Nice video as always. The bandwidth trick does need some more context I think. Yes, the total amount of noise is reduced. The downside is that the signal to noise ratio for lower frequencies is still as bad as before. It's something people seem to overlook sometimes. Especially when the noise is still causing issues at these lower frequencies for data acquisition etc.
You are right Piet! By reducing the upper part of the bandwidth will only reduce the noise in that area of the spectrum. Honestly I haven't worked with data acquisition systems that much.. how would you recommend reducing the noise for such circuits?
@@FesZElectronics Well, maybe a bit boring/lame answer, but that really depends on a lot of things. It depends mostly on the requirements. Are you working with sensors or audio, noisy environment, as well as requirements of being able to "detect" or distinguish a certain signal (SNR)? One way is to use a better opamp, or work with differential signals. Another way is using a DSP or something to add some mathematical tricks. There are some (analoge) tricks to invert the noise signal and sum it with the original. Sometimes it's even about choosing your gain properly and maybe even using two (different) sensors to cover the whole spectrum. In some cases also PCB and ground layout can be key factor. So it really depends where the noise is coming from and WHERE and WHAT you can compromise. In the end that also depends again on the budget.
An interesting method I've seen in some LT application notes is not just to use a "good" op-amp but multiple identical parts in parallel, this they say reduces the noise by a factor of sqrt(nr of parallel elements). Another thing they did was build a discrete input stage with very low noise parts to improve the performance of said "good" op-amp. Both these techniques are used in this document: www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf I'm curious, did you run into any of these methods in practice? I mean they look nice in the app note but seem quite cumbersome to be practical and I haven't seen this implementation anywhere else but maybe I just didn't search enough.
@@FesZElectronics Noise reduction by paralleling devices only gives a 3dB reduction (10*log(x) ) in noise. So to really get anything significant you quickly need something like 4 or 8 devices (6dB to 12dB reduction). By that time you lost so much real estate and budget that it's better to go for a different opamp. But that also depends where the noise is coming from. Sometimes it's actually better to make a differential circuit to just completely cancel the noise out. A discrete input stage could work. Although nowadays there are quite some affordable opamps with a similar or better noise specs. Advantage is that they are easier to work with and require less research time to get the same results.
@@FesZElectronics unfortunately there is no 1 solution to fit all. It depends. generally speaking, important signals should be differential, and if low frequency accuracy is desired, harmonic up conversion is a nice trick to reduce noise. Use of Noise Matching networks is also a nice trick for high frequency noise. But again, At the end of the day, It really depends on where the noise come from, how should the signal be processed and how noise is distributed on the spectrum.
I'm happy my videos helped you! If you are interested in supporting the channel, I do have a Patreon page set-up, there is a link at the end of every video.
I can't find the way to measure nois eof paralleled active devices like JFET sor BJTs. Putting 2 in parallel, the noise drops about a factor of SQR(2) about .. but what I got, is just the sum of the 2 noises. ... certainly I'm making mistakes into the measure in LTSpice but I don't find any help on this.
Hello! I got a ¿? I'm using LTSPICE, I need an operational amplifier but the ones LTSPice give have negative at the top and positive input at the top also, and I need one with both positive input and positive voltaje input at the top, how can I edit the one there? Please help me, you domain LTSpice.
Hello. Well, the placement of the terminals will only have a cosmetic influence; I mean you can set your +supply on the bottom if you wish. Anyway coming back to your question, the opamps with LTspice built in models are not editable but you can use the symbol for "universal Opamp" and edit the symbol, so to swap the location of certain pins. Or you can create a symbol from scratch to look how-ever you want. Just make sure the netlist order of the pins for any new or edited symbol stays the same (pin1:in+ pin 2: in- pin 3:V+ pin 4: V- pin 5: out) Let me know if this works out.
@@missgalaxy6519 You insert it into your schematic, and then right click - a new window pops up and there go to "Open Symbol"; edit and save (maybe use a different name like OpAmp2INV so its different from the default)
@@FesZElectronics Ok, I'm there. I have: Attribute Value Vis Prefix X X InstName U2 Spice Model level.2 Value Value 2 Avol=1Meg GBW=10Meg Slew=10Meg SpiceLine ilimt =25m rail=0 Vos=0 phimargin=45 SpiceLine2 en=0 enk=0 in=0 ink=0 Rin=500Meg Where is the section to edit?
I guess it depends on the model - based on the exact purpose a model can be more or less complicated and take into account more or fewer characteristics; I would recommend a very basic circuit just to test the model before going for a more complex design.
