本页尚未翻译,目前显示英文。
Retevis RT10 hardware
A DMR and FM handheld for 902-928 MHz on the HR_C7000 SoC, with an FD6818 RF chip, no display and a 16-position channel
knob. A value marked (measured) was checked on a radio running RadioFoundry, (unverified) is not confirmed on a radio
or in a datasheet, and (?) is not determined. Unmarked values come from the radio's observed operation and the board's
markings.
1. Chips
| Part | Marking | Role |
|---|---|---|
| SoC | HR_C7000 | C-SKY CK803S core and DMR baseband |
| RF transceiver | FD6818 | receive and transmit, on a two-wire bus (section 7) |
| Reference | 26 MHz TCXO | the FD6818's reference; the modem's DC bias sets its frequency (section 4.1). Marking unverified |
| SoC crystal | unmarked | 24 MHz (unverified: the PLL settings give 168 MHz only from 24 MHz) |
| Program flash | GD25Q16C, JEDEC C8 40 15 (measured) | 2 MB, execute in place at 0x03000000 |
| Data flash | GD25Q16C | 2 MB on SPI0 (section 4) |
| Audio | NJM2904 dual op amp | analog, no control lines |
| Clock | SoC RTC, no backup cell | no coin cell or backup capacitor on either side of the board. The SoC RTC is on I2C2 (address 0x70). The long two-pad crystal marked EAUE beside the SoC is probably its 32.768 kHz crystal (from the package; not measured). Not measured: whether the RTC stays on the main battery with the radio off. RadioFoundry doesn't use the RTC on this radio |
2. Boot and program flash
There is no vendor loader. After reset the boot ROM offers its serial download window on UART0 (the accessory jack's programming cable; measured), then reads the flag at flash 0 and jumps to the address it names.
| Range (flash offset) | Content |
|---|---|
| 0x000000 | ROM flag 74 68 64 0C (thd\x0c), then the entry address, little endian |
| 0x001000-0x00BFFF | unused by RadioFoundry |
| 0x00C000 | the RadioFoundry application (linked for 0x0300C000) |
| 0x0E8000-0x0FFFFF | the codec slot |
| 0x100000-0x1FFFFF | unused |
- RadioFoundry's flag names 0x0300C000.
stub/stubtool.py --write-appwrites the flag sector after the application when the board profile hasrom_entry(board.toml[flash.stub]), so the radio keeps starting whatever was there until the new image is complete. - The original image starts at 0x1000 and runs to about 0x78000, so installing RadioFoundry overwrites it. Only a full
2 MB backup (
stubtool --progmem FILEwith the board profile, or the web flasher's backup) puts it back: the web flasher's restore of that backup writes the application area, then the backup's flag sector (web-flasher.md). - The web flasher installs as stubtool does: the flag sector last, after the application is verified.
- The boot ROM's window is the recovery path: with the radio off,
stubtool PORT --write-app IMAGE, then switch it on. - When no answer to its greeting reaches it, the ROM sends the address it starts, little endian, and starts it:
00 c0 00 03with RadioFoundry's flag, seen 79 ms after the greeting (one power-on, through a PL2303 cable).
Start-up state: the boot ROM leaves the SoC on the 24 MHz oscillator and the XIP controller in single-bit mode. The platform's clock set-up (platform/hrc7000/clock.c) starts the PLL at 168 MHz CPU and 84 MHz APB.
3. Pins
Index = GPIO pin index (0-31 PTA, 32-63 PTB, 64-95 PTC). The IO mux stays at its reset default except SPI0 (IO_DIPLEX1 bits 8:3 cleared, bits 29:27 set).
