Re: [Discuss-gnuradio] Question about stream to complex and fft
Hello S. L.: Short Version: Yes, you are correct! Some picky details: Due to technical limitations, the output timing of the Throttle block is not perfect (but should be good enough, for most practical use cases). Also, due to this, the Throttle block may output more than one samples at a time. GNU Radio blocks use buffers to communicate. A block outputs data by writing the generated data into its output buffer. The next block in the chain considers this buffer as an input buffer, and reads data from it. It then tells the previous block that the data has been read, so as the previous block can generate more data. The Throttle block intentionally slows down this process. But in overall, your idea is correct. Regards, Kyeong Su Shin 보낸 사람: S. L. 보낸 날짜: 2019년 6월 14일 금요일 오후 9:26:45 받는 사람: Kyeong Su Shin; discuss-gnuradio@gnu.org; rear1019 제목: RE: Question about stream to complex and fft Thank you very much for your responses So, if for example, I want to simulate 1 second of a cosine wave of 1K, with sample rate of 32K as the example and I have a very powerful computer the signal source block (cosine) could generate the 32k samples in 0.5 seconds , but the 32K wave samples of a 1K would correspond to real (good) values as if it had been sampling for 1 second ... Is that correct? And if I put a Throttle block at the output of the signal source block at 32K of sample rate, the throttle block will receive the 32k samples as input at a lower time, store them in a buffer and direct them to the output at the precise moment, it is say every 1 / 32K ms ... Is this correct? Thank you very much for your time and for your kindness De: Kyeong Su Shin Enviado: viernes, 14 de junio de 2019 13:31 Para: S. L.; discuss-gnuradio@gnu.org Asunto: RE: Question about stream to complex and fft Hello S.L: That sounds correct to me. One note, however: the sampling rate in GNU Radio is just a conceptual thing, and does not determine the 'actual time that GNU Radio takes to generate or consume the samples', if you do not use hardware sinks/sources (like SDR sources) and/or throttle blocks. If your flowgraph does not have such blocks, then the flowgraph will execute 'as fast as it can'. The data generated by your program will be still correct; it is just that the 'time' is just a conceptual thing (so as you don't have to deal with normalized sampling rates) and may not agree with the actual execution speed of the program. Kyeong Su Shin 보낸 사람: S. L. 대신 Discuss-gnuradio 보낸 날짜: 2019년 6월 14일 금요일 오후 6:08:55 받는 사람: discuss-gnuradio@gnu.org 제목: [Discuss-gnuradio] Question about stream to complex and fft Hello, I have a question about the real-time operation of gnuradio and the "stream-to-vector" and FFT modules " In the example image I have: variable samp_rate to 32K and the following flowchart: Signal source (waveform: cosine, sample_rate: 32K, frecuency: 1K) -> stream-to-vector (Num items: 1024) -> FFT (size: 1024) -> Complex-to-Mag (Vector lenght: 1024) - > QT GUI Vector Sink (Vector size: 1024) Then cosine has 1K of frequency and sample rate is 32K and FFT size is 1024 I'm not really sure how the gnuradio simulation works so I'd like to check it with you and see if I'm right or I'm wrong A whole cosine wave would take 1ms since the sample rate is 32K, a cosine wave would be 32 samples Then 1024 samples would be 1024/32 = 32 cosine waves, that is 32 * 1 ms = 32 ms Then: Since I have vector-to-stream and fft to 1024 samples, I think or I guess it stores the first 1024 samples (from 0 ms to 32 ms) and the vector-to-stream module passes those samples to the FFT module that performs the FFT of those 1024 samples Then another 1024 samples (form 32 ms to 64 ms) would arrive and I would pass those samples to the FFT That is, the FFT block performs its function every 32 ms with 1024 sample vectors Is this operating logic on the stream-to-vector and fft block correct? or I'm wrong Thank you very much everyone for your time a greeting ___ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
Re: [Discuss-gnuradio] Question about stream to complex and fft
Thank you very much for your responses So, if for example, I want to simulate 1 second of a cosine wave of 1K, with sample rate of 32K as the example and I have a very powerful computer the signal source block (cosine) could generate the 32k samples in 0.5 seconds , but the 32K wave samples of a 1K would correspond to real (good) values as if it had been sampling for 1 second ... Is that correct? And if I put a Throttle block at the output of the signal source block at 32K of sample rate, the throttle block will receive the 32k samples as input at a lower time, store them in a buffer and direct them to the output at the precise moment, it is say every 1 / 32K ms ... Is this correct? Thank you very much for your time and for your kindness De: Kyeong Su Shin Enviado: viernes, 14 de junio de 2019 13:31 Para: S. L.; discuss-gnuradio@gnu.org Asunto: RE: Question about stream to complex and fft Hello S.L: That sounds correct to me. One note, however: the sampling rate in GNU Radio is just a conceptual thing, and does not determine the 'actual time that GNU Radio takes to generate or consume the samples', if you do not use hardware sinks/sources (like SDR sources) and/or throttle blocks. If your flowgraph does not have such blocks, then the flowgraph will execute 'as fast as it can'. The data generated by your program will be still correct; it is just that the 'time' is just a conceptual thing (so as you don't have to deal with normalized sampling rates) and may not agree with the actual execution speed of the program. Kyeong Su Shin 보낸 사람: S. L. 