Decoding Time Data

In my previous experiment I introduced the second decoder. This decoder locks to the sync mark phase and thus allows to infer the seconds. It is also the key to get hold of the bit stream that is encoded in the DCF77 signal. Thus I now tackle the issue how to decode the minutes and the hours. I will still ignore the issue with the leap seconds.

From an architecture point of view this means I just add two more decoder boxes to the code.

Decoding Hours and Minutes

So now the reason for introducing the controller in the previous experiment becomes clearer. The controller will control all the decoder boxes. But how do the new decoders work? There are actually two ways to explain their inner mechanics. As the decoders are quite tricky I will explain both approaches.

I will start with the approach that I actually implemented. I will explain it for the hour decoder although the minute decoder works the same way. The point is that the hour decoder only needs to determine 1 out 24 values (instead of 1 out of 60).

The core of the implementation consists of small set of functions.

void process_tick(const uint8_t current_second, const uint8_t tick_value)

template <typename bins_type, uint8_t significant_bits, bool with_parit>
void hamming_binning(bins_type &bins, const BCD::bcd_t input)

template <typename bins_t>
void compute_max_index(bins_t &bins)

template <typename bins_t>
void advance_tick(bins_t &bins)

Lets start with the simplest one. The process_tick function will be called each second. It gets the current second value and the transmitted data. Since it knows it is looking for hours it will collect the hour data bits (with parity). Once it has collected all bits of a minute it will call the hamming_binning function for the real work.

The hamming_binning function then does the heavy lifting. In the hours case it will be called as

hamming_binning<hour_bins, 7, true>(bins, hour_data);

You might want to verify the declaration of the bins as

typedef struct {
    uint8_t data[hours_per_day];
    uint8_t tick;

    uint8_t noise_max;
    uint8_t max;
    uint8_t max_index;
} hour_bins;

Lets have a closer look at its signature. It is a template function. If you are not familiar with templates the technique is not very hard to understand. Basically it introduces compile time parameters. That is the function parameters in the angle brackets will be resolved at compile time instead of run time. Therefore it is admissible to pass types as parameters (which is impossible at run time). It follows that the parameters will be hard wired into the compiled code. Hence a template will give rise to a different piece of code if the parameters in the angle brackets are varied. This is a classical time / memory trade off. The resulting code will usually execute faster but consume more memory. The most important point here is that passing the bins_type parameter allows to infer the number of bins at compile time.

Right now the important detail is that the function will be aware of the number of bins. In this case it knows that the number of bins (=the size of the data array) is “hours_per_day (=24).

Inside the function will vary a “candidate” value for all admissible values. In our case from 0 to 23. It will encode this candidate in BCD. If necessary it will also compute the parity. Then the hamming distance of the candidate and the received data is computed. Actually it is not the distance but something I called score. This score is just the number of matching bits of the candidate vs. the data. This is added to the corresponding bin.

So for each candidate hour we collect the number of bits that are matching. For the correct hour we should thus see sooner or later that it stands out somehow. Time for an example. Suppose the received hour data would be for hour 5 with proper parity: 0b00000101. Then the binning would result in the following values.

Candidate Candidate Data Data Score
0 0b00000000 0b00000101 5
1 0b10000001 0b00000101 6
2 0b10000010 0b00000101 4
3 0b00000011 0b00000101 5
4 0b10000100 0b00000101 6
5 0b00000101 0b00000101 7
6 0b00000110 0b00000101 5
7 0b10000111 0b00000101 6
8 0b10001000 0b00000101 4
9 0b00001001 0b00000101 5
10 0b00010000 0b00000101 3
11 0b10010001 0b00000101 4
12 0b00010010 0b00000101 5
13 0b10010011 0b00000101 6
14 0b10010100 0b00000101 4
15 0b00010101 0b00000101 5
16 0b10010110 0b00000101 4
17 0b00010111 0b00000101 5
18 0b00011000 0b00000101 3
19 0b10011001 0b00000101 4
20 0b00100000 0b00000101 5
21 0b10100001 0b00000101 6
22 0b10100010 0b00000101 4
23 0b00100011 0b00000101 5

So “5” stands out slightly. Each minute this will get more and more pronounced. The nice thing is: even in the presence of noise this works. Of course for each and every minute we might get garbled values. However the proper candidate will have more matching bits on average than any other candidate.

With other words the index of the best candidate is the proper decoded hour and we can infer it with the compute_max_index function. In addition we also get some estimate for the next buest guess which allows an estimate of the overall quality of the retrieved data.

There is one issue with this approach though. What happens if the DCF77 signal advances one hour?

This is where the advance_tick function comes into play. Whenever this function is called the “tick” value of the bin structure is incremented (modulo the number of bins). Whenever scores are distributed to the bins the tick will be considered as well. If the tick is zero the scores go to the corresponding bins. However if it is larger than zero then the scores will be offset. For example for tick == 1 the score for the candidate value 6 would be added to bin 5. And for tick == 2 the score of candidate 7 would be added to bin 5. Hence it is possible to cummulate scores over several hours.

Exactly the same approach is used for minutes, hours, days, weekdays, months and years. The only difference is the number of bins plus the offset (that is if a value starts counting at 0 (like for the minutes) or 1 (like for the days).

As I promised there is another way to interpret the decoder’s action. Instead of considering 24 bins one might focus more on the associated bit streams. If the bits that belong to a candidate value are repeated 60 times (because an hour has 60 minutes) and concatenated, then the scoring is mathmatically equivalent to computing the convolution of the input signal with this test data. It follows that this is a matched filter. This kind of filter is probably the best that you can get. Also it has been studied extensively in the literature. For example in my favourite DSP book you can find it chapter 17 (custom filters / optimal filters).