| Pin | Index | Use | Level |
|---|---|---|---|
| PTA0-PTA4 | 0-4 | JTAG (unverified: the unpopulated 7-pad header) | |
| PTA5 / PTA6 | 5 / 6 | a second two-wire bus, no device found | |
| PTA7 / PTA8 | 7 / 8 | FD6818 clock / data | bit-banged |
| PTA18 | 18 | data flash chip select | low = selected |
| PTA20-22 | 20-22 | SPI0 clock, MOSI, MISO (unverified) | |
| PTB0 | 32 | TX supply | high = on |
| PTB1, PTB2, PTB6, PTB21 | 33, 34, 38, 53 | channel knob code bits 2, 1, 3, 0 | |
| PTB3 | 35 | PA select | high = on |
| PTB4 | 36 | RX supply | high = on |
| PTB5, PTB7 | 37, 39 | LEDs: PTB5 green, PTB7 red (measured: red at power-up and in TX, green on MON) | high = on |
| PTB8 | 40 | speaker amplifier | high = on |
| PTB9, PTB10 | 41, 42 | side keys | low = pressed (unverified) |
| PTB11 | 43 | PTT; also the cable's serial line | low = pressed |
| PTB12 | 44 | power latch | high = stay on |
| PTB13 | 45 | the FD6818's AF into the amplifier: high with an open squelch, low for beeps and DMR voice | high = passes |
| PTB17 | 49 | power switch sense | high = off (measured) |
| PTB26 | 58 | FD6818 GPIO6, its squelch output (measured) | high = open; low with PTB27 high: a held match (section 7) |
| PTB27 | 59 | FD6818 GPIO0, the sub-audio compare result (measured) | high = the tone or code is heard |
| PTB28 / PTB29 | 60 / 61 | voice prompt chip clock / data | |
| DAC B | PA control, after IO_DIPLEX2 bit 28 is cleared | 0 = off | |
| DAC C | routed to its pin (IO_DIPLEX2 bit 29), 1762 at start-up: the modem's AF ADC bias (7.1) | ||
| ADC channel 2 | battery | divider: ? | |
| UART0 | programming cable, 115200 8N1 under RadioFoundry |
The transmitter is off with PTB0 and PTB3 low, PTB4 high and DAC B at 0 (board_rf_safe).
3.1 The cable line
PTT (PTB11) and UART0's RX pad are one line on the K-type jack (two prongs); the PTT gate is in gotchas.md 16.2. UART0's TX is a separate line: the radio doesn't receive its own characters.
The line can't take both directions at once (measured, with a K-type FTDI cable). When the radio starts sending while the host is still sending, the line is clamped low within milliseconds and stays low until the power is switched off (gotchas.md 16.5):
| Check, the line latched | Result |
|---|---|
| The radio | runs on: keys, LEDs, the host link task's loop |
| The host | receives nothing more from the radio; the radio receives nothing |
| PTB11 driven high by the SoC, read while driven | low |
| The cable unplugged; its USB-serial chip reset | low |
| The SoC reset by the watchdog | low |
| A power cycle | released |
| Trigger | pppd: LCP's ConfAck and an Echo-Request back to back; debug link requests sent without waiting for the answers (15 ms). Requests answered in turn never latched it |
RadioFoundry: a host port write waits for 2 ms with no received character first (HRC_HOST_TX_IDLE_MS in
boards/rt10/board.mk), and PPP is off (boards/rt10/config.mk). The MMDVM hotspot and KISS hosts also send at
any time; unchecked. A latched line doesn't count as PTT, and the LEDs flash red and green fast to ask for a power cycle
(gotchas.md 16.5). Checked, with TX inhibited, on a radio latched with debug link requests sent without waiting for the answers: the link died within a second, the LEDs alternated red and green, the PTT key did nothing, and a power cycle restored the link.
Cables by USB-serial chip, one cable each:
| Chip | RT10 |
|---|---|
| FTDI (K-type) | works: boot ROM load, stub, debug link, every HIL run |
| CH340 (K-type, Linux) | no answer to the debug link and no byte received, also after a replug and a power cycle; the same cable works with a DM-32UV |
| PL2303 (K-type, macOS) | the adapter stops sending (below); a PL2303 in a good state is not checked |
The RT10 needs an FTDI cable until another chip is shown to work.
A cable whose adapter stopped sending (gotchas.md 16.6), measured with one K-type PL2303 cable on macOS:
| Check | Result |
|---|---|
| Radio to host | works: the boot ROM's greeting arrives intact at 115200 baud |
| Host to radio | nothing: the ROM takes no answer and starts the firmware; RadioFoundry answers no debug link request |
| The host's port | takes about 1 KB and sends none of it, also with the cable pulled out of the radio |
| Break, DTR and RTS, flow control, baud rates, the port reopened | no change |
| The cable's USB end unplugged and plugged back in | sends again, until a port is closed with data still going out |
| The LEDs, that cable in | flash red and green fast from 2 s after the start; they stop when the cable is pulled out of the radio (no latch) |
| A power cycle without the cable | no flash |
4. Data flash (2 MB, SPI0)
SPI0, mode 0, chip select PTA18 by GPIO. Read 0x03, page program 0x02, 4 KB erase 0x20, status 0x05, write enable 0x06.
