|
The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> RF Design >> TIA in passive mixer https://designers-guide.org/forum/YaBB.pl?num=1322490283 Message started by baohulu on Nov 28th, 2011, 6:24am |
|
Title: TIA in passive mixer Post by baohulu on Nov 28th, 2011, 6:24am hi all, in passive mixer, a common structure is switch+TIA,due to parasitic cap of the switch, the TIA feedback factor is small, maybe 1/10. so the TIA loaded bandwidth is small. however, in a receiver system, there is often a requirement like IIP3 for N+6, N+12 channel, if the channel bandwidth is wide, then TIA can't not have a high gain at 6*bandwidth & 12*bandwidth, then the IIP3 can not be good enough. I want to know how can the TIA be to have enough bandwidth with a good phase margin. I have heard that the TIA can use feedforward method to compensate , who have a detailed description or reference? thanks |
|
Title: Re: TIA in passive mixer Post by buddypoor on Nov 28th, 2011, 7:28am A TIA needs no extra compensation because it is always stable if the feedback resitor is selected properly (data sheet). I think, the bandwidth of a given IC cannot be enlarged. |
|
Title: Re: TIA in passive mixer Post by loose-electron on Nov 28th, 2011, 10:39am TIA? Acronym? TIA = Time Interval Analyzer - but what does it mean here? |
|
Title: Re: TIA in passive mixer Post by buddypoor on Nov 28th, 2011, 1:08pm loose-electron wrote on Nov 28th, 2011, 10:39am:
No, TIA means "Transimpedance Amplifier" - better known as CFA. |
|
Title: Re: TIA in passive mixer Post by aaron_do on Nov 28th, 2011, 5:06pm Hi, First of all, I guess you already know, but the intermodulation distortion is due to the amplification in the op-amp. I suggest you analyze your system very thoroughly. You may find that the switch capacitance or even the low-pass nature of the TIA has the effect of filtering the signal before amplification. In that case, the IIPn can be good even well outside the TIA loop-gain bandwidth. Especially if you turn the TIA into a LPF. Of course, increasing the power in your op-amp's output stage will also help. regards, Aaron |
|
Title: Re: TIA in passive mixer Post by baohulu on Nov 28th, 2011, 9:18pm aaron_do, I can't agree with you even when the TIA is connected with feedback res and cap, that is what you said "LPF type". if you have seen the input resistance of the "LPF TYPE" TIA, you will find that in high frequency, like 50MHz above, the input resistance is high, that means the gain of TIA is small (due to the small feedback factor), then the swing of the TIA input node is large, the large input signal will make the linearity poor, besides, a low loop gain can't help attenuate distortion enough. thus , TIA should accept efficient compensation to enlarge its bandwidth. aaron_do wrote on Nov 28th, 2011, 5:06pm:
|
|
Title: Re: TIA in passive mixer Post by aaron_do on Nov 28th, 2011, 10:35pm Obviously widening the loop gain bandwidth will improve the linearity at higher offset frequencies. I'm not disagreeing with that. However, I suggest that you properly analyze your system to check the out of channel IIPn before going to great efforts to improve your loop gain-bandwidth. The reason is because the out-of-channel signal may be heavily attenuated at the input of the op-amp. I've attached a diagram of what I'm refering to ((a) shows a normal TIA and (b) shows a LPF TIA). I haven't analyzed the Op-Amp input above the loop-gain bandwidth of the Op-Amp (that's why I said you need to check!). |
|
Title: Re: TIA in passive mixer Post by baohulu on Nov 29th, 2011, 2:05am aaron_do, about your figure, I have two questions to ask you 1. what is the difference between a normal TIA and a LPF one in structure, the LPF is formed at the feedback path with RC or just the TIA input node with CAP to VCM 2. what does the opam output line mean in your figure? the other two line are frequency response at opam output and the opam input with open loop, right? |
|
Title: Re: TIA in passive mixer Post by loose-electron on Nov 29th, 2011, 3:46pm I believe - He's talking about out of band signals where the linearity is compromised due to amplifier bandwidth, but the results are not consequential because the content is out of the passband and is being attenuated due to that. |
|
Title: Re: TIA in passive mixer Post by aaron_do on Nov 29th, 2011, 5:03pm Quote:
1. The LPF is formed by feedback RC. I don't know why you would try and form a LPF by putting a CAP at the input of a TIA. 2. opamp output is the desired final response. So in case 1 with a TIA, the output is broadband hence the stright line. In case 2 I want a LPF so I have a low-pass response. Based on the op-amp's open-loop gain, you can backwards calculate the signal that must appear at the input of the op-amp which is what the op-amp input line represents. Those diagrams are from a quick study I did a couple of years ago. I found that my TIA based LPF offered significantly better out-of-band IIP3 than the same TIA with out feedback RC. I reasoned that the out-of-band signal was effectively being attenuated before amplification. Based on some simulations I did, this analysis seemed to be true. However, it was a quick study and that's why I recommended right from the beginning that you do a thorough analysis on your own system to see whether you need more loop-gain bandwidth. Hi loose-electron, Quote:
Is "He" me or baohulu? What do you mean by "the results are not consequential because..." rgds, Aaron |
|
Title: Re: TIA in passive mixer Post by baohulu on Nov 30th, 2011, 1:25am aaron_do, I think we are saying the same thing. if you look at you figure carefully, you will find that at interference B and C frequency, you opam still has enough gain. your opam -3dB bandwidth is higher than your low pass RC corner. I think maybe your system signal has not that big bandwidth. if in TV system, the signal bandwidth is 8MHz if in LOW IF system, that means the low pass RC corner is higher than 10MHz, then the opam -3dB bandwidth is higher than 10MHz, if the opam has 60dB DC gain, then the opam unit gain bandwidth will be higher than 10GHz, this is not that practical in products. thus, the TIA needs very high unit gain bandwidth thanks |
