and um so a graph from everyone and things but anyway style to are representing my work on a turn rate only phase just S and elements rician scheme for sampled data the conversion systems and this work has been done at a university of washington seattle um with michael you ten can which a ring and permission might by of that bit out as a motivation of my talk call present a do we need this yes is good and a lot so that its what's a role of um dish the uh D S B about terms for than about circuits and there and we'll go with the brief or real but it easy to a you that we have selected which in this case would be a segment of thirty C architecture then will propose a low and the lms calibrated to the analog and and then after that will look into that is to back and use be regarded them which uses is under rate and the missed estimation for and finally will complete so and the requirement for as we going to more and more seven and a meter process so you can process we find that a life of an analog design a is becoming harder and harder because for even simpler blocks like operational amplifier you find that the voltages is that shrinking you are strings are reducing and so the challenges is becoming higher and hard to reach a higher again at the same time for the should design a is becoming a relatively easy you have a much higher sampling speed and it can be that um you can really a not achieve i to but so that the basic motivation for this work is this to use a core design approach um a to be use that dish to the S got tends to kind of correct then a lot of functions that we can so before or we step into they at C architecture it is a very general overview of different type of D Cs and this architecture is just a sorry the still clutching is actually a a just compare is actually it for a low to medium band with application and for high resolution applications on a so it's a very generate go we just to give an idea of we stand and if you look we have different types of a arctic for a T Cs like flash i line successive approximation just tear and segment the at C and they have been compared for different parameters like throughput resolution solution they can see power here and for like for low to medium and at that location and high resolution you find a sigma delta a C a one of the most of a lot to that we can have a and so you know use we are actually targeting like for wireless lan uh approximation of like nice i like that and eleven it affected them before we step into the at C are look into that if see architecture the some critical implementation choices is that we should we need to decide and to of the definitions that we have a as for or something ratio which is defined as as a ratio of sampling frequency or what twice a signal time no what's simple of this number is is that i do a so you find that you can get a high uh at the same time for white that of that iterations it means that we need a a has in frequency because if a is increasing it's got at first increases as well so here in this design the focus more on maximizing minimizing the power rather than maximizing this P and so that's of reason real select a low sir of one eight for was design and the question that we want to answer is how little can we should be in of a pen like a is it why able to use like around ten a can be enough ten or twenty and so will will like is twenty six db gain or of easy but generally like a sigma don't applications the are in the range of like fifty to seventy db which is a to achieve a as we use a scaled down the supply so we'll see that if such a thing just yeah use a to be all them uh oh and that such a again if you try to use the in and bold is in the system can you can you can be them which today and and the whole uh process we power yeah fish before we four still uh so this like of a simple um evil you all a second order see my that the U C to large so that i made a C it can be just plus five in terms so it's a two input one our system so these are like to uh a time samples sampled input and then you get a a a a two a a and it can be it in terms of runs functions well the signal process S T F and noise times you so will will to of this out and a be five so if you look into the right side here so so a large score the V four oh no that's what we is thing but is just it's of C you take that out of the sum and you give it to the to integrate is do of one two let's signal that someone O G N S one year time being so this is there since that to integrate as well i not as the we take the output of the someone and we some all the all a and you give it to a for quantizer for like to D C and the we do a C is then back to at least a to compare compared oh sorry and by to which is then some time they have to the right is of the still a for all us as you can see from the from the signals signs here on the Y axis is the a is a range of and you hear and the X is just a times so you see that the was we one and B two have a very small signals planes which means that we do i mean it to use as the requirements for the for the integrate is that we have a all and it also the using this string mean that's so and using a and ones on the entire assist or that that's one of the pretty good advantages of having a a for long the second row that if you can the process from input of the quantizer to a the the second integrate two we find that the last is just a quantization noise so we'll see what's a man to this is to this feature in the next i but is the two things that's well and so for this system we can and a like the signal as a function which is the cost from the input that to