0:00:14so a graph from everyone and things but anyway style
0:00:17to are representing my work on a turn rate only phase
0:00:21just S and elements rician scheme
0:00:23for sampled data the conversion systems
0:00:26and this work has been done at a university of washington seattle um
0:00:30with michael you ten can which a ring and permission might by of that bit out
0:00:36as a motivation of my talk call
0:00:38present a do we need this yes is good and a lot so that its what's a role of
0:00:43dish the uh D S B about terms for than about circuits and there
0:00:47and we'll go with the brief or real
0:00:49but it easy to a you that we have selected which in this case would be a segment of thirty
0:00:53C architecture
0:00:55then will propose a low and the lms calibrated to the analog and
0:01:01then after that will look into that is to back and use be regarded them
0:01:05which uses is under rate and the missed estimation for
0:01:08and finally will complete
0:01:13the requirement for as we going to more and more seven and a meter process so you can process
0:01:19we find that
0:01:20a life of an analog design a is becoming harder and harder because
0:01:24for even simpler blocks like operational amplifier
0:01:28you find that the voltages is that shrinking
0:01:30you are strings are reducing and so
0:01:33the challenges is becoming higher and hard to reach a higher again
0:01:36at the same time for the should design a is becoming a relatively easy you have a much higher sampling
0:01:42and it can be that
0:01:43um you can really a not achieve i to but
0:01:48so that
0:01:48the basic motivation for this work is this to use a core design approach
0:01:53um a to be use that dish to the S got tends to kind of correct then a lot of
0:01:57functions that we can
0:01:59so before or we step into they at C architecture it is a very general overview of
0:02:04different type of D Cs
0:02:06and this architecture is just a sorry
0:02:09the still clutching is actually a a just compare is actually it
0:02:13for a low to medium band with application
0:02:16and for high resolution
0:02:17applications on a so
0:02:19it's a very generate go we just to give an idea of we stand
0:02:22and if you look we have
0:02:23different types of a arctic for a T Cs like flash
0:02:27i line
0:02:28successive approximation just tear and segment the at C
0:02:31and they have been compared for different parameters like throughput resolution solution they can see power here
0:02:38like for low to medium and at that location and high resolution
0:02:42you find a sigma delta a C a one of the most of a lot to that
0:02:46we can have a
0:02:47and so you know use we are actually targeting like for wireless lan uh approximation of like nice i like
0:02:54and eleven it affected them
0:02:58before we step into the at C are look into that
0:03:01if see architecture the some critical implementation choices is that we should we need to decide
0:03:06and to of the definitions that we have a
0:03:08as for or something ratio
0:03:10which is defined
0:03:11as as a ratio of sampling frequency or what twice a signal time
0:03:14no what's simple of this number is is that
0:03:17i do a so you find that you can get a high
0:03:22at the same time
0:03:23for white that of that iterations
0:03:25it means that we need a a has in frequency
0:03:29because if a is increasing
0:03:30it's got at first increases as well
0:03:33so here in this design
0:03:35the focus
0:03:35more on maximizing minimizing the power
0:03:38rather than maximizing this P
0:03:40and so that's of reason real select a low sir of one eight for was design
0:03:46and the question that we want to answer is
0:03:48how little can we should be in of a pen
0:03:51like a is it why able to use like
0:03:53around ten a can be enough ten or twenty
0:03:57and so will will like is twenty six db gain or of easy but generally like a sigma don't applications
0:04:02the are in the range of like fifty to seventy db
0:04:06which is a to achieve a as we use a scaled down the supply
0:04:10so we'll see that if such a thing just yeah use a to be all them uh oh and that
0:04:14such a again
0:04:15if you try to use the in and bold is in the system can you can you can be them
0:04:19which today
0:04:20and and the whole
0:04:21uh process we power
0:04:23yeah fish
0:04:25before we four still uh so this like of a simple um evil you
0:04:30all a second order
0:04:31see my that the U C to large
0:04:34so that i made a C it can be just
0:04:36plus five in terms so
0:04:38it's a two input
0:04:40one our system
0:04:41so these are like to uh a time samples sampled input
0:04:45and then you get a a a a two a a and it can be it in terms of runs
0:04:50well the signal process
0:04:51S T F
0:04:52and noise times
0:04:55so will will to of this out
0:04:57and a be five
0:04:58so if you look into the right side here so
0:05:01so a large
0:05:02score the V four
0:05:04no that's what we is thing but is just it's of C
