Wow great video! In the world of so called professors, just read variables x's y's in their slides and leave their students in sheer confusion, this is how anyone should teach.
Great question. Yes, but since the transmission power is high, the noise at the transmitter is insignificant in comparison, so we ignore it in our models. The receiver only collects a small fraction of the power that was transmitted (eg. in wireless communications it radiates out in all directions), and so the noise in the receiver's electronics can't be ignored. Glad you liked the video.
Quick question: the equation y=hx+n assumes that AWGN is added after the channel. what if part of the AWGN is caused by the transmitter electronics? then we will end up with y=h(n1+x)+n2, where n1 is the AWGN before the channel and n2 is the AWGN after the channel. We just assume that due to the central limit theorem, all the noise is modelled as a single AWGN value and added at the receiver. The question is: how do we model seperate noise sources? are there any reference where noise is added before and after transmission? PS: the same discussion has taken place in the previous post....
Good question. No, we don't add the transmitter and receiver noises and assume it's OK due to the central limit theorem. Actually what happens is we simply ignore the transmitter noise. In most communication systems the transmit signal power is much higher than the transmitter's noise power, so the transmitter's noise has virtually no effect on the transmitted signal. At the receiver it's a different story, since the transmit power has dissipated by the time it reaches the receiver, so the received signal power is much lower, and is often comparable with the receiver noise power.
Thank you for your reply. I guess the confusion comes from the way we model baseband communications, ie y=x+n, which implies noise is added onto the transmitted signal. In reality we should think of (x) as the transmitted signal seen by receiver.
You said that when we have stochastic process (here with the elactrons) with many realizations, we get a GAUSSIAN shape of the probability density function. Is there a special theorem that states that? Thanks
In Gaussian noise pdf curve the speediness of the curve occurs from which parameter. Explain the parameter and give details of the curve with help of example.
hello lain , intigrator can be also viewed as low pass filter how can we relate this low pass filter operation with the collection of energy (the task performed by intigrator at the receiver)
An integrator is not a low pass filter. People sometimes say that it acts like a low pass filter, but that's just confusing, since the gain of the integrator at f=0 is infinity! That's definitely not something you would call a "filter". You can say that high frequencies do not affect the output of the integrator, but that's not the same as saying it is a "filter".
Thanks for the suggestion. I'll put a "white noise" video on my to-do list. In summary though, "white noise" is another way of saying "i.i.d. noise". I explain what i.i.d. means in this video. The reason it is called "white" is that an i.i.d. variable has an auto-correlation function that is an impulse (delta function), which means its power spectral density (psd) is flat. A flat psd means that all frequencies are equally represented. This is analogous to white light, which is made up of equal amounts of all the colours / visible frequencies.
@@iain_explains, Sir I have a doubt,we have an iid random variable ,but it has no width in time,it is discrete in nature,in the case of a random variable with N(0,1) we have a white noise only when it is coupled with a continous func,to make to fit all together,other wise we may need to take DTFT to find psd,I'm I right sir,whether my concepts are correct..
@@vaishnav4035 OK, so the be even more precise, the noise in a digital communication system is a discrete time Random Process where the sampling is happening at the symbol rate of the digital signal (and where each sample is a Gaussian Random Variable).
Well, yes, strictly speaking the integrator just keeps integrating, so technically I should have written the integral limits as being from zero (or negative infinity) to t, and I should have used a different variable for the integration variable (perhaps tau). But doing that can be confusing to people who are not so mathematically minded, so generally the integral in the detector is written the way I wrote it - ie. for just the first symbol (which is over the time period from 0 to T).
What is therm 'n' actually, is it the frequency, is it the amplitude. I am trying to understand this in terms of DSP though, and one thing more, how can we relate white noise to gaussian noise? Please help me understand this
Thanks for your question. OK, so a couple of things to clarify first: n_k is the "noise" that is disrupting the measurement of the k-th data bit, which is measured at time kT. This is a number. It is a voltage. It is random. It could be positive or negative. Now, in the video I have used "n" (without the subscript) to represent a general n_k (ie. for any value of k). The plot shows the probability density function (pdf) for n. If you're not sure what a pdf is, then please check out the video I made on pdf's (th-cam.com/video/jUFbY5u-DMs/w-d-xo.html). Hopefully that answers your question.
