Frequency offset compensation for 802.15.4 (ZigBee-MSK)

Eric Jacobsen <eric.jacobsen@ieee.org> wrote:

>On Sat, 3 Oct 2015 19:39:33 +0000 (UTC), spope33@speedymail.org (Steve
>Pope) wrote:
>
>>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:

>>>It reduces the PAPR of QPSK significantly, 
>>
>>Well, this is true but it's a bit of an odd angle to view it from:
>>fundamentally FM is constant envelope, AM is not.  

>It's not odd at all, because QPSK provides a very power-efficient
>modulation, but the PAPR limited transmit power in some cases.   

Okay..  *to me* it's odd to arrive at MSK by looking at PSK and
modifying it so that it becomes MSK; and then attributing the
various advantages to these modifications.   It seems circuitous.
These advantages were already there in the MSK modulation method,
or any FM method for that matter.

(Maybe this is subtle distinction of interest to nobody but
myself...)

> The slight modification to get OQPSK provided a greatly reduced
> PAPR to get more Tx power in the link without sacrificing any
> of the power efficiency of the modulation.

Maybe the disconnect here is that you are viewing OQPSK as
something that evolved from QPSK.  It did not.  Designers
who were already going to use MSK anyway chose to use the OQPSK 
terminology in some standards submissions, for what I believe
were non-technical reasons.  To a first order that is all
OQPSK is.

>added complexity in the demodulator, but nothing terrible.
>
>FSK and MSK allow full saturation of the transmit PA, but aren't as
>power efficient as QPSK, so you still may not get as efficient of a
>link as OQPSK, and may not be able to close the link in some cases.
>In my experience FSK and MSK were historically used when the cost or
>complexity of a full PSK system was prohibitive.

Well, yes.  If you go long enough back, FSK and MSK were
popular for HF because (in addition to the factors you state), as 
FM methods, they were natually more resistant to multipath.
The capture ratio effect exhibited by FM manifests as multipath
mitigation in FSK systems, making equalization simplified in 
some systems and unnecessary in others.  Partial response systems
in HF were similarly motivated.

As costs went down and more complexity was possible, these
advantages became unimportant as you state.

Steve

On Sat, 3 Oct 2015 23:02:18 +0000 (UTC), spope33@speedymail.org (Steve
Pope) wrote:

>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:
>
>>On Sat, 3 Oct 2015 19:39:33 +0000 (UTC), spope33@speedymail.org (Steve
>>Pope) wrote:
>>
>>>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:
>
>>>>It reduces the PAPR of QPSK significantly, 
>>>
>>>Well, this is true but it's a bit of an odd angle to view it from:
>>>fundamentally FM is constant envelope, AM is not.  
>
>>It's not odd at all, because QPSK provides a very power-efficient
>>modulation, but the PAPR limited transmit power in some cases.   
>
>Okay..  *to me* it's odd to arrive at MSK by looking at PSK and
>modifying it so that it becomes MSK; and then attributing the
>various advantages to these modifications.   It seems circuitous.
>These advantages were already there in the MSK modulation method,
>or any FM method for that matter.
>
>(Maybe this is subtle distinction of interest to nobody but
>myself...)

I never said MSK was arrived at that way.  I think you've
misunderstood me.   I've been noting differences between the two by
comparisons.   That isn't an assertion of origin.

>> The slight modification to get OQPSK provided a greatly reduced
>> PAPR to get more Tx power in the link without sacrificing any
>> of the power efficiency of the modulation.
>
>Maybe the disconnect here is that you are viewing OQPSK as
>something that evolved from QPSK.  It did not.  Designers
>who were already going to use MSK anyway chose to use the OQPSK 
>terminology in some standards submissions, for what I believe
>were non-technical reasons.  To a first order that is all
>OQPSK is.

You got a citation for that?   How do you know it's the earliest use
of the term?

