

  • A Hermes Lite 2 running Thetis could not get CAT working — and Nexus told its operator to check the radio was powered on while quoting that radio's own greeting back at them.

    CAT worked in WSJT-X. It worked in Nexus too, but only after the operator discovered that picking a FlexRadio profile got it going, which is not a thing anyone should have to find out. Three faults, one report. The rig list named no program. For a Hermes Lite 2, an ANAN, a legacy Flex, the SDR console on your PC is the CAT port — the board on the desk has none of its own. The list named only hardware, so searching it for your radio turns up nothing, and the FlexRadio entries are the closest-looking thing there. They do connect. They also cost you the S-meter (Hamlib's FLEX-6000 profile carries no signal-strength reading at all) and send keying without the read-back the PowerSDR-family profiles use. The default rig list — the one you see without ticking "Show all models" — now carries Thetis (Hermes Lite 2 / ANAN / HPSDR), PowerSDR / mRX PS (Apache ANAN / legacy FLEX), piHPSDR / OpenHPSDR (Hermes Lite 2 / ANAN) and SDR Console, each named for the program you launched. PowerSDR is no longer filed under FlexRadio, and no longer described as a TS-2000 emulation, which it never was. These four are list entries only: a program is not a USB device, so Detect and the CAT port auto-test never suggest one and never replace the model you configured. Nexus mistook a rig's CAT port for a rigctld. Before starting its own copy of Hamlib's rigctld, Nexus checks whether one is already running so it can share it rather than fight for the radio. That check accepted any reply as proof — and Thetis greets every program that connects to it. So the greeting alone convinced Nexus a raw CAT port was a rigctld, and it spoke the wrong protocol at a perfectly healthy CAT server until the timeout ran out. The check now reads what came back. When the program announces itself, Nexus names it: "…is Thetis's CAT server, not a rigctld", quotes the greeting it was sent, and names the profile written for that program. If you are on a FlexRadio profile it also spells out what that profile is costing you. Nothing is changed for you — the message says what to set. Both ends of the chain on one port. When Nexus starts its own rigctld, that daemon binds a port and then dials your rig — it cannot dial the port it is listening on, and nothing checked. Giving the rig's Network Address and the rigctld TCP Port the same number is now caught before that daemon is launched, with both numbers read back so you can see the collision instead of being told about it. It is checked at that moment and no earlier, on purpose: when a rigctld you started is already on that port — the setup the manual prescribes for a radio outside the rig list (NET rigctl, Network Address 127.0.0.1:4532, rigctld TCP Port 4532) — the two ends genuinely are one endpoint, and Nexus shares it exactly as it always has. What this does not do: recognition is by greeting, and only Thetis and PowerSDR announce themselves. piHPSDR, SmartSDR CAT and rigctld itself all stay silent until spoken to, and Thetis's greeting can be switched off in its own settings — in every one of those cases Nexus reports what answered and quotes it, rather than guessing a program. No hardware is ever inferred from a greeting: the interesting-looking tail of a Thetis banner is a build label, not your radio. Two smaller things from the same report. The Connection row read "Network (FlexRadio / remote)", which tells an SDR operator the row is not for them; it now reads "Network (host:port — SDR software, or a remote rig)", and the address field says where to read the real port out of your program instead of leaving you to guess (Thetis's TCP/IP CAT server is a different box from its TCI server, and only one of them is CAT). And the two FlexRadio native-stream toggles — SmartSDR panadapter and DAX audio — were offered to any rig whose model name contained "flex", which the PowerSDR entry's did; they now appear only for an actual FLEX-6000.

Every version