0:05:08you take that out of the sum
0:05:09and you give it to the to integrate is
0:05:11do of one
0:05:13let's signal that someone O G N S one year
0:05:16time being
0:05:17so this is there since that to integrate as
0:05:20i not as the
0:05:21we take the output of the someone and we some all the all
0:05:24a and you give it to a for quantizer
0:05:27for like to D C
0:05:28and the we do a C is then
0:05:31to at least a to compare compared
0:05:33oh sorry
0:05:35and by to
0:05:36which is then some time
0:05:39they have to the right is of the still a for all us
0:05:43as you can see from the
0:05:45from the signals signs here
0:05:46on the Y axis is the a is a range of and you hear
0:05:50and the X is just a times
0:05:53so you see that the was we one and B two
0:05:56have a very small signals planes
0:05:58which means that we do i mean it to use as the requirements for the
0:06:02for the integrate is that we have a
0:06:05and it also
0:06:06the using this string mean
0:06:08that's so and using
0:06:10a and ones on the entire assist
0:06:13that that's one of the pretty good advantages of having a a for long
0:06:17the second row
0:06:18that if you can the process
0:06:20from input of the quantizer
0:06:22to a the the second integrate two
0:06:25we find that
0:06:26the last
0:06:27is just a quantization noise
0:06:30so we'll see what's a man to this is to
0:06:32this feature in the next i
0:06:34but is the two things that's well
0:06:38so for this system
0:06:39we can and a like the signal as a function which is the cost
0:06:42from the input that
0:06:44to but
0:06:45and the also
0:06:46and to yeah i post
0:06:48a given by a second one scene
0:06:51if you like to do not at here for we find
0:06:54while see that all put yeah training
0:06:57position of two done
0:06:58X of the month like of the signal process
0:07:01so what implies lies is as the input
0:07:04goes to all without
0:07:05a frame
0:07:07and the one position on
0:07:08is not filtered by a signal
0:07:12so you can think of this as like it basically a portion that in but quantisation noise
0:07:17to go that
0:07:18and in that was is it is a resolution of eighties
0:07:22the next question that the we have to so is is this
0:07:27sufficient for for a requirement of realising a one bit
0:07:30snr or S india
0:07:33so think is like this system but then always are of eight can realise and nine but that's the not
0:07:37so we need an additional two bits
0:07:39so what should we do
0:07:41so what we do here is we
0:07:43we into it to a large you just on a stochastic signal at at at C
0:07:48so it's very simple like take the system from the previous slide
0:07:51we are another system a
0:07:55and then to do that
0:07:56what we do is at the input of this second stage
0:07:59we just stick you would you all
0:08:01Z you minus two from the previous it
0:08:03and added to the input of the second stage
0:08:07feed you it in a more or any uh a a little bit more mathematics a
0:08:12and if you yeah to digital filters
0:08:14S of two D which actually matches
0:08:17uh a that a lot signal as a function of the second stage and then T of one D which
0:08:22match is the noise
0:08:23a function of the first stage
0:08:24and we some though two
0:08:26why of C
0:08:27then after to doing some at which we go now
0:08:30we find that the we can get a higher order of my thing which implies a high resolution
0:08:36uh let's just finish the system here
0:08:38and so once we get this out
0:08:40this is just a decimation estimation for to and it that's mixed down by eight
0:08:44as you can see a three state to to the first as the cascade to get the call to just
0:08:48but you for that about how when only face but to
0:08:51the the estimates it's by two for by a single uh a five compensation for
0:08:55but isn't it split into three state is is just to my as the power
0:09:00let's look into the summation here and you see here is that the input signal
0:09:05X of Z
0:09:06C just a a segment as a function
0:09:08as we saw the previous slide
0:09:10the signal process
0:09:11and is you need so the signal that signal X of C
0:09:14goes out with that
0:09:15you need
0:09:17a second term here
0:09:19which is a forced one quantization noise as you in this is a ideal system
0:09:23would find that
0:09:24if you assume that the and T F one of me is equal to T of one of the
0:09:28and S you of two of these equal to have two of this at this and sits out
0:09:32and so there's is not one position on the first to from position noise doesn't come
0:09:38the total component reduced you to of C
0:09:41sees a four door that a function here
0:09:43and it effectively means that now we are able to
0:09:46noise shape by a fourth order
0:09:49that's it's it's a good thing in the sense that we have
0:09:53we can just take to stable second out of trance that systems
0:09:56put them together and get a higher order