but skis and the also and to yeah i post a given by a second one scene if you like to do not at here for we find while see that all put yeah training but position of two done X of the month like of the signal process okay so what implies lies is as the input goes to all without a frame and the one position on is not filtered by a signal i um so you can think of this as like it basically a portion that in but quantisation noise with to go that and in that was is it is a resolution of eighties the next question that the we have to so is is this system sufficient for for a requirement of realising a one bit snr or S india so think is like this system but then always are of eight can realise and nine but that's the not so we need an additional two bits so what should we do so what we do here is we we into it to a large you just on a stochastic signal at at at C so it's very simple like take the system from the previous slide we are another system a and then to do that what we do is at the input of this second stage we just stick you would you all Z you minus two from the previous it and added to the input of the second stage now feed you it in a more or any uh a a little bit more mathematics a and if you yeah to digital filters S of two D which actually matches uh a that a lot signal as a function of the second stage and then T of one D which match is the noise a function of the first stage and we some though two why of C then after to doing some at which we go now we find that the we can get a higher order of my thing which implies a high resolution ones uh let's just finish the system here and so once we get this out this is just a decimation estimation for to and it that's mixed down by eight as you can see a three state to to the first as the cascade to get the call to just but you for that about how when only face but to the the estimates it's by two for by a single uh a five compensation for but isn't it split into three state is is just to my as the power for let's look into the summation here and you see here is that the input signal X of Z C just a a segment as a function and as we saw the previous slide the signal process and is you need so the signal that signal X of C goes out with that you need a second term here which is a forced one quantization noise as you in this is a ideal system would find that if you assume that the and T F one of me is equal to T of one of the and S you of two of these equal to have two of this at this and sits out and so there's is not one position on the first to from position noise doesn't come the total component reduced you to of C sees a four door that a function here and it effectively means that now we are able to noise shape by a fourth order so that's it's it's a good thing in the sense that we have we can just take to stable second out of trance that systems put them together and get a higher order a a higher or the noise shaping which implies a a high snr what what the disadvantage of the system is this one will be in a cyst uh it is the design it will introduce mismatch between the and a lot a fancy chen and that is to function so how does that come to the house so as we see you know like for a very high gain or and we see that the noise or or for uh a noise of the system is done to but i'm long so here here it's six this is the signal back from zero to ten made and on the bias the or like to rent C V D so right now we can see that for a very large gain it's done it but um and noise but at that it use uses a pretty six db the one position noise starts don on on the right to easy you why that's a reason so we see that that approximation emission that we came a are cross across year one minus the was where all of it true because that are outside as an iir are filter because of that that's that's thing and the gain of whatever "'cause" the system that's to right so what happens is that one position as of the first stage weeks without it and it produces by uh has an impact on a phone so what we do is we take the help of and of the audit to really or a the performance impact that we have so i'll go to the the detail in the blocks of the plots in the next slide what in this like we just want to get a general overview of what we want to do so what we do is we deactivate the input and we inject at to to do great no no no C which is generated on to from a linear feedback shift register at the same time we take the sequence to uh to a set a file for this uh i T have to the in in of one B the output of this a five for is compared with that a with that and a lot of by one of the Y to see and then we use that a missing write them to get you can be the distance a so right so to summarise rate yeah actually using sign to the ms one for a listener was it's one of the most easiest to implement that are most of if and it's it's from you got to maybe compute the next year or efficient from the previous state use that the patient size you know as there if getting in the design and uh and that it right C we take the sign of yeah yeah that the way is about the works is it each be because of the operates to it finds that error is minimised that a miss things and the coefficients is actually get a response to the by so as a function of the signal doesn't that we expect on the so after the long i do calibration you see that the to uh signal thus this same as that of one and for what they stiff and then you oh there is much what do you know how we implement the uh the filters so we inject the thirty two bit to a L that it and m-sequence a thing yeah screen and rate to five to the output of this a five is that can be a the output that