Anotherway toextend the FM responseofaPLL downto
DC,
involvesthe
application
of modulationtoboth the VCO and the reference oscil¬lator
simultaneously.
This scheme is called twopoint
modulation. The blockdiagram
ofa transmitter basedon asingle loop
PLLsynthesizer
modulated
by
way oftwopoint
modulation isdepicted
inFigure
3 7.The
single loop
PLLsynthesizer
consists of aVCO,
twodigital frequency dividers,
aphase-frequency
detector(PFD),
aloop filter,
and a reference oscillator. The purpose of the
loop
is to make thephase
of the VCOprecisely
tracks thephase
of the reference oscillator.Although
the PLL isnonlinear,
when theloop
islocked,J
it can beaccurately
describedby
a linear model[/ '].
5A PLL is said tobelocked, when thefrequencyof the divided VCO signal is thesame asthefrequencyof the reference oscillator and the outputof the PFD is proportionaltothephasedifference of thetwosignals.
22
Chapter
3. Transceiver ArchitecturesShaping VCXO Filter
Figure
3.7: Twopoint modulation transmitter architecture.The transfer function
describing
the instantaneous VCOangular frequency
io0 when themodulating signal Vfm
isapplied
tothe VCOinput (path Phf
inFigure 3.7)
isThf(s)
KvcoKPFDF{s)sKFMN
(3.6)
where
Kyco
isthe VCOgain
when driven from theinput
connectedto theloop filter, Kfm
is the VCOgain
when driven from the modulationinput (including amplifier Aca\), Kppp
is thephase-frequency
detectorgain factor,
N is the divider constant of the dividerdividing
the VCOfrequency,
andF(s)
is theloop
filter transferfunction,
which has alow-pass
characteristic and often is chosento have thefollowing
form:F(s) Ko
1 +STZS 1 + STp
(3.7)
Thf(s)
has ahigh-pass
characteristic: the outputfrequency
canonly
be modulated at rates greater than the PLL
loop
bandwidth. Formodulation rateslower than the PLL
loop bandwidth,
thephase
errorintroduced
by
themodulating signal
is correctedby
the action of theloop
and does not appear at the output.On the other
hand,
the transfer functiondescribing
the instanta¬neous VCO
angular frequency
io0 when themodulating signal Vfm
is3.2. Transmitter Architectures 23
1 10
Frequency (kHz)
100
Figure
3.8: Twopoint modulatorpartial
and total FMtransfer func¬
tions normalized to
Kfm-
applied
tothe reference oscillator(path Plf)
isTlf(s)
=KycoKpFDKRF(s)
R
, KvcoKPFDF{s)
b T N
(3.8)
where R is the dividerconstant of the divider
dividing
the referenceos¬cillator
frequency,
andKr
is thegain
of the reference oscillator whose outputfrequency
must also becapable
ofbeing
modulatedby
avoltage signal. Tpp(s)
has alow-pass
characteristic whichmeans thatthrough
this
path
the PLL outputfrequency
canonly
be modulated at rates slower than theloop
bandwidth. For rates faster thanthis,
thesynthe¬
sizer is not fast
enough
to follow thechanges.
To obtain a modulator with a modulation bandwidth
extending
from DCto
frequencies higher
than theloop bandwidth,
thesynthesizer
has tobe modulated
through
bothpaths simultaneously:
T(s)
=THf(s)A-Tlf(s) (3.9)
24
Chapter
3. Transceiver ArchitecturesInserting (3.6)
and(3.8)
inthe aboveexpression
it can be shown that ifKfm-R
=KR-N (3.10)
T(s) simplify
toT(s)
=Kfm (3.11)
That
is,
ifEquation (3.10)
issatisfied,
thetwopoint
modulator becomesa modulator witha flatresponse
extending
from DC toanupper limit determinedby
the VCO modulationbandwidth, independently
fromthe PLL
loop
bandwidth. Theloop
bandwidth can thus be chosenaccording
tootherspecifications
such as the maximum lock time.The
partial
FM transfer functionsTpp(s)
andThf(s) together
withthe total FM transfer function
T(s)
ofatwopoint
modulator basedon atype2third-order
loop
withabandwidth of 20 kHz andaphase margin
of 55° are illustrated inFigure
3.8. The curves have been normalized toKfm-
One
disadvantage
of this architecture is the need for avoltage
con¬trolled
crystal
oscillator(VCXO)
inplace
ofa conventionalcrystal
os¬cillator.