//
//  www.blinkenlight.net
//
//  Copyright 2013 Udo Klein
//
//  This program is free software: you can redistribute it and/or modify
//  it under the terms of the GNU General Public License as published by
//  the Free Software Foundation, either version 3 of the License, or
//  (at your option) any later version.
//
//  This program is distributed in the hope that it will be useful,
//  but WITHOUT ANY WARRANTY; without even the implied warranty of
//  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
//  GNU General Public License for more details.
//
//  You should have received a copy of the GNU General Public License
//  along with this program. If not, see http://www.gnu.org/licenses/


//#define DEBUG 1


const uint8_t first_output_pin = 5;

namespace Debug {
    void debug_helper(char data);
    void bcddigit(uint8_t data);
    void bcddigits(uint8_t data);
}

namespace Hamming {
    typedef struct {
        uint8_t lock_max;
        uint8_t noise_max;
    } lock_quality_t;
}

namespace BCD {
    typedef union {
        struct {
            uint8_t lo:4;
            uint8_t hi:4;
        } digit;

        struct {
            uint8_t b0:1;
            uint8_t b1:1;
            uint8_t b2:1;
            uint8_t b3:1;
            uint8_t b4:1;
            uint8_t b5:1;
            uint8_t b6:1;
            uint8_t b7:1;
        } bit;

        uint8_t val;
    } bcd_t;

    void increment(bcd_t &value);

    bcd_t int_to_bcd(const uint8_t value);
    uint8_t bcd_to_int(const uint8_t value);
}

typedef union {
    struct {
        uint8_t b0:2;
        uint8_t b1:2;
        uint8_t b2:2;
        uint8_t b3:2;
    } signal;
    uint8_t byte;
} tData;

namespace DCF77 {
    const uint8_t long_tick  = 3;
    const uint8_t short_tick = 2;
    const uint8_t undefined  = 1;
    const uint8_t sync_mark  = 0;

    typedef struct {
        BCD::bcd_t hour;     // 0..23
        BCD::bcd_t minute;   // 0..59
        uint8_t second;      // 0..60

        boolean undefined_minute_output;
    } time_data;
}

namespace DCF77_Encoder {

    void reset(DCF77::time_data &now);
    void advance_second(DCF77::time_data &now);

    uint8_t get_current_signal(const DCF77::time_data &now);

    void debug(const DCF77::time_data &clock);
    void debug(const DCF77::time_data &clock, const uint16_t cycles);

    // Bit      Bezeichnung     Wert    Pegel   Bedeutung
    // 0        M                       0       Minutenanfang (

    // 1..14    n/a                             reserviert

    // 15       R                               Reserveantenne aktiv (0 inaktiv, 1 aktiv)
    // 16       A1                              Ankündigung Zeitzonenwechsel (1 Stunde vor dem Wechsel für 1 Stunde, d.h ab Minute 1)
    // 17       Z1               2              Zeitzonenbit Sommerzeit (MEZ = 0, MESZ = 1); also Zeitzone = UTC + 2*Z1 + Z2
    // 18       Z2               1              Zeitzonenbit Winterzeit (MEZ = 1, MESZ = 0); also Zeitzone = UTC + 2*Z1 + Z2
    // 19       A2                              Ankündigung einer Schaltsekunde (1 Stunde vor der Schaltsekunde für 1 Stunde, d.h. ab Minute 1)

    // 20       S                       1       Startbit für Zeitinformation

    // 21                        1              Minuten  1er
    // 22                        2              Minuten  2er
    // 23                        4              Minuten  4er
    // 24                        8              Minuten  8er
    // 25                       10              Minuten 10er
    // 26                       20              Minuten 20er
    // 27                       40              Minuten 40er
    // 28       P1                              Prüfbit 1 (gerade Parität)

    // 29                        1              Stunden  1er
    // 30                        2              Stunden  2er
    // 31                        4              Stunden  4er
    // 32                        8              Stunden  8er
    // 33                       10              Stunden 10er
    // 34                       20              Stunden 20er
    // 35       P2                              Prüfbit 2 (gerade Parität)

    // 36                        1              Tag  1er
    // 37                        2              Tag  2er
    // 38                        4              Tag  4er
    // 39                        8              Tag  8er
    // 40                       10              Tag 10er
    // 41                       20              Tag 20er

    // 42                        1              Wochentag 1er (Mo = 1, Di = 2, Mi = 3,
    // 43                        2              Wochentag 2er (Do = 4, Fr = 5, Sa = 6,
    // 44                        4              Wochentag 4er (So = 7)

    // 45                        1              Monat  1er
    // 46                        2              Monat  2er
    // 47                        4              Monat  4er
    // 48                        8              Monat  8er
    // 49                       10              Monat 10er

    // 50                        1              Jahr  1er
    // 51                        2              Jahr  2er
    // 52                        4              Jahr  4er
    // 53                        8              Jahr  8er
    // 54                       10              Jahr 10er
    // 55                       20              Jahr 20er
    // 56                       40              Jahr 40er
    // 57                       80              Jahr 80er

    // 58       P3                              Prüftbit 3 (gerade Parität)

    // 59       sync                            Sync Marke, kein Impuls (übliches Minutenende)
    // 59                               0       Schaltsekunde (sehr selten, nur nach Ankündigung)
    // 60       sync                            Sync Marke, kein Impuls (nur nach Schaltsekunde)

    // Falls eine Schaltsekunde eingefügt wird, wird bei Bit 59 eine Sekundenmarke gesendet.
    // Der Syncimpuls erfolgt dann in Sekunde 60 statt 59. Üblicherweise wird eine 0 als Bit 59 gesendet

    // Üblicherweise springt die Uhr beim Wechsel Winterzeit nach Sommerzeit von 1:59:59 auf 3:00:00
    //                               beim Wechsel Sommerzeit nach Winterzeit von 2:59:59 auf 2:00:00

    // Die Zeitinformation wird immer 1 Minute im Vorraus übertragen. D.h. nach der Syncmarke hat
    // man die aktuelle Zeit

    // http://www.dcf77logs.de/SpecialFiles.aspx

    // Schaltsekunden werden in Deutschland von der Physikalisch-Technischen Bundesanstalt festgelegt,
    // die allerdings dazu nur die international vom International Earth Rotation and Reference Systems
    // Service (IERS) festgelegten Schaltsekunden übernimmt. Im Mittel sind Schaltsekunden etwa alle 18
    // Monate nötig und werden vorrangig am 31. Dezember oder 30. Juni, nachrangig am 31. März oder
    // 30. September nach 23:59:59 UTC (also vor 1:00 MEZ bzw. 2:00 MESZ) eingefügt. Seit der Einführung
    // des Systems 1972 wurden ausschließlich die Zeitpunkte im Dezember und Juni benutzt.
}

namespace DCF77_Hour_Decoder {
    void setup();
    void process_tick(const uint8_t current_second, const uint8_t tick_value);
    void advance_hour();
    BCD::bcd_t get_hour();
    void get_quality(Hamming::lock_quality_t &lock_quality);

    void debug();
}

namespace DCF77_Minute_Decoder {
    void setup();
    void process_tick(const uint8_t current_second, const uint8_t tick_value);
    void advance_minute();
    BCD::bcd_t get_minute();

    void debug();
}

namespace DCF77_Second_Decoder {
    void setup();
    void process_single_tick_data(const uint8_t tick_data);
    uint8_t get_second();
    void get_quality(Hamming::lock_quality_t &lock_quality);

    void debug();
}

namespace DCF77_Clock_Controller {
    void setup();
    void process_single_tick_data(const uint8_t tick_data);

    void flush();
    typedef void (*flush_handler)(const DCF77::time_data &decoded_time);

    void default_handler(const DCF77::time_data &decoded_time);
    void set_flush_handler(const flush_handler output_handler);

    typedef Hamming::lock_quality_t lock_quality_t;

    typedef struct {
        struct {
            uint32_t lock_max;
            uint32_t noise_max;
        } phase;

        lock_quality_t second;
        lock_quality_t minute;
        lock_quality_t hour;
    } clock_quality_t;

    void get_quality(clock_quality_t &clock_quality);