0:09:59a a higher or the noise shaping which implies a a high snr
0:10:02what what
0:10:03the disadvantage of the system is this one will be in a cyst uh it is the design
0:10:08it will introduce mismatch between the and a lot
0:10:11a fancy
0:10:12chen and that is to
0:10:15so how does that come to the house
0:10:17so as we see you know like for a very high gain or and we see that
0:10:21the noise or or for uh
0:10:24a noise of the system is done to but i'm long
0:10:28so here here it's six
0:10:29this is the signal back from zero
0:10:32to ten made
0:10:34and on the bias
0:10:34the or like to rent C V
0:10:37so right now
0:10:38we can see that for a very large gain it's done it but um and noise
0:10:42but at that it use uses a pretty six db
0:10:44the one position noise
0:10:46starts don
0:10:49on the right to easy you why that's a reason
0:10:51so we see that
0:10:52that approximation emission that we came a are cross across year one minus the was where
0:10:58all of it true because that are outside
0:11:00as an iir are filter
0:11:02because of that that's that's thing
0:11:04and the gain of whatever
0:11:05"'cause" the system that's
0:11:07to right
0:11:08so what happens is that one position as of the first stage weeks without it
0:11:12and it produces
0:11:13by uh
0:11:14has an impact on a phone
0:11:19what we do is we take the help of and of the audit
0:11:23to really
0:11:24or a the performance impact that we have
0:11:27so i'll go to the
0:11:28the detail in the blocks of the plots in the next slide
0:11:32in this like we just want to get a general overview of what we want to do
0:11:35so what we do is we deactivate the input and we inject
0:11:38at to to do great
0:11:40no no no C
0:11:41which is generated on to
0:11:43from a linear feedback shift register
0:11:46at the same time we take the sequence
0:11:48to uh
0:11:49to a set a file for this
0:11:51uh i T have to the in in of one B
0:11:53the output of this a five for is compared with that
0:11:56a with that and a lot of by one of the Y to see
0:11:59and then we use that a missing write them to get you can be the distance
0:12:04so right
0:12:05so to summarise rate yeah actually using
0:12:07sign to the ms one
0:12:10a listener was it's one of the most easiest to implement that are most of if
0:12:16and it's it's from you got to maybe compute the next year or efficient
0:12:20from the previous state
0:12:21use that the patient size
0:12:22you know as there
0:12:23if getting in the design and
0:12:26uh and that it right C
0:12:28we take the sign of
0:12:29yeah yeah
0:12:32the way is about the works is it each be because of the operates
0:12:36to it finds that error is minimised that a miss things
0:12:39and the coefficients is actually get a response to the
0:12:43by so as a function of the signal doesn't
0:12:45that we expect on the
0:12:47so after the long i do calibration
0:12:50you see that the to uh signal thus
0:12:53this same as that of one
0:12:54and for what they stiff and then you
0:12:59there is much what do you know how we implement the
0:13:03the filters
0:13:04so we inject the thirty two bit to a L that it and m-sequence a thing
0:13:11screen and rate to five to
0:13:13the output of this a five
0:13:15is that can be a the output
0:13:18that you see why
0:13:19the Z Y two Z
0:13:20and there is then you meant to have a isn't going
0:13:23and the coefficients of an adaptive guys
0:13:26so to to use the truncation errors
0:13:28in there five for it is use an eight to five
0:13:32one disadvantage that the
0:13:34system them can have a
0:13:35is that you in for a set of the asians
0:13:37we are actually sound the system
0:13:39i a sampling frequency of the N T one fifty two hundred meg
0:13:44a what as is that
0:13:45you do you have a very low or and
0:13:47because now we are using very
0:13:50simple simple fires and optimising and
0:13:53a channel addition
0:13:55three can
0:13:56approximate three times if you do a
0:13:58a force that
0:13:59as compared to that
0:14:01so what should we do
0:14:02how should be uh
0:14:04uh or should be in signal processing techniques here
0:14:07to make sure that you get a good efficient says
0:14:10what we do a is in that to mention the police that you just
0:14:15we are using a noise cancellation the also
0:14:18which is same as
0:14:19basically C the N T F and S T E
0:14:22uh yeah on again as and C S
0:14:25and the are basically using that a
0:14:27as we talked in the previous slide
0:14:29no but ones get a is given decimation fig
0:14:32which on
0:14:34so the proposed approach what we do it
0:14:38we use one these techniques
0:14:39do really to do a convolution of
0:14:41the noise
0:14:42the noise cancellation fit
0:14:44and india for this stiff
0:14:46and that is for that
0:14:48and get a what if easy
0:14:51then we do the same per
0:14:53we reject the a random sequence
0:14:55we get the desired signal
0:14:57and B J the all output of the four and a stage
0:15:01and we make sure that sees the C