That's because the instantaneous noise value is fully random and therefore follow a gaussian distribution. The RMS of a noise signal is basically the standard deviation (Sigma) of the gauss graph. Statistically, X-factor of Sigma gives you the probability of something happening. 6 times Sigma or 8 in some literature, give you the worst case scenario possible. Hence the 8 times RMS magic number.
I use them both, for very different purposes. Microcap isn't the easiest program to use, but can be very powerful. I personally use LTSpice more often for quick simulations, for more in debt research, or very specifics comparisons (especially with graphs), Microcap is much better.
@Mai Mariarti Both run totally fine under Wine. I really don't see why it would be barbaric? You can't expect that every vendor/company is going to compile his software for every thinkable platform.
I honestly haven't used the Hobo OS :)) although I am fully aware its better than Win, its just that I don't really want to spend the time to learn to use it. I guess its the same with Microcap for me, at the moment I just want to focus on learning LTspice; the best tool in general is not the one with the most features and best performance but rather the one you actually know how to use, that's my opinion on this at least.
I think this is the weirdest New Years Resolution I ever hear, but I guess its legit. I'm not sure why LT (now analog) decided not to make LTspice Linux compatible, I mean its a pretty bear bones simulator, it doesn't have any fancy bling bling graphics and interfaces, just the minimum you need to run a simulation well and fast; and I guess this is the same philosophy behind linux also - keep things simple and good, so the 2 should work well together. Oh well, maybe in the future LT will also be linux compatible...
The result should be the same, just the exact calculated frequencies and the way in which the data is represented may be different; with a linear simulation, by default the data is plotted on a linear scale, and with decade, a logarithmic scale is used; you can however change the scale type by rightclicking on the X axis and selecting the "logarithmic" checkbox.
All videos are great! Thank you! Can I ask if we can use LTSpice in any communications system for experiment like in Transmission media and antenna system?
Thank you for the nice video. I have a question about the comparison of the results obtained from LTSPICE, with one from theoretical formulae for resistor noise. Usually, voltage noise density is given by = sqrt ( 4KRT) where we have Boltzmann constant, temperature, and resistance. But somehow, My voltage density values obtained from simulation and formula are not matching. For 1 k Ohm resistor, it should be around 4nV/rtHz but from simulation we can see that its below 3 nV/rtHz.
I'm not sure what the problem is - I tried to add a 1K resistor connected to ground and I get a 4.07nV/rtHz density as expected. My circuit was (V1 added just to make the simulator happy) R1 N001 0 1k V1 N002 0 1 .noise v(n001) v1 dec 100 1 2 Maybe you can give more details on your circuit?
@@FesZElectronics Hi. I was simulating the simple voltage divider available in the video at 03:54. If we connect both of them ( resistor and power supply) to ground, as shown in the video at 03:54 we can obtain the noise voltage as 2.88 nV/rtHz at the output node with the contribution of R2 as 2.035 nV/rtHz ( as shown in your video). Now, if I make the R1 noiseless, I obtain 2.035 nV/rtHz at the output node which is all coming from R2 of course. However, ideally, I think that we should have 4.07 nV/rtHz at the output node because now only R2 ( 1kOhm) is causing the noise and according to V=sqrt(KRT), we should have 4nV/rtHz for R2=1kOhm. Therefore, to obtain the 4.07nV/rtHz, which is the ideal value I think we should only connect the power supply to the ground and leave the R1 bottom side as an open circuit. ( In other words, there should be only one ground in this voltage divider shown at 03:54. HOWEVER, this is what I think. Please correct me, as I might be wrong. ( I am not from electrical background )
![ตีสิบเดย์ [FULL] | นางเอกหมอลำเงินล้าน "อุ๋งอิ๋ง เพชรบ้านแพง" & ไฮโซสาวสุดฮอต "ลิลลี่ เหงียน"](http://i.ytimg.com/vi/jPxQS_EgrJ4/mqdefault.jpg)
I started watching you to pass the time at uni, now Im using your tutorials to do my job, thank you for your efforts man :)
the gem of circuit simulation using spice on whole youtube
Great video Fesz. The way you explained ways to decrease the noise was very intuitive. Great work. Keep making more of these videos. Love you!
Great presentation, great knowledge 👏🏼
Brilliant, Thanks Fesz
I'm happy you enjoyed it! Thanks for watching!
si-a aratat fataaaa!!! Bravo, Feri! Craciun Fericit! :)
deja de video-ul trecut... Craciun Fericit!
Great playlist! Thanks a lot!
Nice video as always.
The bandwidth trick does need some more context I think. Yes, the total amount of noise is reduced. The downside is that the signal to noise ratio for lower frequencies is still as bad as before. It's something people seem to overlook sometimes. Especially when the noise is still causing issues at these lower frequencies for data acquisition etc.