    // blocking, will unblock at the start of the second
    void get_current_time(DCF77::time_data &now);
}

namespace DCF77_Demodulator {
    void setup();
    void detector(const uint8_t sampled_data);
    void get_quality(uint32_t &lock_max, uint32_t &noise_max);
    void get_noise_indicator(uint32_t &noise_indicator);

    void debug();
}

namespace Debug {
    void debug_helper(char data) { Serial.print(data == 0? 'S': data == 1? '?': data - 2 + '0', 0); }

    void bcddigit(uint8_t data) {
        if (data <= 0x09) {
            Serial.print(data, HEX);
        } else {
            Serial.print('?');
        }
    }

    void bcddigits(uint8_t data) {
        bcddigit(data >>  4);
        bcddigit(data & 0xf);
    }
}

namespace BCD {
    void increment(bcd_t &value) {
        if (value.digit.lo < 9) {
            ++value.digit.lo;
        } else {
            value.digit.lo = 0;

            if (value.digit.hi < 9) {
                ++value.digit.hi;
            } else {
                value.digit.hi = 0;
            }
        }
    }

    bcd_t int_to_bcd(const uint8_t value) {
        const uint8_t hi = value / 10;

        bcd_t result;
        result.digit.hi = hi;
        result.digit.lo = value-10*hi;

        return result;
    }

    uint8_t bcd_to_int(const bcd_t value) {
        return value.digit.lo + 10*value.digit.hi;
    }
}

namespace Arithmetic_Tools {
    template <uint8_t N> inline void bounded_increment(uint8_t &value) __attribute__((always_inline));
    template <uint8_t N>
    void bounded_increment(uint8_t &value) {
        if (value >= 255 - N) { value = 255; } else { value += N; }
    }

    template <uint8_t N> inline void bounded_decrement(uint8_t &value) __attribute__((always_inline));
    template <uint8_t N>
    void bounded_decrement(uint8_t &value) {
        if (value <= N) { value = 0; } else { value -= N; }
    }

    inline void bounded_add(uint8_t &value, const uint8_t amount) __attribute__((always_inline));
    void bounded_add(uint8_t &value, const uint8_t amount) {
        if (value >= 255-amount) { value = 255; } else { value += amount; }
    }

    inline void bounded_sub(uint8_t &value, const uint8_t amount) __attribute__((always_inline));
    void bounded_sub(uint8_t &value, const uint8_t amount) {
        if (value <= amount) { value = 0; } else { value -= amount; }
    }

    inline uint8_t bit_count(const uint8_t value) __attribute__((always_inline));
    uint8_t bit_count(const uint8_t value) {
        const uint8_t tmp1 = (value & 0b01010101) + ((value>>1) & 0b01010101);
        const uint8_t tmp2 = (tmp1  & 0b00110011) + ((tmp1>>2) & 0b00110011);
        return (tmp2 & 0x0f) + (tmp2>>4);
    }

    inline uint8_t parity(const uint8_t value) __attribute__((always_inline));
    uint8_t parity(const uint8_t value) {
        uint8_t tmp = value;

        tmp = (tmp & 0xf) ^ (tmp >> 4);
        tmp = (tmp & 0x3) ^ (tmp >> 2);
        tmp = (tmp & 0x1) ^ (tmp >> 1);

        return tmp;
    }
}

namespace Hamming {
    template <uint8_t significant_bits>
    void score (uint8_t &bin, const BCD::bcd_t input, const BCD::bcd_t candidate) {
        using namespace Arithmetic_Tools;

        const uint8_t the_score = significant_bits - bit_count(input.val ^ candidate.val);
        bounded_add(bin, the_score);
    }

    template <typename bins_t>
    void advance_tick(bins_t &bins) {
        const uint8_t number_of_bins = sizeof(bins.data) / sizeof(bins.data[0]);
        if (bins.tick < number_of_bins - 1) {
            ++bins.tick;
        } else {
            bins.tick = 0;
        }
    }

    template <typename bins_type, uint8_t significant_bits, bool with_parity>
    void hamming_binning(bins_type &bins, const BCD::bcd_t input) {
        using namespace Arithmetic_Tools;
        using namespace BCD;

        const uint8_t number_of_bins = sizeof(bins.data) / sizeof(bins.data[0]);

        if (bins.max > 255-significant_bits) {
            // If we know we can not raise the maximum any further we
            // will lower the noise floor instead.
            for (uint8_t bin_index = 0; bin_index <number_of_bins; ++bin_index) {
                bounded_decrement<significant_bits>(bins.data[bin_index]);
            }
            bins.max -= significant_bits;
            bounded_decrement<significant_bits>(bins.noise_max);
        }

        const uint8_t offset = number_of_bins-1-bins.tick;
        uint8_t bin_index = offset;

        bcd_t candidate;
        candidate.val = 0x00;
        for (uint8_t pass=0; pass < number_of_bins; ++pass) {

            if (with_parity) {
                candidate.bit.b7 = parity(candidate.val);
                score<significant_bits>(bins.data[bin_index], input, candidate);
                candidate.bit.b7 = 0;
            } else {
                score<significant_bits>(bins.data[bin_index], input, candidate);
            }

            bin_index = bin_index < number_of_bins-1? bin_index+1: 0;
            increment(candidate);
        }
    }

    template <typename bins_t>
    void compute_max_index(bins_t &bins) {
        const uint8_t number_of_bins = sizeof(bins.data) / sizeof(bins.data[0]);

        bins.noise_max = 0;
        bins.max = 0;
        bins.max_index = 255;
        for (uint8_t index = 0; index < number_of_bins; ++index) {
            const uint8_t bin_data = bins.data[index];