|
Title: Re: TIA in passive mixer Post by loose-electron on Nov 30th, 2011, 11:02am Quote:
The comment about spectral content versus the bandwiidth of the amplifier. The comment was about the higher frequency content that does not sit within the bandwidth of the op-amp, and thus has poorer lineraity due to less feedback gain. "less consequential" should be a better way of saying it - if the out of band content has poorer linearity, but exists in the rolloff region of the filter, then it may not be as much of an issue, due to the fact that the results are attenuated. Have to look at the filter profile and how it interacts with the spurs, and quantitatively determine if its sufficient linearity. Essentially I am saying something similar to what Aaron said. Perhaps I should have just said: "Me Too!" :D Maybe we need a "Like" button, similar to Facebook? LOL! |
|
Title: Re: TIA in passive mixer Post by aaron_do on Nov 30th, 2011, 5:24pm Baohulu, Quote:
No we are not. You are saying Quote:
but I am saying, Quote:
They don't seem to be the same to me. Those impractical numbers you showed don't constitute a thorough analysis. You need to actually analyze the IIPn out of the amplifiers loop-gain bandwidth. Which, as I have said multiple times, may be good. One reason that the linearity can be good outside of the loop-gain bandwidth is because the dominant path to the output is directly through the feedback path (not through the opamp), which is a passive network. Think about it. Quote:
I drew the figure, so probably I was looking at it carefully when I drew it. I already mentioned, Quote:
Also, you read too much into my figure. Quote:
My figure is only illustrating that out-of-band interferers are attenuated before amplification in a LPF-based TIA. It is not trying to tell you how to design the op-amp or filter response. You have presented an arguement that you definitely need more loop-gain bandwidth. I am saying that you may or may not, and have presented you with a situation in which you don't need more loop-gain bandwidth. So now all you need to do is properly analyze your system to see whether you really need more loop-gain bandwidth. However, it seems that you are intent on extending the loop-gain bandwidth even if it means you end up overdesigning. So maybe somebody else can help you with that. Considering my signature (at the bottom), I probably should have given up on this discussion by now... Aaron P.S. If I have come across as impolite, I appologize. But I have the flu right now, and it is frustrating having to say the same thing over and over again. |
|
Title: Re: TIA in passive mixer Post by baohulu on Nov 30th, 2011, 5:50pm aaron_do, your method is efficiency only when your opam -3dB bandwidth is higher than your low path RC corner. if your opam -3dB bandwidth is lower, then the opam input figure is not what you have drwan, it is much like the left normal figure. so, your method is better than the normal one. but signal has ~MHz bandwidth, you need very high unit gain bandwidth the thing that I am always saying is that high bandwidth is hard to design |
|
Title: Re: TIA in passive mixer Post by aaron_do on Nov 30th, 2011, 6:53pm Baohulu, Quote:
yes. thanks for pointing that out. Quote:
Incorrect. Compared to the normal TIA case, the effect of adding the LPF response is the same regardless of whether the LPF transfer function or the op-amp open-loop gain has a higher bandwidth. Compared to the normal TIA case, the signal effectively sees the LPF response in front of the op-amp. Therefore the improvement is very significant. Quote:
yes. But it is not my method. I didn't invent the active-RC LPF. Quote:
Incorrect. You need to analyze your system to see whether you need to improve the loop-gain bandwidth. As you already agreed, the LPF TIA shows better (actually much better) linearity than the standard TIA. Quote:
that is true. And that is why its better to analyze your system to see if you really need that high bandwidth. That is what I was always saying. Also note that you have control over the TIA DC gain which will affect the linearity. You can always reduce the DC gain to improve the linearity. Since the output is filtered, the linearity requirements of the following stages will be greatly relaxed. Aaron |
|
Title: Re: TIA in passive mixer Post by wave3x on Dec 8th, 2011, 6:05pm Hi baohulu, I suppose you are a young engineer from China who are looking for the way to be a good designer. But firstly, you must be patient and be honest. Everybody here understood what aaron_do said, but you cannot??? Here, I have two questions too, 1. You have talked many times that you need to handle 8MHz IF bandwidth, what a wide bandwidth??? Indeed I have designed passive mixer with ten times IF BW of yours, and I followed a similar methedology that aaron_do mentioned, that really helps. 2. You use Low-IF Rx to design Silicon TV Tuner system??? That's really not a good choose, maybe you should debate this point with your system guy, but not waste your time here and make us frustrating. aaron_do just want to share his insight and experience with you. Sorry for my not answering your questions. You can mail me if you are still struggling, but definitely I haven't such a good patience as aaron_do held. BTW, I am also Chinese, and I knew so many designers at Chinese Mainland. They are young, smart, hardworking, graduated from top University, and continuously accumulated experiences. But many of them still need to learn how to be patient and modest, just as those their Mom told them once and once again. The world belongs to the youth, but will wait until they become matured. |
|
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |