HDMI: Random micro-cuts of sound

[sy2002]

User Tonton#0297 reported on Discord: "Random micro-cuts of the sound through the hdmi port. No issue with the audio jack of the Mega65" Discord Link here.

@amb5l and @MJoergen

Challenge: Does seem to be an extremely rare problem: Nobody ever reported this. So I asked Tonton to make a video with his phone or so, so that we understand the problem a bit better.

[amb5l]

This has to be* a small but cumulative delta between perfect and actual frequencies of the pixel and PCM clocks. Would one of you be able to summarise the MMCM recipe (ref clock, multiplier, divider, output divider) for each?

Is the PCM clocking done in the "clock enable" way? If so there may be a tiny but cumulative error in the counting logic. It should be quite easy to create a small/fast simulation focused on looking for this.

(*) I suppose it's also possible that it's an issue with ACR packets. That their frequency is a little too low and that the audio clock PLL in the TV/monitor doesn't like that.

[kaloun34]

hello
This issue is random but can happen quickly

Mega65.mp4

[sy2002]

The video link does not seem to work on GitHub, here is a link to Discord that shows the glitch after about 7 seconds: The audio goes away and then comes back:
https://discord.com/channels/719326990221574164/794775503818588200/1066690068531658813

[sy2002]

Tonton confirmed that the problem occurs in our old (pre-audio refactoring) version of the core V4 as well as in the new alpha version A7 for version 5 after @MJoergen refactored some stuff - so it looks like it was not the audio refactoring and we can use the refactored version in the develop branch to investigate further.

@amb5l about your questions:

Here is the clocking for the Commodore 64:
https://github.com/MJoergen/C64MEGA65/blob/develop/CORE/vhdl/clk.vhd

As you can see there, we are - when the mode "HDMI: Flicker-fix" is ON slowing down the whole C64 system to make sure that the framerate is much closer to 50 Hz(*) and therefore we don't have tearing/flickering effects while scrolling (more details about this are here but not sure if this is relevant for the problem).

(*) re 50 Hz: The core is a PAL-only core at this moment, therefore 50 Hz.

The MiSTer2MEGA65 framework is then translating the C64 output to HDMI using this clocking:
https://github.com/MJoergen/C64MEGA65/blob/develop/M2M/vhdl/clk_m2m.vhd

As you can see there, everything is pretty plain vanilla the frequency multipliers and dividers for example for 74.25 MHz for 720p (50 Hz or 60 Hz) and 5x74.25 MHz = 371.25 MHz for TMDS result in "nice" values. For example:

f_VCO = (100 MHz / 5) x 37.125 = 742.5 MHz
742.5 / 2 = 371.25 MHz
742.5 / 10 = 74.25

About your question "Is the PCM clocking done in the "clock enable" way?":

Yes, it is generated from the HDMI clock via a clock enable:

   -- ACR packet rate should be 128fs/N = 1kHz
   p_pcm_acr : process (hdmi_clk_i)
   begin
      if rising_edge(hdmi_clk_i) then
         if pcm_clken = '1' then
            -- Generate 1KHz ACR pulse train from 48 kHz
         if pcm_acr_counter /= (pcm_acr_cnt_range - 1) then
            pcm_acr_counter <= pcm_acr_counter + 1;
            pcm_acr <= '0';
         else
            pcm_acr <= '1';
            pcm_acr_counter <= 0;
         end if;
      end if;
      end if;
   end process p_pcm_acr;

The clock enable pcm_clken is generated like this:

   -- Generate PCM sample rate (48 kHz)
   i_clk_synthetic_enable : entity work.clk_synthetic_enable
      port map (
         clk_i       => hdmi_clk_i,
         src_speed_i => hdmi_video_mode.CLK_KHZ * 1000,
         dst_speed_i => HDMI_PCM_SAMPLING,
         enable_o    => pcm_clken
      ); -- i_clk_synthetic_enable

And the clk_synthetic_enable component works like this:
https://github.com/MJoergen/C64MEGA65/blob/develop/M2M/vhdl/av_pipeline/clk_synthetic_enable.vhd

[sy2002]

@kaloun34

Which HDMI mode are you using? (Which resolution, Hz) AND: Do you have HDMI: Flicker-free on or off when this happens?

[kaloun34]

[sy2002]

@kaloun34 Can you please test this core? hdmi_tpg_mega65r3.zip

It is an HDMI test core. @amb5l built it and it uses an advanced version of the HDMI engine that the C64 core is using. The test core shows a test image and plays a test sound.

Are the random micro-cuts still audible on your monitor with this test core?

When the core starts, please press MEGA65's RETURN key 8 times so that you are in 576p @ 50 Hz. When finished testing this mode, press the RETURN key two more times (10 times overall) to enter the mode 720p @ 50 Hz. These two modes are used by the C64 core.

[sy2002]

Just in case the .bit above is not sufficient, here is a .cor: hdmi_tpg_mega65r3.zip

[kaloun34]

@sy2002
Hello
Random micro-cuts are not audible with this Core test, but I would need continuous audio to be sure.

[sy2002]

@kaloun34 Thank you for your feedback and @amb5l we might need a version with continuous audio to be sure.