            if (bin_data >= bins.max) {
                bins.noise_max = bins.max;
                bins.max = bin_data;
                bins.max_index = index;
            } else if (bin_data > bins.noise_max) {
                bins.noise_max = bin_data;
            }
        }
    }

    template <typename bins_t>
    void setup(bins_t &bins) {
        const uint8_t number_of_bins = sizeof(bins.data) / sizeof(bins.data[0]);

        for (uint8_t index = 0; index < number_of_bins; ++index) {
            bins.data[index] = 0;
        }
        bins.tick = 0;

        bins.max = 0;
        bins.max_index = 255;
        bins.noise_max = 0;
    }

    template <typename bins_t>
    BCD::bcd_t get_time_value(const bins_t &bins) {
        // there is a trade off involved here:
        //    low threshold --> lock will be detected earlier
        //    low threshold --> if lock is not clean output will be garbled
        //    a proper lock will fix the issue
        //    the question is: which start up behaviour do we prefer?
        const uint8_t threshold = 2;

        const uint8_t number_of_bins = sizeof(bins.data) / sizeof(bins.data[0]);
        const uint8_t offset = (number_of_bins == 60 || number_of_bins == 24 || number_of_bins == 100)? 0x00: 0x01;

        if (bins.max-bins.noise_max >= threshold) {
            return BCD::int_to_bcd((bins.max_index + bins.tick + 1) % number_of_bins + offset);
        } else {
            BCD::bcd_t undefined;
            undefined.val = 0xff;
            return undefined;
        }
    }

    template <typename bins_t>
    void get_quality(const bins_t bins, Hamming::lock_quality_t &lock_quality) {
        lock_quality.lock_max = bins.max;
        lock_quality.noise_max = bins.noise_max;
    }

    template <typename bins_t>
    void debug (const bins_t &bins) {
        const uint8_t number_of_bins = sizeof(bins.data) / sizeof(bins.data[0]);
        const bool uses_integrals = sizeof(bins.max) == 4;

        Serial.print(get_time_value(bins).val, HEX);
        Serial.print(F(" Tick: "));
        Serial.print(bins.tick);
        Serial.print(F(" Quality: "));
        Serial.print(bins.max, DEC);
        Serial.print('-');
        Serial.print(bins.noise_max, DEC);
        Serial.print(F(" Max Index: "));
        Serial.print(bins.max_index, DEC);
        Serial.print('>');

        for (uint8_t index = 0; index < number_of_bins; ++index) {
            if (index == bins.max_index ||
                (!uses_integrals && index == (bins.max_index+1) % number_of_bins) ||
                (uses_integrals && (index == (bins.max_index+10) % number_of_bins ||
                                   (index == (bins.max_index+20) % number_of_bins)))) {
                Serial.print('|');
            }
            Serial.print(bins.data[index],HEX);
        }
        Serial.println();
    }
}

namespace DCF77_Encoder {
    using namespace DCF77;

    void reset(DCF77::time_data &now) {
        now.second     = 0;
        now.minute.val = 0x00;
        now.hour.val   = 0x00;

        now.undefined_minute_output = false;
    }
    void advance_second(DCF77::time_data &now) {
        // in case some value is out of range it will not be advanced
        // this is on purpose
        if (now.second < 59) {
            ++now.second;
        } else {
            now.second = 0;

            if (now.minute.val < 0x59) {
                increment(now.minute);
            } else if (now.minute.val == 0x59) {
                now.minute.val = 0x00;

                if (now.hour.val < 0x23) {
                    increment(now.hour);
                } else if (now.hour.val == 0x23) {
                    now.hour.val = 0x00;
                }
            }
        }
    }

    uint8_t get_current_signal(const DCF77::time_data &now) {
        using namespace Arithmetic_Tools;

        if (now.second >= 1 && now.second <= 14) {
            // weather data or other stuff we can not compute
            return undefined;
        }

        bool result;
        switch (now.second) {
            case 0:  // start of minute
                return short_tick;

            case 20:  // start of time information
                return long_tick;

            case 21:
                if (now.undefined_minute_output || now.minute.val > 0x59) { return undefined; }
                result = now.minute.digit.lo & 0x1; break;
            case 22:
                if (now.undefined_minute_output || now.minute.val > 0x59) { return undefined; }
                result = now.minute.digit.lo & 0x2; break;
            case 23:
                if (now.undefined_minute_output || now.minute.val > 0x59) { return undefined; }
                result = now.minute.digit.lo & 0x4; break;
            case 24:
                if (now.undefined_minute_output || now.minute.val > 0x59) { return undefined; }
                result = now.minute.digit.lo & 0x8; break;

            case 25:
                if (now.undefined_minute_output || now.minute.val > 0x59) { return undefined; }
                result = now.minute.digit.hi & 0x1; break;
            case 26:
                if (now.undefined_minute_output || now.minute.val > 0x59) { return undefined; }
                result = now.minute.digit.hi & 0x2; break;
            case 27:
                if (now.undefined_minute_output || now.minute.val > 0x59) { return undefined; }
                result = now.minute.digit.hi & 0x4; break;

            case 28:
                if (now.undefined_minute_output || now.minute.val > 0x59) { return undefined; }
                result = parity(now.minute.val); break;

            case 29:
                if (now.hour.val > 0x23) { return undefined; }
                result = now.hour.digit.lo & 0x1; break;
            case 30:
                if (now.hour.val > 0x23) { return undefined; }
                result = now.hour.digit.lo & 0x2; break;
            case 31:
                if (now.hour.val > 0x23) { return undefined; }
                result = now.hour.digit.lo & 0x4; break;
            case 32:
                if (now.hour.val > 0x23) { return undefined; }
                result = now.hour.digit.lo & 0x8; break;

            case 33:
                if (now.hour.val > 0x23) { return undefined; }
                result = now.hour.digit.hi & 0x1; break;
            case 34:
                if (now.hour.val > 0x23) { return undefined; }
                result = now.hour.digit.hi & 0x2; break;

            case 35:
                if (now.hour.val > 0x23) { return undefined; }
                result = parity(now.hour.val); break;
            case 59:
                return sync_mark;

            default:
                return undefined;
        }

        return result? long_tick: short_tick;
    }

    void debug(const DCF77::time_data &clock) {
        using namespace Debug;

        bcddigits(clock.hour.val);
        Serial.print(':');
        bcddigits(clock.minute.val);
        Serial.print(':');
        if (clock.second < 10) {
            Serial.print('0');
        }
        Serial.println(clock.second, DEC);
   }

    void debug(const DCF77::time_data &clock, const uint16_t cycles) {
        DCF77::time_data local_clock = clock;
        DCF77::time_data decoded_clock;