| Range | Content |
|---|---|
| 0x001300-0x031FFF | the original firmware's settings and channels (kept, never written by RadioFoundry) |
| 0x100000-0x1FDFFF | RadioFoundry: settings, journal, open-firmware codeplug images, native codeplug (boards/rt10/board.c) |
| 0x1FE000-0x1FEFFF | activation block tied to the SoC's eFuse ID: never written |
| 0x1FF000-0x1FFFFF | factory calibration (section 4.1): never written |
4.1 Factory calibration
Byte tables of 15 entries, one per 10 MHz step. The step for a frequency is (frequency / 10 MHz) minus the base byte,
held between 0 and 14. Offsets are from 0x1FF000. 0xFF is an entry the factory didn't fill. RadioFoundry copies the
tables at start-up (boards/rt10/calibration_rt10.c) and transmits only with a valid base.
| Offset | Content |
|---|---|
| 0x1A9 | the modem's TX center (SIG_CENTER), one byte: the TCXO's tune (below) |
| 0x470 | base: the first step's frequency / 10 MHz (90 on the radio checked; 80-95 accepted) |
| 0x51D / 0x52C | PA control level / 16 (DAC B), high / low power |
| 0x559 | 15 pairs: the first byte the modulation balance (the modem's MOD_BIAS), the second not used |
| 0x5B5 / 0x5C4 | FD6818 PA gain (0x28), high / low power; the high one is capped at 7 |
| 0x5D3 / 0x5F1 | key-up delay / 10 us, high / low power: RX supply off to the TX state |
| 0x600 | key-down delay / 10 us, before the ramp down |
The TX center and the step's modulation byte go into the modem's SIG_CENTER and RF_MOD_BIAS_CTRL at start-up and at
every receive set-up, as on the other HR_C7000 radios (boards/rt10/radio_rt10.c). The modem's MOD1 DC level is the
TCXO's control voltage, so these bytes set the frequency of receive and transmit alike. With the modem never powered,
the radio measured 28.2 ppm low: its carrier at 927.4738 MHz for 927.5, and its receiver's peak 25 kHz low. With them
written, the carrier measured +0.47 ppm (section 5) and the receiver peaks on the bench's frequency.
5. Power and transmitter
| Item | Value |
|---|---|
| Power latch | PTB12 high from start-up; low lets the radio turn off |
| Power switch | PTB17, high = off (measured). RadioFoundry lets go of the latch after five 100 ms polls read "off" (power_rt10.c) |
| Battery | ADC channel 2; scale ? |
| PA | DAC B drives the PA's power control; levels per frequency step come from the calibration (section 4.1) |
| Power levels | L and H: the calibration's low and high sets |
FM key-up, in order. Measured: the chip's TX state and frequency, the pin levels and the signal on air (927.5 MHz at L, the modem's bias written as in section 4.1: the carrier 436 Hz high, +0.47 ppm, on the SDR bench, whose own correction is good to about 0.5 ppm).
- The FD6818's TX words while it still receives (0x40 0x3AD0 with or without sub-audio, the sub-audio, PA gain 0x28 with the gain in both 3-bit fields), the speaker off, PTB3 and PTB0 low.
- The key-up delay, then PTB4 (RX supply) low.
- The FD6818 to the TX frequency, idle, then TX state 0xC1FE (0xC1FA for a tone).
- DAC B powered, PTB3 and PTB0 high, DAC B ramped to the calibration's level: 180 steps of 2 us on a smoothstep curve.
Key-down: the FD6818 back to receive, the key-down delay, the ramp down, DAC B 0, PTB0 and PTB3 low, DAC B powered down, 100 us, the receiver retuned, PTB4 high.