You are right Piet! By reducing the upper part of the bandwidth will only reduce the noise in that area of the spectrum. Honestly I haven't worked with data acquisition systems that much.. how would you recommend reducing the noise for such circuits?
@@FesZElectronics Well, maybe a bit boring/lame answer, but that really depends on a lot of things.
It depends mostly on the requirements.
Are you working with sensors or audio, noisy environment, as well as requirements of being able to "detect" or distinguish a certain signal (SNR)?
One way is to use a better opamp, or work with differential signals. Another way is using a DSP or something to add some mathematical tricks. There are some (analoge) tricks to invert the noise signal and sum it with the original.
Sometimes it's even about choosing your gain properly and maybe even using two (different) sensors to cover the whole spectrum. In some cases also PCB and ground layout can be key factor.
So it really depends where the noise is coming from and WHERE and WHAT you can compromise.
In the end that also depends again on the budget.
An interesting method I've seen in some LT application notes is not just to use a "good" op-amp but multiple identical parts in parallel, this they say reduces the noise by a factor of sqrt(nr of parallel elements). Another thing they did was build a discrete input stage with very low noise parts to improve the performance of said "good" op-amp. Both these techniques are used in this document: www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf
I'm curious, did you run into any of these methods in practice? I mean they look nice in the app note but seem quite cumbersome to be practical and I haven't seen this implementation anywhere else but maybe I just didn't search enough.
@@FesZElectronics Noise reduction by paralleling devices only gives a 3dB reduction (10*log(x) ) in noise. So to really get anything significant you quickly need something like 4 or 8 devices (6dB to 12dB reduction). By that time you lost so much real estate and budget that it's better to go for a different opamp. But that also depends where the noise is coming from. Sometimes it's actually better to make a differential circuit to just completely cancel the noise out.
A discrete input stage could work. Although nowadays there are quite some affordable opamps with a similar or better noise specs. Advantage is that they are easier to work with and require less research time to get the same results.
@@FesZElectronics unfortunately there is no 1 solution to fit all. It depends. generally speaking, important signals should be differential, and if low frequency accuracy is desired, harmonic up conversion is a nice trick to reduce noise. Use of Noise Matching networks is also a nice trick for high frequency noise. But again, At the end of the day, It really depends on where the noise come from, how should the signal be processed and how noise is distributed on the spectrum.
thnx
@FesZ Electronics Why do we multiply by 6 (or 8) ?
I think it was some "rule of thumb" I found in some documentation source; its just a good approximation
Dude, is there any way to sponsor you? Your videos have helped me a way that I must help you back
I'm happy my videos helped you! If you are interested in supporting the channel, I do have a Patreon page set-up, there is a link at the end of every video.
I can't find the way to measure nois eof paralleled active devices like JFET sor BJTs. Putting 2 in parallel, the noise drops about a factor of SQR(2) about .. but what I got, is just the sum of the 2 noises. ... certainly I'm making mistakes into the measure in LTSpice but I don't find any help on this.
Hello , Great video .
Aren"t we able to plt SNR (dB) in LTSpice ?
What does the values on the y axis represents exactly?
Hello!
I got a ¿?
I'm using LTSPICE, I need an operational amplifier but the ones LTSPice give have negative at the top and positive input at the top also, and I need one with both positive input and positive voltaje input at the top, how can I edit the one there?
Please help me, you domain LTSpice.
Hello. Well, the placement of the terminals will only have a cosmetic influence; I mean you can set your +supply on the bottom if you wish. Anyway coming back to your question, the opamps with LTspice built in models are not editable but you can use the symbol for "universal Opamp" and edit the symbol, so to swap the location of certain pins. Or you can create a symbol from scratch to look how-ever you want. Just make sure the netlist order of the pins for any new or edited symbol stays the same (pin1:in+ pin 2: in- pin 3:V+ pin 4: V- pin 5: out)
Let me know if this works out.
@@FesZElectronics hello!
Im gonna try it ;)
Thank U 😊👍
@@FesZElectronics another ¿? How can I edit the Universal Amplifier and switch the possitive input?
@@missgalaxy6519 You insert it into your schematic, and then right click - a new window pops up and there go to "Open Symbol"; edit and save (maybe use a different name like OpAmp2INV so its different from the default)
@@FesZElectronics Ok, I'm there.
I have:
Attribute Value Vis
Prefix X X
InstName U2
Spice Model level.2
Value
Value 2 Avol=1Meg GBW=10Meg Slew=10Meg
SpiceLine ilimt =25m rail=0 Vos=0 phimargin=45
SpiceLine2 en=0 enk=0 in=0 ink=0 Rin=500Meg
Where is the section to edit?