        Serial.print(F("M ?????????????? RAZZA S mmmmMMMP hhhhHHP ddddDD www mmmmM yyyyYYYYP S"));
        for (uint16_t second = 0; second < cycles; ++second) {
            switch (local_clock.second) {
                case  0: Serial.println(); break;
                case  1: case 15: case 20: case 21: case 29:
                case 36: case 42: case 45: case 50: case 59: Serial.print(' ');
            }

            const uint8_t tick_data = get_current_signal(local_clock);
            Debug::debug_helper(tick_data);
       }

        Serial.println();
        Serial.println();
    }
}
namespace DCF77_Hour_Decoder {
    const uint8_t hours_per_day = 24;

    typedef struct {
        uint8_t data[hours_per_day];
        uint8_t tick;

        uint8_t noise_max;
        uint8_t max;
        uint8_t max_index;
    } hour_bins;

    hour_bins bins;


    void advance_hour() {
        Hamming::advance_tick(bins);
    }

    void process_tick(const uint8_t current_second, const uint8_t tick_value) {
        using namespace Hamming;

        static BCD::bcd_t hour_data;

        switch (current_second) {
            case 29: hour_data.val +=      tick_value; break;
            case 30: hour_data.val +=  0x2*tick_value; break;
            case 31: hour_data.val +=  0x4*tick_value; break;
            case 32: hour_data.val +=  0x8*tick_value; break;
            case 33: hour_data.val += 0x10*tick_value; break;
            case 34: hour_data.val += 0x20*tick_value; break;
            case 35: hour_data.val += 0x80*tick_value;        // Parity !!!
                    hamming_binning<hour_bins, 7, true>(bins, hour_data); break;

            case 36: compute_max_index(bins);
                    // fall through on purpose
            default: hour_data.val = 0;
        }
    }

    void get_quality(Hamming::lock_quality_t &lock_quality) {
        Hamming::get_quality(bins, lock_quality);
    }

    BCD::bcd_t get_hour() {
        return Hamming::get_time_value(bins);
    }

    void setup() {
        Hamming::setup(bins);
    }

    void debug() {
        Serial.print(F("Hour: "));
        Hamming::debug(bins);
    }
}

namespace DCF77_Minute_Decoder {
    const uint8_t minutes_per_hour = 60;

    typedef struct {
        uint8_t data[minutes_per_hour];
        uint8_t tick;

        uint8_t noise_max;
        uint8_t max;
        uint8_t max_index;
    } minute_bins;

    minute_bins bins;

    void advance_minute() {
        Hamming::advance_tick(bins);
    }

    void process_tick(const uint8_t current_second, const uint8_t tick_value) {
        using namespace Hamming;

        static BCD::bcd_t minute_data;

        switch (current_second) {
            case 21: minute_data.val +=      tick_value; break;
            case 22: minute_data.val +=  0x2*tick_value; break;
            case 23: minute_data.val +=  0x4*tick_value; break;
            case 24: minute_data.val +=  0x8*tick_value; break;
            case 25: minute_data.val += 0x10*tick_value; break;
            case 26: minute_data.val += 0x20*tick_value; break;
            case 27: minute_data.val += 0x40*tick_value; break;
            case 28: minute_data.val += 0x80*tick_value;        // Parity !!!
                    hamming_binning<minute_bins, 8, true>(bins, minute_data); break;
            case 29: compute_max_index(bins);
            // fall through on purpose
            default: minute_data.val = 0;
        }
    }

    void setup() {
        Hamming::setup(bins);
    }

    void get_quality(Hamming::lock_quality_t &lock_quality) {
        Hamming::get_quality(bins, lock_quality);
    }

    BCD::bcd_t get_minute() {
        return Hamming::get_time_value(bins);
    }

    void debug() {
        Serial.print(F("Minute: "));
        Hamming::debug(bins);
    }
}

namespace DCF77_Second_Decoder {
    using namespace DCF77;

    const uint8_t seconds_per_minute = 60;
    // this is a trick threshold
    //    lower it to get a faster second lock
    //    but then risk to garble the successive stages during startup
    //    --> to low and total startup time will increase
    const uint8_t lock_threshold = 12;

    typedef struct {
        uint8_t data[seconds_per_minute];
        uint8_t tick;

        uint8_t noise_max;
        uint8_t max;
        uint8_t max_index;
    } sync_bins;

    sync_bins bins;

    void sync_mark_binning(const uint8_t tick_data) {
        // We use a binning approach to find out the proper phase.
        // The goal is to localize the sync_mark. Due to noise
        // there may be wrong marks of course. The idea is to not
        // only look at the statistics of the marks but to exploit
        // additional data properties:

        // Bit position  0 after a proper sync is a 0.
        // Bit position 20 after a proper sync is a 1.

        // The binning will work as follows:

        //   1) A sync mark will score +6 points for the current bin
        //      it will also score -2 points for the previous bin
        //                         -2 points for the following bin
        //                     and -2 points 20 bins later
        //  In total this will ensure that a completely lost signal
        //  will not alter the buffer state (on average)

        //   2) A 0 will score +1 point for the previous bin
        //      it also scores -2 point 20 bins back
        //                 and -2 points for the current bin

        //   3) A 1 will score +1 point 20 bins back
        //      it will also score -2 point for the previous bin
        //                     and -2 points for the current bin

        //   4) An undefined value will score -2 point for the current bin
        //                                    -2 point for the previous bin
        //                                    -2 point 20 bins back

        //   5) Scores have an upper limit of 255 and a lower limit of 0.