대신 Discuss-gnuradio 보낸 날짜: 2019년 6월 14일 금요일 오후 6:08:55 받는 사람: discuss-gnuradio@gnu.org 제목: [Discuss-gnuradio] Question about stream to complex and fft Hello, I have a question about the real-time operation of gnuradio and the "stream-to-vector" and FFT modules " In the example image I have: variable samp_rate to 32K and the following flowchart: Signal source (waveform: cosine, sample_rate: 32K, frecuency: 1K) -> stream-to-vector (Num items: 1024) -> FFT (size: 1024) -> Complex-to-Mag (Vector lenght: 1024) - > QT GUI Vector Sink (Vector size: 1024) Then cosine has 1K of frequency and sample rate is 32K and FFT size is 1024 I'm not really sure how the gnuradio simulation works so I'd like to check it with you and see if I'm right or I'm wrong A whole cosine wave would take 1ms since the sample rate is 32K, a cosine wave would be 32 samples Then 1024 samples would be 1024/32 = 32 cosine waves, that is 32 * 1 ms = 32 ms Then: Since I have vector-to-stream and fft to 1024 samples, I think or I guess it stores the first 1024 samples (from 0 ms to 32 ms) and the vector-to-stream module passes those samples to the FFT module that performs the FFT of those 1024 samples Then another 1024 samples (form 32 ms to 64 ms) would arrive and I would pass those samples to the FFT That is, the FFT block performs its function every 32 ms with 1024 sample vectors Is this operating logic on the stream-to-vector and fft block correct? or I'm wrong Thank you very much everyone for your time a greeting ___ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
Re: [Discuss-gnuradio] Question about stream to complex and fft
On Fri, 14 Jun 2019 at 09:08:55 +, S. L. wrote: > Since I have vector-to-stream and fft to 1024 samples, I think or I > guess it stores the first 1024 samples (from 0 ms to 32 ms) and the > vector-to-stream module passes those samples to the FFT module that > performs the FFT of those 1024 samples Then another 1024 samples (form > 32 ms to 64 ms) would arrive and I would pass those samples to the FFT > > That is, the FFT block performs its function every 32 ms with 1024 > sample vectors > > Is this operating logic on the stream-to-vector and fft block correct? > or I'm wrong Your description of what happens is correct: The stream to vector block waits for N samples and passes them to the FFT block (N being the vector size). Strictly speaking the stream to vector blocks waits for n·N samples where n is a positive integer, i.e. the delay can be bigger than N samples. The value of n is determined by GNU Radio’s scheduler and is not necessarily constant during the simulation. The temporal aspect of your description is only correct if you refer to the time passed “inside” the simulation. The time passed from the point of view of a user watching the simulation (“real time”) may be different. Your flowgraph does not restrict the processing speed, i.e. generating and processing 32k samples will probably (depending on CPU) take less than 1 second of real time. See also reply by Kyeong Su Shin. ___ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
Re: [Discuss-gnuradio] Question about stream to complex and fft
Hello S.L: That sounds correct to me. One note, however: the sampling rate in GNU Radio is just a conceptual thing, and does not determine the 'actual time that GNU Radio takes to generate or consume the samples', if you do not use hardware sinks/sources (like SDR sources) and/or throttle blocks. If your flowgraph does not have such blocks, then the flowgraph will execute 'as fast as it can'. The data generated by your program will be still correct; it is just that the 'time' is just a conceptual thing (so as you don't have to deal with normalized sampling rates) and may not agree with the actual execution speed of the program. Kyeong Su Shin 보낸 사람: S. L. 대신 Discuss-gnuradio 보낸 날짜: 2019년 6월 14일 금요일 오후 6:08:55 받는 사람: discuss-gnuradio@gnu.org 제목: [Discuss-gnuradio] Question about stream to complex and fft Hello, I have a question about the real-time operation of gnuradio and the "stream-to-vector" and FFT modules " In the example image I have: variable samp_rate to 32K and the following flowchart: Signal source (waveform: cosine, sample_rate: 32K, frecuency: 1K) -> stream-to-vector (Num items: 1024) -> FFT (size: 1024) -> Complex-to-Mag (Vector lenght: 1024) - > QT GUI Vector Sink (Vector size: 1024) Then cosine has 1K of frequency and sample rate is 32K and FFT size is 1024 I'm not really sure how the gnuradio simulation works so I'd like to check it with you and see if I'm right or I'm wrong A whole cosine wave would take 1ms since the sample rate is 32K, a cosine wave would be 32 samples Then 1024 samples would be 1024/32 = 32 cosine waves, that is 32 * 1 ms = 32 ms Then: Since I have vector-to-stream and fft to 1024 samples, I think or I guess it stores the first 1024 samples (from 0 ms to 32 ms) and the vector-to-stream module passes those samples to the FFT module that performs the FFT of those 1024 samples Then another 1024 samples (form 32 ms to 64 ms) would arrive and I would pass those samples to the FFT That is, the FFT block performs its function every 32 ms with 1024 sample vectors Is this operating logic on the stream-to-vector and fft block correct? or I'm wrong Thank you very much everyone for your time a greeting ___ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