>>added complexity in the demodulator, but nothing terrible.
>>
>>FSK and MSK allow full saturation of the transmit PA, but aren't as
>>power efficient as QPSK, so you still may not get as efficient of a
>>link as OQPSK, and may not be able to close the link in some cases.
>>In my experience FSK and MSK were historically used when the cost or
>>complexity of a full PSK system was prohibitive.
>
>Well, yes.  If you go long enough back, FSK and MSK were
>popular for HF because (in addition to the factors you state), as 
>FM methods, they were natually more resistant to multipath.
>The capture ratio effect exhibited by FM manifests as multipath
>mitigation in FSK systems, making equalization simplified in 
>some systems and unnecessary in others.  Partial response systems
>in HF were similarly motivated.
>
>As costs went down and more complexity was possible, these
>advantages became unimportant as you state.
>
>Steve

Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

On Sat, 03 Oct 2015 15:20:44 +0000, Eric Jacobsen wrote:

> On Sat, 03 Oct 2015 07:54:15 -0400, Randy Yates
> <yates@digitalsignallabs.com> wrote:
> 
>>"tarikkazaz" <50642@DSPRelated> writes:
>>
>>> Hi all,
>>>
>>> I am working on implementation of zigbee device on FPGA. Till now I
>>> have made Tx which is compatible with commercial devices. Also I have
>>> Rx which is working with my own Tx, but does not work with commercial
>>> devices.
>>> Major reason why my receiver can not decode signals from commercial
>>> devices is frequency offset. So far I tried to apply zero crossing
>>> approach to decode signal, but that does not help
>>> (http://ieeexplore.ieee.org/xpl/login.jsp?
tp=&arnumber=1615158&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%
2Fabs_all.jsp%3Farnumber%3D1615158).
>>> Previously I have been doing correlation of known chips with received
>>> signal (on that way I was able to decode signals from my own Tx).
>>>
>>> Because now I am planning to start with implementation of frequency
>>> offset compensation for my receiver, I would like to get advice. 
>>> Which estimator do you suggest me to use? Or maybe I could use some
>>> simplified method, as zigbee has known preamble of 256 chips, which is
>>> 128 chips in I branch,
>>> and 128 chips in Q branch. Sampling rate of my ADC is 64Msps, and I
>>> was doing down sapling to 4Msps.
>>
>>Hi,
>>
>>MSK is FM, right?
> 
> No, MSK is not FM.   It is essentially OQPSK with crappy filtering.

To-may-to to-mah-to.  It's both FSK with a deviation ratio of 1/2, and 
OQPSK with half-wave cosine pulse shaping ("crappy filtering" if you 
will).

When I was designing an MSK receiver in the late 80's one of the 
references I was working from suggested that in low noise environments 
one could use a discriminator.  I didn't try it.

-- 
www.wescottdesign.com

Eric Jacobsen <eric.jacobsen@ieee.org> wrote:

>On Sat, 3 Oct 2015 23:02:18 +0000 (UTC), spope33@speedymail.org (Steve

>>>It's not odd at all, because QPSK provides a very power-efficient
>>>modulation, but the PAPR limited transmit power in some cases.   

>>Okay..  *to me* it's odd to arrive at MSK by looking at PSK and
>>modifying it so that it becomes MSK; and then attributing the
>>various advantages to these modifications.   It seems circuitous.
>>These advantages were already there in the MSK modulation method,
>>or any FM method for that matter.

>>(Maybe this is subtle distinction of interest to nobody but
>>myself...)

>I never said MSK was arrived at that way.  I think you've
>misunderstood me.   I've been noting differences between the two by
>comparisons.   That isn't an assertion of origin.

Understood.  You have been, however forming phrases such as
"MSK is not FM" and "MSK ... is essentially OQPSK with crappy 
filtering."

While these aren't explicit statements about origin, they seem
to downplay that MSK in fact originated from FM; and that OQPSK is 
in fact terminology that came along later.

>> Designers who were already going to use MSK anyway chose
>> to use the OQPSK terminology in some standards submissions,
>> for what I believe were non-technical reasons.

>You got a citation for that?   How do you know it's the earliest use
>of the term?

I don't.  It is my personal reading of the sequence of events,
but I haven't researched it, partly because it isn't a huge issue,
and partly because it can be difficult to prove that a term was
not in use before some given point in time.  (e.g. at what point did
parallel tone modem start getting called OFDM?)