        // Summary: sync mark earns 6 points, a 0 in position 0 and a 1 in position 20 earn 1 bonus point
        //          anything that allows to infer that any of the "connected" positions is not a sync will remove 2 points

        // It follows that the score of a sync mark (during good reception)
        // may move up/down the whole scale in slightly below 64 minutes.
        // If the receiver should glitch for whatever reason this implies
        // that the clock will take about 33 minutes to recover the proper
        // phase (during phases of good reception). During bad reception things
        // are more tricky.
        using namespace Arithmetic_Tools;

        const uint8_t previous_tick = bins.tick>0? bins.tick-1: seconds_per_minute-1;
        const uint8_t previous_21_tick = bins.tick>20? bins.tick-21: bins.tick + seconds_per_minute-21;

        switch (tick_data) {
            case sync_mark:
                bounded_increment<6>(bins.data[bins.tick]);

                bounded_decrement<2>(bins.data[previous_tick]);
                bounded_decrement<2>(bins.data[previous_21_tick]);

                { const uint8_t next_tick = bins.tick< seconds_per_minute-1? bins.tick+1: 0;
                bounded_decrement<2>(bins.data[next_tick]); }
                break;

            case short_tick:
                bounded_increment<1>(bins.data[previous_tick]);

                bounded_decrement<2>(bins.data[bins.tick]);
                bounded_decrement<2>(bins.data[previous_21_tick]);
                break;

            case long_tick:
                bounded_increment<1>(bins.data[previous_21_tick]);

                bounded_decrement<2>(bins.data[bins.tick]);
                bounded_decrement<2>(bins.data[previous_tick]);
                break;

            case undefined:
            default:
                bounded_decrement<2>(bins.data[bins.tick]);
                bounded_decrement<2>(bins.data[previous_tick]);
                bounded_decrement<2>(bins.data[previous_21_tick]);
        }
        bins.tick = bins.tick<seconds_per_minute-1? bins.tick+1: 0;

        // determine sync lock
        if (bins.max - bins.noise_max <=lock_threshold ||
            get_second() == 3) {
            // after a lock is acquired this happens only once per minute and it is
            // reasonable cheap to process,
            //
            // that is: after we have a "lock" this will be processed whenever
            // the sync mark was detected

            Hamming::compute_max_index(bins);
        }
    }

    void get_quality(Hamming::lock_quality_t &lock_quality) {
        Hamming::get_quality(bins, lock_quality);
    }

    uint8_t get_second() {
        if (bins.max - bins.noise_max >= lock_threshold) {
            // at least one sync mark and a 0 and a 1 seen
            // the threshold is tricky:
            //   higher --> takes longer to acquire an initial lock, but higher probability of an accurate lock
            //
            //   lower  --> higher probability that the lock will oscillate at the beginning
            //              and thus spoil the downstream stages

            // we have to subtract 2 seconds
            //   1 because the seconds already advanced by 1 tick
            //   1 because the sync mark is not second 0 but second 59

            uint8_t second = 2*seconds_per_minute + bins.tick - 2 - bins.max_index;
            while (second >= seconds_per_minute) { second-= seconds_per_minute; }

            return second;
        } else {
            return 0xff;
        }
    }

    void process_single_tick_data(const uint8_t tick_data) {
        sync_mark_binning(tick_data);
    }

    void setup() {
        Hamming::setup(bins);
    }

    void debug() {
        static uint8_t prev_tick;

        if (prev_tick == bins.tick) {
            return;
        } else {
            prev_tick = bins.tick;

            Serial.print(F("second: "));
            Serial.print(get_second(), DEC);
            Serial.print(F(" Sync mark index "));
            Hamming::debug(bins);

            Serial.println();
        }
    }
}

namespace DCF77_Clock_Controller {
    flush_handler output_handler = 0;

    DCF77::time_data decoded_time;
    volatile bool second_toggle;

    void get_current_time(DCF77::time_data &now) {
        for (bool stopper = second_toggle; stopper == second_toggle; ) {
            // wait for second_toggle to toggle
            // that is wait for decoded time to be ready
        }
        now = decoded_time;
    }

    void set_DCF77_encoder(DCF77::time_data &now) {
        using namespace DCF77_Second_Decoder;
        using namespace DCF77_Minute_Decoder;
        using namespace DCF77_Hour_Decoder;

        now.second  = get_second();
        now.minute  = get_minute();
        now.hour    = get_hour();
    }

    void flush() {
        // this is called at the end of each second / before the next second begins
        // it is most interesting to propagate this further at the end of a sync marks
        // it is also interesting to propagate this to reference clocks
        DCF77::time_data now;
        DCF77::time_data now_1;

        set_DCF77_encoder(now);
        now_1 = now;

        DCF77_Encoder::advance_second(now);

        decoded_time.second = now.second;
        if (now.second == 0) {
            // the decoder will always decode the data for the NEXT minute
            // thus we have to keep the data of the previous minute
            decoded_time = now_1;
            decoded_time.second = 0;
        }

        second_toggle = !second_toggle;
        if (output_handler) {
            output_handler(decoded_time);
        }
    }

    void set_flush_handler(const flush_handler new_output_handler) {
        output_handler = new_output_handler;
    }

    void get_quality(clock_quality_t &clock_quality) {
        DCF77_Demodulator::get_quality(clock_quality.phase.lock_max, clock_quality.phase.noise_max);
        DCF77_Second_Decoder::get_quality(clock_quality.second);
        DCF77_Minute_Decoder::get_quality(clock_quality.minute);
        DCF77_Hour_Decoder::get_quality(clock_quality.hour);
    }

    void setup() {
        DCF77_Second_Decoder::setup();
        DCF77_Minute_Decoder::setup();
        DCF77_Hour_Decoder::setup();

        DCF77_Encoder::reset(decoded_time);
    }

    void process_single_tick_data(const uint8_t tick_data) {
        using namespace DCF77;
        using namespace DCF77_Second_Decoder;
        using namespace DCF77_Minute_Decoder;
        using namespace DCF77_Hour_Decoder;

        time_data now;
        set_DCF77_encoder(now);

        DCF77_Encoder::advance_second(now);

        DCF77_Second_Decoder::process_single_tick_data(tick_data);

        if (now.second == 0) {
            DCF77_Minute_Decoder::advance_minute();
            if (now.minute.val == 0x00) {

                // "while" takes automatically care of timezone change
                while (get_hour().val <= 0x23 && get_hour().val != now.hour.val) { advance_hour(); }
            }
        }
        const uint8_t tick_value = (tick_data == long_tick || tick_data == undefined)? 1: 0;
        DCF77_Minute_Decoder::process_tick(now.second, tick_value);
        DCF77_Hour_Decoder::process_tick(now.second, tick_value);
    }
}

namespace DCF77_Demodulator {
    using namespace DCF77;

    const uint8_t bin_count = 100;

    typedef struct {
        uint16_t data[bin_count];
        uint8_t tick;

        uint32_t noise_max;
        uint32_t max;
        uint8_t max_index;
    } phase_bins;

    phase_bins bins;

    const uint16_t samples_per_second = 1000;

    const uint16_t samples_per_bin = samples_per_second / bin_count;
    const uint16_t bins_per_10ms  = bin_count / 100;
    const uint16_t bins_per_50ms  =  5 * bins_per_10ms;
    const uint16_t bins_per_60ms  =  6 * bins_per_10ms;
    const uint16_t bins_per_100ms = 10 * bins_per_10ms;
    const uint16_t bins_per_200ms = 20 * bins_per_10ms;
    const uint16_t bins_per_500ms = 50 * bins_per_10ms;

    void setup() {
        Hamming::setup(bins);