you see why the Z Y two Z and there is then you meant to have a isn't going and the coefficients of an adaptive guys so to to use the truncation errors in there five for it is use an eight to five one disadvantage that the system them can have a is that you in for a set of the asians we are actually sound the system i a sampling frequency of the N T one fifty two hundred meg a what as is that you do you have a very low or and because now we are using very simple simple fires and optimising and the oh a channel addition but three can approximate three times if you do a a force that as compared to that so what should we do how should be uh uh or should be in signal processing techniques here to make sure that you get a good efficient says what we do a is in that to mention the police that you just about oh we are using a noise cancellation the also which is same as basically C the N T F and S T E uh yeah on again as and C S and the are basically using that a as we talked in the previous slide no but ones get a is given decimation fig which on so the proposed approach what we do it we use one these techniques do really to do a convolution of the noise the noise cancellation fit and india for this stiff yeah and that is for that and get a what if easy then we do the same per we reject the a random sequence we get the desired signal and B J the all output of the four and a stage and we make sure that sees the C and then compared what the stages do you do in N and now this approach is very similar to what a good subband adaptive filter but one pretty good point two that we are using or sampling in this case so using or so i it tells us to use the you using right and it helps service to ensure that of that the white one of C oh which we used to basically uh this you estimate is not a very aggressive anti aliasing filter and it just has to be very many so that this approach for a or so as well so that is the proposed a one some and the the system that we show that us like reject it and see here we are a a lot of the estimate it and we lost a if it's like filter take out right the same time we take the out of that of well as the this yeah so for or or in this this meeting but a of this right to that it are and take their compute coefficient so we do that for both the parts of are here and finally we can speak good thing about this is we found that but the meant in a similar it's you using this to two at which means that you can re read to not to my based on the whole system in innovation um do does have a small one right because of the increase in the number of taps of you and C T E M Z but was that have can is based on and it can he's the optimized when we do not you issue we friend the power decreases when you decoding case oh that's television we shouldn't a points the number of points in or are pretty much the same for all all the different techniques so is proposing but the system that we have yeah we have that a lot yeah we we see that using the low you know that introduces some some in what errors which is a mismatch in the and and and to process function and here as the but different flavours of the back and that we have a and to again an F P G and we got and mention that a misty with the one face but to a and what if is better or there is a they should signal to noise ratio in of you of this nation so without any yeah a missing uh without any kind vision find a signal to noise ratio yeah as actually don't around fifty six points db after calibration uh we see that all the three seems give hear a D or form eleven bits oh six is a T V C i i C six i um so here is a matrix fixed calibration sampling frequency V using as one have oh of the option that i i what what or supply voltage one point two five or to a spare i meant one had in a seamless as a comparison the it's uh a a here is the forced to be so we use like a in a six db but it's just a lot of them to yes and before that up to my the rest of the cost of i'd a reference for years and actually a uh and the quantizer and i think this and the number can be really optimized to last what's for a forty five design yeah we uh as compared to the other the words we see that we have a can't uh a or savings for that for the by we do a approximate analysis of finding the power for years so the use that are i see that uh you a power as it can be at fourteen time uh a small errors G i become got this the actions of you for many words what is you get on that a before for the demise as this so we do not that it at your had because of some for does that you have to put to match it so is a final completion format um are don't was to the use of a six E uh you know sees you said that addition white them to go go from the gain errors and uh we proposed something techniques which uses like month at it uh are it important is only polyphase i and we find that using this you that as in using at or the machine especially for high fit by thank you if you have any question sure oh yeah also yeah we compared to the convention the then use i don't i a moving bad to are of people yeah the whole system can see uh okay okay i just space and min yeah i would say like a uh i think though the presence and this project has been on like a design between that a lot in by so if you compare the uh i figure of matter it's with the current architecture use one just the an up front and that we have it's actually not uh it's not as good as scum it's and the ball but but it's not as good as the best ones and there but the more mean as is here is how can you really meant like a back and when you have a like a very cheap and a i oh i oh thank you