high noise/signal in a simple serial to mono dac module
in [FPGA]
|
From: Fei Liu on 11 Apr 2008 17:32 Hello I am working on playing simple 11 kbytes/second 8bit WAV data through my FPGA stereo jack. I used the following synthesis to play the sound data delivered through the serial interface. I also used the mono_dac demo module. My problem is I am getting a very high noise to signal ratio although I can hear the speech. The sound file plays crystal clear on my host computer. Another problem I have with simple WAV file is I don't know how to extract volume information. So I hardwired 3'b100 as the volume. I could use some help figuring out the source of the high noise. Is this caused by timing issue? Fei WAV file info: File : 11k8bitpcm.wav Length : 152312 RIFF : 152304 WAVE fmt : 16 Format : 0x1 => WAVE_FORMAT_PCM Channels : 1 Sample Rate : 11025 Block Align : 1 Bit Width : 8 Bytes/sec : 11025 data : 152267 End ---------------------------------------- Sample Rate : 11025 Frames : 152267 Channels : 1 Format : 0x00010005 Sections : 1 Seekable : TRUE Duration : 00:00:13.811 Signal Max : 128 (0.00 dB) `include "timescale.v" module serial_snd( clk, rx, tx, led0, led1, led2, aud_l, aud_r ); parameter clk_freq=50000000; parameter blink_freq = 25; (* LOC="E12" *) input clk; (* LOC="F16" *) input rx; (* LOC="E15" *) output tx; (* LOC="R20" *) output led0; (* LOC="T19" *) output led1; (* LOC="U20" *) output led2; (* LOC="Y10" *) output aud_l; (* LOC="V10" *) output aud_r; wire data_ready; wire RxD_idle; wire [7:0] rx_data; reg [blink_freq:0] count = 0; // serial port receiver async_receiver #(.clk_freq(clk_freq)) ar( .clk(clk), .RxD(rx), .RxD_data_ready(data_ready), .RxD_data(rx_data), .RxD_idle(RxD_idle) ); // Serial port transmitter async_transmitter #(.clk_freq(clk_freq)) serializer( .clk(clk), .TxD(tx), .TxD_start(data_ready), .TxD_data(rx_data) ); // Blink LED0 with a frequency slightly slower than once per second assign led0 = count[blink_freq]; assign led2 = RxD_idle; // LED1 lights on when there is incoming data from serial line, it lasts 2^20 clks state_holder #(.clk_hold(20)) bl(clk, data_ready, led1); //PWM audio( // .clk(clk), // .PWM_in(rx_data), // .PWM_out(aud_l) //); wire audio; FD fd_aud_lt_inst (.D(audio),.Q(aud_l),.C(clk)); mono_dac mono( .signed_data(rx_data), .volume(3'b100), .audio(audio), .clk(clk) ); assign aud_r = 8'bz; always @(posedge clk) begin count <= count + 1; end endmodule
From: kevin93 on 11 Apr 2008 22:02 On Apr 11, 2:32 pm, Fei Liu <fei....(a)gmail.com> wrote: > Hello > I am working on playing simple 11 kbytes/second 8bit WAV data through my > FPGA stereo jack. I used the following synthesis to play the sound data > delivered through the serial interface. I also used the mono_dac demo > module. > > My problem is I am getting a very high noise to signal ratio although I > can hear the speech. The sound file plays crystal clear on my host > computer. <.. snip ..> > > //PWM audio( > // .clk(clk), > // .PWM_in(rx_data), > // .PWM_out(aud_l) > //); > wire audio; > FD fd_aud_lt_inst (.D(audio),.Q(aud_l),.C(clk)); > > mono_dac mono( > .signed_data(rx_data), > .volume(3'b100), > .audio(audio), > .clk(clk) > ); <.. snip ..> 8-bit WAV files use offset binary rather than signed values (i.e. 0 is the most negative, 128 is mid-scale 255 is full-scale) kevin