        DCF77_Clock_Controller::setup();
    }

    void decode_220ms(const uint8_t input, const uint8_t bins_to_go) {
        // will be called for each bin during the "interesting" 220ms

        static uint8_t count = 0;
        static uint8_t decoded_data = 0;

        count += input;
        if (bins_to_go >= bins_per_100ms + bins_per_10ms) {
            if (bins_to_go == bins_per_100ms + bins_per_10ms) {
                decoded_data = count > bins_per_50ms? 2: 0;
                count = 0;
            }
        } else {
            if (bins_to_go == 0) {
                decoded_data += count > bins_per_50ms? 1: 0;
                count = 0;
                // pass control further
                // decoded_data: 3 --> 1
                //               2 --> 0,
                //               1 --> undefined,
                //               0 --> sync_mark
                DCF77_Clock_Controller::process_single_tick_data(decoded_data);
            }
        }
    }

    uint16_t wrap(const uint16_t value) {
        // faster modulo function which avoids division
        uint16_t result = value;
        while (result >= bin_count) {
            result-= bin_count;
        }
        return result;
    }

    void phase_detection() {
        // We will compute the integrals over 200ms.
        // The integrals is used to find the window of maximum signal strength.
        uint32_t integral = 0;

        for (uint16_t bin = 0; bin < bins_per_100ms; ++bin)  {
            integral += ((uint32_t)bins.data[bin])<<1;
        }

        for (uint16_t bin = bins_per_100ms; bin < bins_per_200ms; ++bin)  {
            integral += (uint32_t)bins.data[bin];
        }

        bins.max = 0;
        bins.max_index = 0;
        for (uint16_t bin = 0; bin < bin_count; ++bin) {
            if (integral > bins.max) {
                bins.max = integral;
                bins.max_index = bin;
            }

            integral -= (uint32_t)bins.data[bin]<<1;
            integral += (uint32_t)(bins.data[wrap(bin + bins_per_100ms)] + bins.data[wrap(bin + bins_per_200ms)]);
        }

        // max_index indicates the position of the 200ms second signal window.
        // Now how can we estimate the noise level? This is very tricky because
        // averaging has already happened to some extend.

        // The issue is that most of the undesired noise happens around the signal,
        // especially after high->low transitions. So as an approximation of the
        // noise I test with a phase shift of 200ms.
        bins.noise_max = 0;
        const uint16_t noise_index = wrap(bins.max_index + bins_per_200ms);

        for (uint16_t bin = 0; bin < bins_per_100ms; ++bin)  {
            bins.noise_max += ((uint32_t)bins.data[wrap(noise_index + bin)])<<1;
        }

        for (uint16_t bin = bins_per_100ms; bin < bins_per_200ms; ++bin)  {
            bins.noise_max += (uint32_t)bins.data[wrap(noise_index + bin)];
        }
    }

    uint8_t phase_binning(const uint8_t input) {
        // how many seconds may be cummulated
        // this controls how slow the filter may be to follow a phase drift
        // N times the clock precision shall be smaller 1
        // clock 30 ppm => N < 300
        const uint16_t N = 300;

        Hamming::advance_tick(bins);

        if (input) {
            if (bins.data[bins.tick] < N) {
                ++bins.data[bins.tick];
            }
        } else {
            if (bins.data[bins.tick] > 0) {
                --bins.data[bins.tick];
            }
        }
        return bins.tick;
    }

    void detector_stage_2(const uint8_t input) {
        const uint8_t current_bin = bins.tick;

        const uint8_t threshold = 30;

        if (bins.max-bins.noise_max < threshold ||
            wrap(bin_count + current_bin - bins.max_index) == 53) {
            // Phase detection far enough out of phase from anything that
            // might consume runtime otherwise.
            phase_detection();
        }

        static uint8_t bins_to_process = 0;

        if (bins_to_process == 0) {
            if (wrap((bin_count + current_bin - bins.max_index)) <= bins_per_100ms ||   // current_bin at most 100ms after phase_bin
                wrap((bin_count + bins.max_index - current_bin)) <= bins_per_10ms ) {   // current bin at most 10ms before phase_bin
                // if phase bin varies to much during one period we will always be screwed in may ways...

                // last 10ms of current second
                DCF77_Clock_Controller::flush();

                // start processing of bins
                bins_to_process = bins_per_200ms + 2*bins_per_10ms;
            }
        }

        if (bins_to_process > 0) {
            --bins_to_process;

            // this will be called for each bin in the "interesting" 220ms
            // this is also a good place for a "monitoring hook"
            decode_220ms(input, bins_to_process);
        }
    }

    void detector(const uint8_t sampled_data) {
        static uint8_t current_sample = 0;
        static uint8_t average = 0;

        // detector stage 0: average 10 samples (per bin)
        average += sampled_data;

        if (++current_sample >= samples_per_bin) {
            // once all samples for the current bin are captured the bin gets updated
            // that is each 10ms control is passed to stage 1
            const uint8_t input = (average> samples_per_bin/2);

            phase_binning(input);

            detector_stage_2(input);

            average = 0;
            current_sample = 0;
        }
    }

    void get_quality(uint32_t &lock_max, uint32_t &noise_max) {
        lock_max = bins.max;
        noise_max = bins.noise_max;
    }

    void debug() {
        Serial.print(F("Phase: "));
        Hamming::debug(bins);
    }
}


const uint8_t dcf77_sample_pin = 19; // A5
const uint8_t dcf77_analog_sample_pin = 5;
const uint8_t dcf77_monitor_pin = 18;


void process_one_sample() {
    // The Blinkenlight LEDs will cut off the input signal
    // below a logical high. This is due to the weak
    // output signal of the DCF module.
    // Hence for the Blinkenlighty
    // and the Blinkenlight shield we can not use
    // digitalRead.