6. Keys and knob
Two side keys and PTT are direct inputs. The 16-position channel knob gives a 4-bit code on four inputs (bits 3..0 on PTB6, PTB1, PTB2, PTB21, read high = 1). The code is not the position in order and not a Gray code (measured, one position at a time from the counterclockwise stop):
| Position | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Code | 6 | 2 | 9 | 12 | 5 | 1 | 10 | 14 | 15 | 7 | 0 | 8 | 13 | 4 | 3 | 11 |
Between positions 3 and 4 the code read 13 (position 13) for up to about 100 ms (measured), so the firmware takes a
code once it has held 150 ms. The knob is read as a position, not as steps: position N selects the zone's channel N,
at power-on too (ui.md 3). The diagnostics page shows the raw code (knob raw), the debug link view the
position (input.switch_pos, 0-based).
7. FD6818 bus
A two-wire bus (I2C style) on PTA7 (clock) and PTA8 (data), device address 0xE2 write / 0xE3 read, 16-bit registers,
bit-banged (measured). The register map is the FD6818's (drivers/rf/fd6818/fd6818sh.h): state 0x03, page select
0x7F. On this bus page 1's registers are addressed from 0x80 (0x7F = 1, then register | 0x80).
| Item | Value |
|---|---|
| ID (0x00) | 0x6818 (measured) |
| States (0x03) | receive 0xBFF1 (reads back 0xBFF5), TX 0xC1FE, TX with a tone 0xC1FA; every change through 0 |
| Band | 902-928 MHz is the chip's band above 600 MHz: 0x1A = 0x1F80, 0x40 = 0x3AD0 (measured: -34 dBm read from a signal the band word 0xBF24 left at the floor) |
| Bandwidth (0x43) | 0x4048 at 12.5 kHz, 0x3028 at 25 kHz |
| Receive set | the power-on tables and the analog words in boards/rt10/radio_rt10.c |
| AF output (0x47) | 0x7140 receive, 0x7040 mute (TX) |
| GPIO | 0x25-0x27: GPIO0 the sub-audio compare, to PTB27; GPIO6 squelch, to PTB26 (measured) |
| Noise (0x65) | no signal: 38-44 raw at 904-927.5 MHz (measured), inside the core's squelch range, so the driver adds the FD6818 offset (fd6818_noise) |
Sub-audio decode (measured with CTCSS 100.0 Hz and DCS 023N): the right tone or code raises PTB27, a wrong one and no carrier leave it low. A match with CTCSS is held: PTB27 stays high after the carrier (9 s watched) and through the next carrier, even one with another tone (131.8 Hz after 100.0 Hz, 4.6 s). Writing the sub-audio word (0x51) again clears it, and a new match then takes about 120 ms. The driver counts PTB27 only with PTB26 (squelch) high, and writes the sub-audio set again when it finds PTB26 low with PTB27 high. The chip doesn't tell DCS polarity apart: 023N is heard with 023 set normal or inverted.
7.1 DMR
The HR_C7000's 4FSK goes on the TCXO through MOD1 (section 4.1), and each burst runs the FM key-up and key-down,
started by the modem's RF TX interrupt. RadioFoundry runs the bursts from the shared RF-burst task
(modem-hrc7000.md 3.1a) with the sequences in boards/rt10/radio_rt10.c. The values:
| Item | Value |
|---|---|
| FD6818 digital set | the DMR entry: 0x04 = 0x0300, 0x48 = 0x6C5C, 0x4B = 0xE442, 0x73 = 0x569A (AFC off), 0x7E = 0x349E (TX DC filter bypassed), 0x23 = 0x8F87 (AF DAC level 7), then 0x43 = 0x7E08 and 0x77 = 0xC0FF (the analog set writes 0x80EF). Without the 0x77 word the speaker hisses through a DMR key-up, with a pulse at the burst rate; with it the speaker is silent (listened to). The FD6818 guide's DMR bypass mode lists the same 0x48, 0x4B, 0x73 and 0x7E words ("cic mode"). Leaving DMR, the analog set writes the power-on words back (0x48 = 0x6C4C) |
| 0x4B | 0xE442 puts the demodulated 4FSK on the AF output. With the power-on 0x223C the modem's PHASE_OUT range stays within about ±40 counts of its center during a call (a few millivolts at its AF ADC) and the modem never syncs; with 0xE442 it reaches the ±2432 limit (measured) |