Moreover,
in order tosatisfy Equation (3.10),
thegain
of the VCXO or thegain
of the VCO needs to be calibrated. This can be achievedby inserting
a variablegain amplifier (or attenuator)
in oneof the two
paths Ppp
orPhf
asexemplified by
theamplifier Acai
inFigure
3.7.Fortunately,
the BERperformance
seems not to be very sensitive togain
mismatches between the twopaths [23.].
Often,
in atransceiver,
the referencefrequency
of the PLL is also used asbasic clockfrequency
for the basebandprocessor. The effects of theslight frequency change
introducedby
thetwopoint
modulator on theoperation
of the latter have thus tobe taken intoaccount. In caseof
problems,
the referencefrequency
for the mainsynthesizer
has tobegenerated
withaseparatecrystal
oscillatororwith thehelp
ofasecondPLL
using
the basebandcrystal
oscillatorasreferencefrequency.
In thelatter case, the reference
frequency
of the main PLL canbe modulatedby choosing
a verylowsecondary loop
bandwidth andby applying
thelow-frequency
components of themodulating signal
tothe VCO of thesecondary loop.
The reference oscillator isnotthe
only place
where thelow-frequency
modulation components can be
injected
into theloop.
An alternative3.2. Transmitter Architectures 25
point
ise.g. infront of theloop
filter. Asimple analysis
however shows that toachievefrequency
modulation down toDC,
an idealintegrator
is needed. Eventhough
aTDD/TDMA
systemlike Bluetoothprovides enough
time intervalsduring
which theintegrator
canbe reset, modu¬lation
through
the reference oscillator isparticularly
attractive as theVCXO
automatically performs
the neededintegration
without the need forextracircuitry.
Theuseofa
fractional-N
PLL allows foranextramethodtomodu¬late the
frequency
ofasynthesizer.
Asa fractional-N PLL allowa very fine control of thefrequency,
the modulation information candirectly
be added tothe nominal divide value
controlling
the carrierfrequency.
By including
adigital compensation filter,
this method allows todig¬
itally
modulate the carrier at rates slower and faster than theloop
bandwidth
[21].
The maindisadvantage
of this method is the fact that thecompensation
filterdepends
on the PLLloop
filter.26
Chapter
3. Transceiver ArchitecturesChapter 4
System Planning
In the
previous chapter
we have discussed ingeneral
termsreceiverandtransmitterarchitectures
capable
ofproviding
theperformance required by
Bluetooth. For eachone, we havepointed
outthe varioustrade-offs, merits,
and weaknesses.Here,
we firstanalyze
therequirements
relevant for the RF trans¬ceiver dictated
by
the Bluetooth standard. The Bluetooth type ap¬proval specify
test conditions for acomplete
system. Somespecifica¬
tions as e.g. the nominal output power are
directly applicable
as re¬quirements
for one or more parts of the transceiver. Others need to be translated in a metricappropriate
for thedesign
of the transceiver front-end. Forinstance,
the receiversensitivity
isspecified
as a maxi¬mal BER
given
a referencesignal.
The BER in its turndepends
fromthe type of
demodulator/detector
used and from the noisegenerated by
the front-end. Thus afterhaving specified
the kind of detector to beused,
the noisefigure
canbe usedas a better metric for the receiver front-end.Next,
wewillidentify
themostappropriate
architecturecompatible
with the target
applications,
andfinally
we willassign
therequired performance figures
toeach block of thesystem.27
28