From: Fei Liu on 12 Apr 2008 00:36 On Apr 11, 10:02 pm, kevin93 <ke...(a)whitedigs.com> wrote: > On Apr 11, 2:32 pm, Fei Liu <fei....(a)gmail.com> wrote: > > > Hello > > I am working on playing simple 11 kbytes/second 8bit WAV data through my > > FPGA stereo jack. I used the following synthesis to play the sound data > > delivered through the serial interface. I also used the mono_dac demo > > module. > > > My problem is I am getting a very high noise to signal ratio although I > > can hear the speech. The sound file plays crystal clear on my host > > computer. > <.. snip ..> > > > //PWM audio( > > // .clk(clk), > > // .PWM_in(rx_data), > > // .PWM_out(aud_l) > > //); > > wire audio; > > FD fd_aud_lt_inst (.D(audio),.Q(aud_l),.C(clk)); > > > mono_dac mono( > > .signed_data(rx_data), > > .volume(3'b100), > > .audio(audio), > > .clk(clk) > > ); > > <.. snip ..> > > 8-bit WAV files use offset binary rather than signed values (i.e. 0 is > the most negative, 128 is mid-scale 255 is full-scale) > > kevin Ah, that's sweet to know. Where can I find similar information for 16bit and even 32bit WAV files? Thanks a lot! Fei
From: Alan Nishioka on 12 Apr 2008 01:35 On Apr 11, 7:02 pm, kevin93 <ke...(a)whitedigs.com> wrote: > On Apr 11, 2:32 pm, Fei Liu <fei....(a)gmail.com> wrote: > > > Hello > > I am working on playing simple 11 kbytes/second 8bit WAV data through my > > FPGA stereo jack. I used the following synthesis to play the sound data > > delivered through the serial interface. I also used the mono_dac demo > > module. > > > My problem is I am getting a very high noise to signal ratio although I > > can hear the speech. The sound file plays crystal clear on my host > > computer. > <.. snip ..> > > > //PWM audio( > > // .clk(clk), > > // .PWM_in(rx_data), > > // .PWM_out(aud_l) > > //); > > wire audio; > > FD fd_aud_lt_inst (.D(audio),.Q(aud_l),.C(clk)); > > > mono_dac mono( > > .signed_data(rx_data), > > .volume(3'b100), > > .audio(audio), > > .clk(clk) > > ); > > <.. snip ..> > > 8-bit WAV files use offset binary rather than signed values (i.e. 0 is > the most negative, 128 is mid-scale 255 is full-scale) > > kevin And to convert signed to offset or to convert offset to signed you just have to invert the top bit (MSB). Since it is so easy, it might be worth it to just try it and see if it works. Alan Nishioka
From: Fei Liu on 13 Apr 2008 00:40
Alan Nishioka wrote: > On Apr 11, 7:02 pm, kevin93 <ke...(a)whitedigs.com> wrote: >> On Apr 11, 2:32 pm, Fei Liu <fei....(a)gmail.com> wrote: >> >>> Hello >>> I am working on playing simple 11 kbytes/second 8bit WAV data through my >>> FPGA stereo jack. I used the following synthesis to play the sound data >>> delivered through the serial interface. I also used the mono_dac demo >>> module. >>> My problem is I am getting a very high noise to signal ratio although I >>> can hear the speech. The sound file plays crystal clear on my host >>> computer. >> <.. snip ..> >> >>> //PWM audio( >>> // .clk(clk), >>> // .PWM_in(rx_data), >>> // .PWM_out(aud_l) >>> //); >>> wire audio; >>> FD fd_aud_lt_inst (.D(audio),.Q(aud_l),.C(clk)); >>> mono_dac mono( >>> .signed_data(rx_data), >>> .volume(3'b100), >>> .audio(audio), >>> .clk(clk) >>> ); >> <.. snip ..> >> >> 8-bit WAV files use offset binary rather than signed values (i.e. 0 is >> the most negative, 128 is mid-scale 255 is full-scale) >> >> kevin > > And to convert signed to offset or to convert offset to signed you > just have to invert the top bit (MSB). Since it is so easy, it might > be worth it to just try it and see if it works. > > Alan Nishioka Hmm, it doesn't seem to work. I use libsndfile and libserial to decode and deliver data to serial port in this code: static const sf_count_t nframes = 16; signed char data; char ptr[nframes]; int count = 0; while(sfh.readRaw(ptr, nframes)){ count ++; if(!(count%1000)) printf("Frames: %d\n", count); int i = 0; while(i < nframes){ data = ptr[i]; data -= 0x80; my_serial_stream << data; i++; } } I just get a load of noise. I analyzed the signal sent to serial port (from the Tx output), there is no glitch. Fei |