| AF output (0x47) | receive 0x4140 (to the modem; analog 0x7140); at a burst's TX 0x6740, written before the tune |
| 0x40 | 0x3AD0: the band select in every tune writes it, so a burst transmits with it |
| States (0x03) | receive 0xBFF1, TX 0xC1FE, each through 0 |
| Modem RF_MODE | 0x034C9860: AF gain 0x4C, TX symbol mapping not inverted (bit 7 clear) |
| RF TX interrupt lead | RF_CONTROL b13:8 = 0x1A (2.6 ms) |
| Modem ADC gain (GCR_ADCL) | 23 dB, listened to through a parrot. The other HR_C7000 radios use 11 dB. make RT10_DMR_MIC_DB=N |
| DAC C | the modem's AF ADC bias (its N input): each step moves the PHASE_OUT center about 9 counts. The set-up level is 1762; RadioFoundry writes RT10_DMR_DAC_C (1850) entering DMR and 1762 leaving it. AMBE bits corrected per frame on a bench voice call (234 frames each, measured): 1762 0.83-0.86, 1800 0.51, 1825 0.45, 1850 0.05, 1875 0.09, 1900 0.16, 1950 0.50, 2050 0.89; at 1680 no sync. make RT10_DMR_DAC_C=N |
Burst TX on, from the RF TX interrupt: the calibration's key-up delay (section 4.1), PTB4 (RX supply) low, the FD6818 to the TX frequency in its TX state, PTB3 and PTB0 high, DAC B ramped to the calibration's level. TX off: the ramp down, PTB0 and PTB3 low, 100 us, the FD6818 back to the receive frequency (0x4140, 0xBFF1), 100 us, PTB4 high. The calibration's key-down delay is counted from the modem's RF RX interrupt; RadioFoundry holds the TX until the burst's last symbol instead.
RadioFoundry's lead is not a fixed 2.6 ms: it is the shortest 100 us step that has the PA up for the first
symbol, 0.85 ms after the lead (rt10_dmr_rf_tx_lead, boards/rt10/rt10.h). After the RF TX interrupt plus the
calibration's key-up delay, the TX on sequence has the PA within 1 dB of the plateau at 2.53-2.66 ms and its ramp done
at 3.0-3.12 ms; the model takes 2.9 ms, which puts the 1 dB point at least 0.24 ms ahead of the first symbol. With the
low set's 450 us key-up delay at 927.5 MHz the lead is 2.5 ms, and the receiver goes off 2.9-3.0 ms before the first
symbol whatever the key-up delay. The ramp's 180 steps of 2 us take about 0.5 ms with the level's arithmetic, which
is 32-bit (core/radio/apc_ramp.h); a 64-bit divide per step made it 0.7 ms.
The key-up delay counts from the RF TX interrupt, not from the RF task's wake (up to 0.42 ms later), and a burst whose interrupt came late (up to 0.32 ms: the modem's interrupts are masked while the slot engine holds its lock) is timed from the 60 ms frame grid of the burst before (modem-hrc7000.md 3.1a). Before both, one or two bursts in 35 started their TX on 0.3-0.5 ms late, with the PA coming up in their first symbols. The first burst of a transmission has no grid yet: its interrupt comes about 0.25 ms late, which the lead's margin covers.
Measured (927.5 MHz, lowest power, 2 s key-ups, against the first symbol; PA within 1 dB of the plateau):
| Build | PA up | Carrier gone after the last symbol |
|---|---|---|
| 64-bit ramp, lead 2.6 ms | -0.31..-0.27 ms | +0.62..+0.69 ms |
| 32-bit ramp, lead 2.6 ms | -0.48..-0.40 ms | +0.23..+0.40 ms |
| 32-bit ramp, lead 2.3 ms | -0.12..+0.02 ms, late bursts to +0.35 ms | +0.12..+0.40 ms |
| Timed from the interrupt and the grid, lead 2.3 ms (3 key-ups, 114 bursts) | -0.23..-0.04 ms | median +0.41 ms |
| Timed from the interrupt and the grid, lead 2.5 ms (5 key-ups, 190 bursts) | -0.48..-0.30 ms; the first burst of a key-up -0.06..-0.02 ms | median +0.40 ms; the first burst of a key-up up to +0.73 ms |
Measured on the SDR bench (927.5 MHz, lowest power):
| Test | Result |
|---|---|