0:15:04and then compared what the stages
0:15:06do you do in N
0:15:09now this approach is very similar to what a good subband adaptive filter
0:15:13one pretty good point two
0:15:15that we are using or sampling in this case
0:15:18so using or so
0:15:20it tells us to use the you using
0:15:22and it helps service
0:15:23to ensure
0:15:25that of
0:15:26the white one of C
0:15:28oh which we used to basically
0:15:30uh this you estimate
0:15:32is not a very aggressive anti aliasing filter
0:15:34and it just has to be very many
0:15:37that this approach
0:15:39for a or so as well
0:15:43so that is the proposed a one
0:15:47the the system that we show that us like reject
0:15:51it and see
0:15:53we are a a lot of the estimate it
0:15:55and we lost
0:15:55a if it's like filter
0:15:57take out right
0:15:58the same time we take the out of that of
0:16:01well as the
0:16:02this yeah
0:16:03so for or or in this this meeting but
0:16:06a of this
0:16:09right to that it are and take their
0:16:11compute coefficient
0:16:12so we do that for both the parts of are here
0:16:15and finally
0:16:16we can speak
0:16:19good thing about this is we found that
0:16:21but the meant in a similar
0:16:23it's you using this to two at which means that you can
0:16:26re read to not to my
0:16:28based on the whole system in innovation
0:16:34do does have a small one right
0:16:36because of the increase in the number of taps of you and C T E M Z but was that
0:16:40have can is based on
0:16:42and it can he's the optimized when we do not you
0:16:46we friend the power decreases when you decoding
0:16:51that's television we shouldn't a points the number of points in or
0:16:55are pretty much the same for all all the different techniques
0:17:00is proposing but the system that we have
0:17:02yeah we have that a lot
0:17:04we we see that
0:17:05using the low you know that
0:17:07introduces some some in what errors which is a mismatch in the
0:17:10and and and to process function
0:17:12and here as the but different flavours of the back and that we have a
0:17:15and to again an F P G
0:17:17and we got and mention that a misty with the one face but to a and what if is better
0:17:26there is a
0:17:27they should signal to noise ratio in of you of this nation
0:17:30so without any yeah
0:17:31a missing uh without any kind vision find a signal to noise ratio
0:17:36as actually don't around fifty six points
0:17:39after calibration
0:17:41we see that
0:17:42all the three seems give hear a D or form
0:17:44eleven bits
0:17:46oh six is a T V C
0:17:48C six
0:17:54so here is a matrix fixed calibration
0:17:57sampling frequency V using as
0:18:00oh of the option that i
0:18:06or supply voltage
0:18:07one point two five or to
0:18:10a spare
0:18:12i meant one had in a seamless
0:18:16as a comparison
0:18:17the it's uh
0:18:18a a here is the forced to be so we use like a in a six db
0:18:23but it's just a lot of them
0:18:25to yes
0:18:26and before that up to my
0:18:28the rest of the cost of i'd a reference for years
0:18:32actually a uh and the quantizer
0:18:35and i think this and the number can be really optimized to last
0:18:39what's for a forty five
0:18:44we uh as compared to the other the words we see that we have a
0:18:48uh a or savings for that
0:18:51for the by
0:18:52we do a
0:18:53approximate analysis of finding the power for years
0:18:58so the use that are
0:18:59i see that
0:19:01uh you
0:19:02a power as it can be at fourteen time
0:19:04uh a small errors
0:19:07i become got this the actions of you for many words
0:19:12what is
0:19:12you get on that
0:19:15a before for the demise
0:19:16as this
0:19:18so we do not that it at your had
0:19:20because of some
0:19:21for does that you have to put to match
0:19:26so is a final completion format
0:19:28um are don't was to the use of a six E uh you know
0:19:34you said that
0:19:35addition white them to go go from the gain errors
0:19:37and uh we proposed something
0:19:40which uses like month at it uh are it important is
0:19:44only polyphase
0:19:47and we find that
0:19:48using this you
0:19:49that as in using at or
0:19:51the machine
0:19:52especially for high fit
0:19:57thank you
0:19:58if you have any question
0:20:22yeah we compared to the convention
0:20:27then use
0:20:28i don't
0:20:30i a moving bad to are of people
0:20:33yeah the whole system can see
0:20:46just space
0:20:58yeah i would say like a uh i think though the presence and this project has been on like a
0:21:02design between that a lot in by
0:21:05so if you compare the uh
0:21:07i figure of matter it's with the current architecture use one just
0:21:10the an up front and that we have
0:21:12it's actually not uh it's not
0:21:14as good as
0:21:15scum it's and the ball but but it's not as good as the best ones and there
0:21:19but the more mean as is here is how can you really meant like a back and
0:21:23when you have a like a very cheap and a
0:21:35thank you