And "white noise" is another way of saying "i.i.d. noise". I explain what i.i.d. means in the video. The reason it is called "white" is that an i.i.d. variable has an auto-correlation function that is an impulse (delta function), which means its power spectral density (psd) is flat. A flat psd means that all frequencies are equally represented. This is analogous to white light, which is made up of equal amounts of all the colours / visible frequencies.
Thanks for the suggestion. It's great timing, since I'm just about to make a video on the topic of Noise Power, and explain that all "real" noise is band limited (ie. filtered). White noise, with its infinite power, doesn't really exist (we can't have infinite bandwidth - not until we find a way to have things travel faster than the speed of light, anyway).
I make my videos on topics that I feel are not well explained in the existing literature/textbooks. You might like to check out my other videos on the channel, and let me know of any particular topics you'd like me to cover, if I haven't already. Check out the full list here: iaincollings.com
Wow great video! In the world of so called professors, just read variables x's y's in their slides and leave their students in sheer confusion, this is how anyone should teach.
Glad you found it helpful.
Well said.
Great video. I got one question:
- What about the noise that is added during transmission of the signal? Is it also a gaussian noise?
Great question. Yes, but since the transmission power is high, the noise at the transmitter is insignificant in comparison, so we ignore it in our models. The receiver only collects a small fraction of the power that was transmitted (eg. in wireless communications it radiates out in all directions), and so the noise in the receiver's electronics can't be ignored. Glad you liked the video.
Quick question: the equation y=hx+n assumes that AWGN is added after the channel. what if part of the AWGN is caused by the transmitter electronics? then we will end up with y=h(n1+x)+n2, where n1 is the AWGN before the channel and n2 is the AWGN after the channel. We just assume that due to the central limit theorem, all the noise is modelled as a single AWGN value and added at the receiver. The question is: how do we model seperate noise sources? are there any reference where noise is added before and after transmission?
PS: the same discussion has taken place in the previous post....
Good question. No, we don't add the transmitter and receiver noises and assume it's OK due to the central limit theorem. Actually what happens is we simply ignore the transmitter noise. In most communication systems the transmit signal power is much higher than the transmitter's noise power, so the transmitter's noise has virtually no effect on the transmitted signal. At the receiver it's a different story, since the transmit power has dissipated by the time it reaches the receiver, so the received signal power is much lower, and is often comparable with the receiver noise power.
Thank you for your reply. I guess the confusion comes from the way we model baseband communications, ie y=x+n, which implies noise is added onto the transmitted signal. In reality we should think of (x) as the transmitted signal seen by receiver.
The noise is gaussian because of central limit theorem
Wow great explanation..❤️
Thanks, I'm glad you liked it.
thank you sir, love you from Vietnam 🇻🇳
I'm so glad you like the videos.
Thank you for the quick summary
You're welcome!
You said that when we have stochastic process (here with the elactrons) with many realizations, we get a GAUSSIAN shape of the probability density function. Is there a special theorem that states that? Thanks
This video explains it: "What is the Central Limit Theorem?" th-cam.com/video/Xd0_kez9smk/w-d-xo.html
In Gaussian noise pdf curve the speediness of the curve occurs from which parameter. Explain the parameter and give details of the curve with help of example.
Thanks Iain
Great video. Ty
very clear thanks .I will watch all your videos
Glad you found it useful.
hello lain ,
intigrator can be also viewed as low pass filter how can we relate this low pass filter operation with the collection of energy (the task performed by intigrator at the receiver)
An integrator is not a low pass filter. People sometimes say that it acts like a low pass filter, but that's just confusing, since the gain of the integrator at f=0 is infinity! That's definitely not something you would call a "filter". You can say that high frequencies do not affect the output of the integrator, but that's not the same as saying it is a "filter".
Good explanation, Gaussian process and white noise comparison video please, is there any connection between those two ideas, please explain sir ..