    // Comment the line below if you are not using this code with a
    // Blinkenlighty or Blinkenlight shield.
    const uint8_t sampled_data = analogRead(dcf77_analog_sample_pin)>200? 1: 0;
    // Uncomment the line below if you are using this code with a standard Arduino
    //const uint8_t sampled_data = digitalRead(dcf77_sample_pin);

    digitalWrite(dcf77_monitor_pin, sampled_data);

    DCF77_Demodulator::detector(sampled_data);
}

ISR(TIMER2_COMPA_vect) {
    process_one_sample();
}

void initTimer2() {
    // Timer 2 CTC mode, prescaler 64
    TCCR2B = (0<<WGM22) | (1<<CS22);
    TCCR2A = (1<<WGM21) | (0<<WGM20);

    // 249 + 1 == 250 == 250 000 / 1000 =  (16 000 000 / 64) / 1000
    OCR2A = 249;

    // enable Timer 2 interrupts
    TIMSK2 = (1<<OCIE2A);
}

void stopTimer0() {
    // ensure that the standard timer interrupts will not
    // mess with msTimer2
    TIMSK0 = 0;
}


void output_handler(const DCF77::time_data &decoded_time) {
    uint8_t lo;
    uint8_t hi;
    uint8_t out;
    
    if (decoded_time.second < 60) {
        // bcd conversion
        hi = decoded_time.second / 10;
        lo = decoded_time.second - 10 * hi;
        out = (hi<<4) + lo;
        
        uint8_t pin = first_output_pin;
        for (uint8_t bits=0; bits<8; ++bits) {
            digitalWrite(pin++, out & 0x1);
            out >>= 1;
        }
    }
}

DCF77::time_data clock;

void setup() {
    using namespace DCF77_Encoder;

    Serial.begin(115200);
    Serial.println();

    pinMode(dcf77_monitor_pin, OUTPUT);

    pinMode(dcf77_sample_pin, INPUT);
    digitalWrite(dcf77_sample_pin, HIGH);

    for (uint8_t pin=first_output_pin; pin<first_output_pin+8; ++pin) {
        pinMode(pin, OUTPUT);
        digitalWrite(pin, HIGH);
    }

    DCF77_Demodulator::setup();
    DCF77_Clock_Controller::set_flush_handler(&output_handler);

    #ifndef DEBUG
    Serial.flush();

    initTimer2();

    stopTimer0();
    #endif
}

#ifdef DEBUG
int free_Ram () {
    extern int __heap_start, *__brkval;
    int v;
    return (int) &v - (__brkval == 0 ? (int) &__heap_start : (int) __brkval);
}

bool noise() {
    const uint8_t r = random(97);
    return r < 0? r&1: 0;
}
#endif

void loop() {
    //while (tick != 0) {};

#ifdef DEBUG
    using namespace Debug;
    Serial.print(F("free Ram: "));
    Serial.println(free_Ram());

    for (uint8_t i=0; i<60; ++i) {
        DCF77_Encoder::advance_second(clock);

        const uint8_t signal = DCF77_Encoder::get_current_signal(clock);
        Debug::debug_helper(signal);

        for (uint8_t passes=0; passes<100; ++passes) {
            DCF77_Demodulator::detector((signal >= DCF77::undefined? 1: 0));//^noise());
        }
        for (uint8_t passes=0; passes<100; ++passes) {
            DCF77_Demodulator::detector((signal == DCF77::long_tick? 1: 0));//^noise());
        }
        for (uint16_t passes=0; passes<800; ++passes) {
            DCF77_Demodulator::detector(0&&noise());
        }
    }

    Serial.println();
    Serial.println();

    Serial.println(F("-----------"));
    Serial.print(F("reference: hh:mm:ss  "));
    Serial.print(clock.hour.val, HEX);
    Serial.print(':');
    Serial.print(clock.minute.val, HEX);
    Serial.print(':');
    Serial.println(clock.second, DEC);
#endif

    DCF77::time_data now;

    DCF77_Clock_Controller::get_current_time(now);
    if (now.second <= 60) {
        Serial.println();
        Serial.print(F("Decoded time: "));

        DCF77_Encoder::debug(now);
    }

    DCF77_Clock_Controller::clock_quality_t clock_quality;
    DCF77_Clock_Controller::get_quality(clock_quality);

    Serial.print(F("Quality (p,s,m,h): "));
    Serial.print('(');
    Serial.print(clock_quality.phase.lock_max, DEC);
    Serial.print('-');
    Serial.print(clock_quality.phase.noise_max, DEC);
    Serial.print(')');

    Serial.print('(');
    Serial.print(clock_quality.second.lock_max, DEC);
    Serial.print('-');
    Serial.print(clock_quality.second.noise_max, DEC);
    Serial.print(')');

    Serial.print('(');
    Serial.print(clock_quality.minute.lock_max, DEC);
    Serial.print('-');
    Serial.print(clock_quality.minute.noise_max, DEC);
    Serial.print(')');

    Serial.print('(');
    Serial.print(clock_quality.hour.lock_max, DEC);
    Serial.print('-');
    Serial.print(clock_quality.hour.noise_max, DEC);
    Serial.print(')');

    /**/

    //DCF77_Hour_Decoder::debug();
    //DCF77_Minute_Decoder::debug();
    //DCF77_Second_Decoder::debug();
}

3 Responses to Decoding Time Data

  1. Markus Gritsch says:

    Hi Udo,

    amazing code. I really like your matched filter approach 🙂
    One question: How does the hour decoder respond to changes from and to daylight saving time?

    Cheers,
    Markus

    • Hi Markus,

      the current published version does not yet deal with timezone change. The current code line (under test) was running during the timezone change and performed flawless. However it is not yet fully documented. For the final version there will be several different cases:

      1) The clock was synced at prior to the timezone changed and is still locked.
      –> Simple, it can ignore the flags as the hour of the change is ruled by law –> as long at it can keep the phase lock the transition will happen smoothly.
      1a) In case of a sync during the “double hour” that happens at CEST -> CET transition and at very high noise levels it may lock to the wrong hour. The chance is very tiny but it might happen. In this case it might take some time for it to recognize the issue. (Slightly longer than the time required for the initial lock).

      2) In case it is syncing while the transition happens but can not acquire a lock before the transition is finished and the timezone data is noisy the impact is that sync time will double. If the data is clean and you happen to start syncing at the worst possible time it may take about 5-10 Minutes longer for the first sync.

      I will publish the parts as I finish and test them. The next steps are:
      a) add more decoder stages
      b) add a local clock to deal with signal loss
      c) some surprise how to keep a significantly better phase lock than the previous versions (once a first sync is acquired)

      Cheers, Udo

  2. Markus Gritsch says:

    Great, looking forward to it. I really like reading your excellent articles. Keep up the good work!

    Cheers,
    Markus

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s