| DMR transmit, 6 s key-up, decoded by the bench | 106 bursts at 60.0 ms: 3 LC headers, 17 superframes, terminator; CC 2; group voice from the radio's ID; every FEC check passed, 0 AMBE bits corrected; deviation 1904-1913 Hz, frequency error within ±60 Hz. The diagnostics line: 106 keyed, 106 ended, none late |
| DMR receive, a bench group call of recorded speech (78 voice bursts, 234 AMBE frames) | the call decoded (source, talkgroup, group, CC 2); 78 of 78 voice bursts, 3 headers, 2 terminators, no CRC errors; 234 of 234 frames played, none concealed or lost, 4 and 5 bits corrected on two calls; PCM peak 16707 |
7.2 APRS
The FD6818 has no AFSK path of its own, so APRS goes through the HR_C7000 (platform/hrc7000/hrc_afsk.c):
| Direction | Path |
|---|---|
| Transmit | The FD6818 keys up in its TX state, then goes to its tone state with no tone (0xC1FA: carrier on, microphone ADC off, fd6818sh_mic_off). The HR_C7000's FM transmitter plays the AFSK from its prompt FIFO onto MOD1 (the TCXO, as DMR does). Ring play alone sent a bare carrier, so the tones take the codec mixer (hrc7000_codec_dac_to_tx, modem-hrc7000.md 8.4), which also keeps the microphone out. The FM limiter (fm_tx_limit) is 0x0F at both bandwidths; the default 4 held the peak to 1.66 kHz. Level: amp_per_level 55 in boards/rt10/radio_rt10.c |
| Receive | The HR_C7000 in FM receive copies the FD6818's AF output (0x47 = 0x7140) to the CPU (fm_rx_audio.c). The AF output stays on while APRS listens; the modem's FM_PLAY stays off |
Measured on air (927.5 MHz, lowest power): the frames decode with a valid FCS, 3.0 kHz peak deviation at the default level, 2.8 dB less at the 1200 Hz tone than at the 2200 Hz tone. Receive is not yet measured on air.
7.3 Air interface (CONFIG_P25, CONFIG_NXDN)
board.air and board.baseband are the HR_C7000's provider and receive tap (modem-hrc7000.md 10),
under CONFIG_DMR. A session runs in DMR: the FD6818 keeps its DMR set-up (7.1), so the receive copy carries the 4FSK
the modem's AF ADC sees, and a transmission keys the RF through the burst sequences of 7.1 once for its whole length.
The radio has no VFO, so the modes run on stored channels. Checked on air at 927.5 MHz against the SDR bench, lowest
power, the radio on its antenna:
| What | Test | Result |
|---|---|---|
| P25 transmit on a stored channel | test_p25_stored_channel_tx |
decoded off air: the channel's NAC, talk group and unit ID |
| P25 receive | test_p25_scan_from_bench (zone, dual) |
a bench call stops a zone scan and dual watch on the P25 channel |
| P25 trunked transmit (p25-trunking.md) | test_p25_trunk_tx |
control channel at 927.0, inbound 927.5: registration, affiliation, a voice request, and a call on the granted voice channel (927.25), twice |
| NXDN transmit, 4800 bit/s, stored channel | test_nxdn_stored_channel_tx |
decoded off air: the channel's RAN, unit and group |
| NXDN receive, 4800 bit/s | test_nxdn_scan_from_bench (zone, dual) |
a bench call stops a zone scan and dual watch on the NXDN channel |
Not checked: NXDN at 9600 bit/s (the stored-channel tests use 4800), P25 receive frame counts (that test runs in the VFO), P25 trunked receive (needs two bench transmitters), a radio of another make, a live trunked system, sensitivity.
8. Codec
The AMBE+2 vocoder is part of the original image: its code runs from RAM, its tables are copied to modem SRAM. No
vendor file is published, so the codec comes from a read of the radio's program flash (V1.1.2, the build name in the
image). RadioFoundry hosts it in the codec slot like the other HR_C7000 radios (boards/rt10/memory.ld): the firmware
copies the code to the codec window and moves the words its codec map lists (codec-placement.md).