Hi. What about the current noise of OpAmps? Does LTSPICE take it into account or do you need to do something special?
I guess it depends on the model - based on the exact purpose a model can be more or less complicated and take into account more or fewer characteristics; I would recommend a very basic circuit just to test the model before going for a more complex design.
Around 13:40, why did you multiply RMS noise by 8? That seemed like it came out of nowhere?
That's because the instantaneous noise value is fully random and therefore follow a gaussian distribution. The RMS of a noise signal is basically the standard deviation (Sigma) of the gauss graph. Statistically, X-factor of Sigma gives you the probability of something happening. 6 times Sigma or 8 in some literature, give you the worst case scenario possible. Hence the 8 times RMS magic number.
Great video, by the way Micro-Cap is free now, would be interesting to see the direction thing will go in the future.
I use them both, for very different purposes. Microcap isn't the easiest program to use, but can be very powerful.
I personally use LTSpice more often for quick simulations, for more in debt research, or very specifics comparisons (especially with graphs), Microcap is much better.
@@p_mouse8676 totally agree.
@Mai Mariarti Both run totally fine under Wine. I really don't see why it would be barbaric? You can't expect that every vendor/company is going to compile his software for every thinkable platform.
I honestly haven't used the Hobo OS :)) although I am fully aware its better than Win, its just that I don't really want to spend the time to learn to use it. I guess its the same with Microcap for me, at the moment I just want to focus on learning LTspice; the best tool in general is not the one with the most features and best performance but rather the one you actually know how to use, that's my opinion on this at least.
I think this is the weirdest New Years Resolution I ever hear, but I guess its legit.
I'm not sure why LT (now analog) decided not to make LTspice Linux compatible, I mean its a pretty bear bones simulator, it doesn't have any fancy bling bling graphics and interfaces, just the minimum you need to run a simulation well and fast; and I guess this is the same philosophy behind linux also - keep things simple and good, so the 2 should work well together. Oh well, maybe in the future LT will also be linux compatible...
why when i do linear instead of decade i get different results of noise?
The result should be the same, just the exact calculated frequencies and the way in which the data is represented may be different; with a linear simulation, by default the data is plotted on a linear scale, and with decade, a logarithmic scale is used; you can however change the scale type by rightclicking on the X axis and selecting the "logarithmic" checkbox.
All videos are great! Thank you! Can I ask if we can use LTSpice in any communications system for experiment like in Transmission media and antenna system?
You have a built in model for transmission lines, but this is not really a simulator for fields - so its not really made for antennas...
Thank you for the nice video. I have a question about the comparison of the results obtained from LTSPICE, with one from theoretical formulae for resistor noise. Usually, voltage noise density is given by = sqrt ( 4KRT) where we have Boltzmann constant, temperature, and resistance. But somehow, My voltage density values obtained from simulation and formula are not matching. For 1 k Ohm resistor, it should be around 4nV/rtHz but from simulation we can see that its below 3 nV/rtHz.
I'm not sure what the problem is - I tried to add a 1K resistor connected to ground and I get a 4.07nV/rtHz density as expected.
My circuit was (V1 added just to make the simulator happy)
R1 N001 0 1k
V1 N002 0 1
.noise v(n001) v1 dec 100 1 2
Maybe you can give more details on your circuit?
@@FesZElectronics Thank you so much for your response. I found the error. It is working correctly now.
What was the error?
@@FesZElectronics Hi. I was simulating the simple voltage divider available in the video at 03:54. If we connect both of them ( resistor and power supply) to ground, as shown in the video at 03:54 we can obtain the noise voltage as 2.88 nV/rtHz at the output node with the contribution of R2 as 2.035 nV/rtHz ( as shown in your video). Now, if I make the R1 noiseless, I obtain 2.035 nV/rtHz at the output node which is all coming from R2 of course. However, ideally, I think that we should have 4.07 nV/rtHz at the output node because now only R2 ( 1kOhm) is causing the noise and according to V=sqrt(KRT), we should have 4nV/rtHz for R2=1kOhm. Therefore, to obtain the 4.07nV/rtHz, which is the ideal value I think we should only connect the power supply to the ground and leave the R1 bottom side as an open circuit. ( In other words, there should be only one ground in this voltage divider shown at 03:54. HOWEVER, this is what I think. Please correct me, as I might be wrong. ( I am not from electrical background )
Can u teach how to simulate ask, psk and fsk using LT spice plsss
You sound exactly like Agadmator, both in tone and accent. Even a slight resemblance. Are you guys related?
Not really... Sry