Thanks for the suggestion. I'll put a "white noise" video on my to-do list. In summary though, "white noise" is another way of saying "i.i.d. noise". I explain what i.i.d. means in this video. The reason it is called "white" is that an i.i.d. variable has an auto-correlation function that is an impulse (delta function), which means its power spectral density (psd) is flat. A flat psd means that all frequencies are equally represented. This is analogous to white light, which is made up of equal amounts of all the colours / visible frequencies.
@@iain_explains,
Sir I have a doubt,we have an iid random variable ,but it has no width in time,it is discrete in nature,in the case of a random variable with N(0,1) we have a white noise only when it is coupled with a continous func,to make to fit all together,other wise we may need to take DTFT to find psd,I'm I right sir,whether my concepts are correct..
@@vaishnav4035 OK, so the be even more precise, the noise in a digital communication system is a discrete time Random Process where the sampling is happening at the symbol rate of the digital signal (and where each sample is a Gaussian Random Variable).
I've just uploaded a video on this topic which will hopefully help: th-cam.com/video/QfUQMzHfbxs/w-d-xo.html
The integrator ouput is no longer a function of 't'. How do you sample at t = kT?
Well, yes, strictly speaking the integrator just keeps integrating, so technically I should have written the integral limits as being from zero (or negative infinity) to t, and I should have used a different variable for the integration variable (perhaps tau). But doing that can be confusing to people who are not so mathematically minded, so generally the integral in the detector is written the way I wrote it - ie. for just the first symbol (which is over the time period from 0 to T).
wow dude this is solid, keep on
Glad you found it useful. Any topics you'd like new videos on?
What is therm 'n' actually, is it the frequency, is it the amplitude. I am trying to understand this in terms of DSP though, and one thing more, how can we relate white noise to gaussian noise? Please help me understand this
Thanks for your question. OK, so a couple of things to clarify first: n_k is the "noise" that is disrupting the measurement of the k-th data bit, which is measured at time kT. This is a number. It is a voltage. It is random. It could be positive or negative. Now, in the video I have used "n" (without the subscript) to represent a general n_k (ie. for any value of k). The plot shows the probability density function (pdf) for n. If you're not sure what a pdf is, then please check out the video I made on pdf's (th-cam.com/video/jUFbY5u-DMs/w-d-xo.html). Hopefully that answers your question.
And "white noise" is another way of saying "i.i.d. noise". I explain what i.i.d. means in the video. The reason it is called "white" is that an i.i.d. variable has an auto-correlation function that is an impulse (delta function), which means its power spectral density (psd) is flat. A flat psd means that all frequencies are equally represented. This is analogous to white light, which is made up of equal amounts of all the colours / visible frequencies.
I've just uploaded a video on this topic which will hopefully help: th-cam.com/video/QfUQMzHfbxs/w-d-xo.html
Thanks for sharing! Very useful!
Glad it was helpful!
very well, thank you!
Does this gaussian shape imply that there's high possibility for lower amplitude noise
You might like to check out this video about p.d.f.s th-cam.com/video/jUFbY5u-DMs/w-d-xo.html
Thanks for the video
I'm your new subscriber
Glad you found it useful.
Thanks for the beautiful presentation. can you make a video presentation on color and white noises? I subscribed your channel
Thanks for the suggestion. It's great timing, since I'm just about to make a video on the topic of Noise Power, and explain that all "real" noise is band limited (ie. filtered). White noise, with its infinite power, doesn't really exist (we can't have infinite bandwidth - not until we find a way to have things travel faster than the speed of light, anyway).
Thank you kind sir.
You're very welcome
Great one ! But where do we find these insights what would you recommend us to follow(and also what you are following) for these intuitions
I make my videos on topics that I feel are not well explained in the existing literature/textbooks. You might like to check out my other videos on the channel, and let me know of any particular topics you'd like me to cover, if I haven't already. Check out the full list here: iaincollings.com
Greate video
Glad you liked it
Brilliant as always
Thanks. Glad you found it useful.
cool!
Voice out put is low
I hope i can speak english well to understand your lecture