No codec bytes are in the repository or in a release.
| Item | Value |
|---|---|
| Code start (the two signature words) | flash 0x0305E438 |
| Code | 0x12FFC bytes (77,820), sha256 3f3b8740df14c81b44f53f984e4e570a609ec467233341729157f5c176cc3c15. The window ends where memcpy starts |
| Code run address | 0x16DD8: the start-up code copies flash 0x03057A60 to RAM 0x10400..0x2E324 |
| Compile | The DB25-D 020 compile, with one function inside the window (at +0x614) 20 bytes shorter: the codec code past it sits 20 bytes lower. The same 153 functions are reachable from the entry points |
| Entry points (offsets) | decode 0, encode 0x294, dec_init 0xE8, dec_set 0xA4, enc_init 0x400, enc_set 0x32C, fec_init 0x1CD0 |
| Init | The order, arguments and state offsets platform/hrc7000/vocoder.c uses |
| Tables | flash 0x0303C76C, 0x2D50 bytes (11,600), sha256 e1e435822471894c437048396f2ea96bea5803acc08446b8f1073714978e219e (the DB25-D's tables). The start-up code copies flash 0x03038D60 to modem SRAM 0x18000800, so they run at 0x1800420C..0x18006F5C |
| Codec map | rt10-112 (tools/rfoundry/rfoundry/codec_maps/rt10/): 3 self-referencing words (one reachable from the entry points), 69 table words (52 targets), no instruction pairs, no references outside the code and table windows. Measured: the merged codec decodes bench speech and encodes the radio's transmissions (7.1) |
memory.ld reserves modem SRAM 0x1800420C..0x18007184, 0x228 bytes more than the tables. tools/hrc7000_codec.py --board rt10 --stock READ merges the codec into a built image; READ is a 1 MB or the whole 2 MB program flash read
(stub/stubtool.py --progmem), and the codec is found by its hashes. The slot's stock tag is 112.
9. Bring-up state
| Item | State |
|---|---|
| Boot from the ROM flag at 0x0300C000 | written |
| Host link on UART0, stub entry, watchdog, crash log | measured: the hang and stall checks keep their records and the radio restarts by itself (test_crash.py). The restart is a watchdog reset: a reboot answers again after 3 s, and every crash and watchdog test passes through it (test_crash.py, test_watchdog.py, 12 of 12) |
| Battery scale, key levels, knob map | open |
| FD6818 two-wire bus, FM receive, sub-audio decode | measured |
| Factory calibration read | measured (the radio's own sector) |
| FM transmit | measured on air (section 5) |
| Battery save | measured: asleep is the FD6818 idle and PTB4 (RX supply) low; waking retunes the receiver |
| Reference | measured: the modem's bias from the calibration puts receive and transmit on frequency (section 4.1) |
| Reference trim | written: the trim moves the synthesizer frequency at the next tune; the modem's bias stays the factory's tune. Untested on air |
| DMR transmit and receive, codec | measured on air (7.1): voice both ways with the codec map rt10-112 (section 8) |
| DMR microphone | measured: a 7 s spoken key-up, received on the SDR bench, decoded as speech (351 of 360 AMBE frames distinct, the parameters spread as speech spreads them) |
| Speaker | listened to: the amplifier is PTB8 (section 3); FM monitor, the talk permit and deny tones and a parrot reply play, FM squelch stays closed after an unkey, no hiss in a DMR key-up |
| IP data over the air | measured on air: UDP to and from the radio, compressed and full headers, off-net, confirmed, bursts from the radio. The IP gateway is off with PPP (3.1) |
| IP link to the computer | written: PPP on the host link (UART0, 115200 baud), off: pppd sends while the radio sends, which latches the cable line low until a power cycle (3.1). It builds with CONFIG_NET_PPP=y, for a check with another cable |
| P25, P25 trunked, NXDN | measured on air (7.3): P25 both ways and trunked transmit; NXDN both ways at 4800 bit/s. NXDN at 9600 bit/s and P25 trunked receive untested |
| APRS | written (7.2): the radio keys up and sends its frames; deviation, twist and decoding on air not yet measured |
| Hotspot | written, off: the MMDVM host sends at any time, as pppd does, so it may latch the cable line (3.1). It builds with CONFIG_HOTSPOT=y, for that check |
| Hotspot split, USRP | not written: no continuous base-station outbound; USRP's 8 kHz audio both ways needs more than 115200 baud |
| Screen-only features | left out: no display (boards/rt10/config.mk) |
| Microphone through the programming cable's line | measured: barely couples. FM key-ups on the SDR bench with 0x55 bursts at 6000 and 2400 baud on the line: the audio rose from about 5 Hz rms deviation to 6-8 Hz rms during the bursts, no tone. tests/hil/test_mic_uart_inject.py leaves the RT10 out |