Steve

On Sun, 4 Oct 2015 05:13:49 +0000 (UTC), spope33@speedymail.org (Steve
Pope) wrote:

>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:
>
>>On Sat, 3 Oct 2015 23:02:18 +0000 (UTC), spope33@speedymail.org (Steve
>
>>>>It's not odd at all, because QPSK provides a very power-efficient
>>>>modulation, but the PAPR limited transmit power in some cases.   
>
>>>Okay..  *to me* it's odd to arrive at MSK by looking at PSK and
>>>modifying it so that it becomes MSK; and then attributing the
>>>various advantages to these modifications.   It seems circuitous.
>>>These advantages were already there in the MSK modulation method,
>>>or any FM method for that matter.
>
>>>(Maybe this is subtle distinction of interest to nobody but
>>>myself...)
>
>>I never said MSK was arrived at that way.  I think you've
>>misunderstood me.   I've been noting differences between the two by
>>comparisons.   That isn't an assertion of origin.
>
>Understood.  You have been, however forming phrases such as
>"MSK is not FM" and "MSK ... is essentially OQPSK with crappy 
>filtering."

In my realm FM is analog or brute-force digital, e.g., an FM receiver 
with a Schmitt trigger on the output to get PWM.   MSK is not that.

It is, essentially, OQPSK with crappy filtering.

>While these aren't explicit statements about origin, they seem
>to downplay that MSK in fact originated from FM; and that OQPSK is 
>in fact terminology that came along later.

You're explicitly stating this as fact again without any evidence.  I
think the MSK terminology came along much later.

>>> Designers who were already going to use MSK anyway chose
>>> to use the OQPSK terminology in some standards submissions,
>>> for what I believe were non-technical reasons.
>
>>You got a citation for that?   How do you know it's the earliest use
>>of the term?
>
>I don't.  It is my personal reading of the sequence of events,
>but I haven't researched it, partly because it isn't a huge issue,
>and partly because it can be difficult to prove that a term was
>not in use before some given point in time.  (e.g. at what point did
>parallel tone modem start getting called OFDM?)

The original 1970 patent for OFDM used the terminology from the
beginning.   

You seemed to be stating it as fact and it doesn't align with my
understanding of history.   Since I've been working on this stuff
since the early 90s there has been the occasional opportunity to see
very old equipment or hear from people about the history of the
applications or links we were supplying equipment for.   PSK goes back
way earlier than I would have otherwise thought.

There's an old book, "Digital Phase Modulation", by Anderson, Aulin,
and Sundberg, that has a reasonably good treatment of history in the
first two chapters. Another book by John Anderson, "Coded Modulation
Systems", also has a pretty good treatment of some of the history and
chronology.   PSK, mostly BPSK and QPSK, was around in practical use
in the 1960s.  "Fast FSK", (aka, MSK), wasn't published until 1972
(de Buda), and much of the published research refining it was in the
middle and late 1970s.  MSK was actually patented in 1961 (Doelz and
Heald) but was apparently not published until much later.  de Buda's
paper is often referred to as the publication of the technique despite
the earlier patent, and the term MSK apparently didn't stick until
later than that.   As often happens, the patent may have significantly
slowed the adaptation of the technology and/or terminology until
later.   The publication by another author, de Buda, with another name
on the technology, "Fast FSK", more than ten years after the patent
suggests the term "MSK" was not in the common lexicon at that time.

CPM started getting attention toward the late-70s, early-80s, and
things like MSK started getitng more attention at that time, along
with the idea of "coded modulation", which came from the idea of the
phase trellis in CPM, which could be applied to "Fast-FSK" and MSK.  

From that standpoint the practical application of PSK pre-dates the
common knowledge or refinement of MSK-ish signals by a decade or so,
and PSK pre-dates CPM by nearly twice that.   Since many of the early
PSK application were for space (satellite, probes), where amplifier
linearity was an issue, OQPSK was a natural derivative that solved
some real problems, and the naming convention was pretty obvious.

There weren't so many common standards in those days.


Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

Eric Jacobsen <eric.jacobsen@ieee.org> wrote:

>On Sun, 4 Oct 2015 05:13:49 +0000 (UTC), spope33@speedymail.org (Steve


>>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:

>>Understood.  You have been, however forming phrases such as
>>"MSK is not FM" and "MSK ... is essentially OQPSK with crappy 
>>filtering."

>In my realm FM is analog or brute-force digital, e.g., an FM receiver 
>with a Schmitt trigger on the output to get PWM.   MSK is not that.

Wow, okay.

In my realm FM is any system in which a (real-valued) signal is fed into 
an FM modulator.  FSK and its sub-categories happen when you pass a 
telegraph (boxcar?) or staircase waveform into an FM modulator.

>It is, essentially, OQPSK with crappy filtering.

2-MSK is OQPSK with filtering, I think we can agree.

But I'm still hoping to tease a working definition of "crappy filtering"
out of this subthread.  My notion at the moment is that it is filtering 
that degrades capacity out of proportion to any narrowing of bandwidth.

>>While these aren't explicit statements about origin, they seem
>>to downplay that MSK in fact originated from FM; and that OQPSK is 
>>in fact terminology that came along later.

>You're explicitly stating this as fact again without any evidence.  I
>think the MSK terminology came along much later.

This could be; I don't know for certain.  I first encountered the MSK
terminology in one of Oldenwater's papers, which was from the 70's, that 
I personally ran across in the 80's.  It did not seem like he was 
introducing a new term.

(Tangentially, the "minimal" in "minimal shift keying" is kind-of
tenuous, as one can operate FSK with a modulation depth below
0.5, and even some of the aforementioned 802.15 standards have
lower modulation depths, such as 0.33.  It isn't necessarily a great idea 
in general, but people have found reason to do it.  More recently
very-low-modulation-index FSK has found application.)

>>>> Designers who were already going to use MSK anyway chose
>>>> to use the OQPSK terminology in some standards submissions,
>>>> for what I believe were non-technical reasons.

>You seemed to be stating it as fact and it doesn't align with my
>understanding of history.   Since I've been working on this stuff
>since the early 90s there has been the occasional opportunity to see

It is not surprising my sense of the history differs from yours,
since we are differently-experienced.

>There's an old book, "Digital Phase Modulation", by Anderson, Aulin,
>and Sundberg, that has a reasonably good treatment of history in the
>first two chapters. Another book by John Anderson, "Coded Modulation
>Systems", also has a pretty good treatment of some of the history and
>chronology.   PSK, mostly BPSK and QPSK, was around in practical use
>in the 1960s.  "Fast FSK", (aka, MSK), wasn't published until 1972
>(de Buda), and much of the published research refining it was in the
>middle and late 1970s.  MSK was actually patented in 1961 (Doelz and
>Heald) but was apparently not published until much later.  

Thanks for the information and the discussion.

Steve

On Mon, 5 Oct 2015 06:44:56 +0000 (UTC), spope33@speedymail.org (Steve
Pope) wrote:

>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:
>
>>On Sun, 4 Oct 2015 05:13:49 +0000 (UTC), spope33@speedymail.org (Steve
>
>
>>>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:
>
>>>Understood.  You have been, however forming phrases such as
>>>"MSK is not FM" and "MSK ... is essentially OQPSK with crappy 
>>>filtering."
>
>>In my realm FM is analog or brute-force digital, e.g., an FM receiver 
>>with a Schmitt trigger on the output to get PWM.   MSK is not that.
>
>Wow, okay.
>
>In my realm FM is any system in which a (real-valued) signal is fed into 
>an FM modulator.  FSK and its sub-categories happen when you pass a 
>telegraph (boxcar?) or staircase waveform into an FM modulator.

Like a lot of terms in this business, "FM" is overloaded.   In order
to distinguish the wide variety of analog-FM applications from actual
FSK, many people reserve "FM" for analog and "FSK" (or something
appropriate) for digital applications.   I think it helps make an
important distinction, so I use it that way, too.

>>It is, essentially, OQPSK with crappy filtering.
>
>2-MSK is OQPSK with filtering, I think we can agree.

Not really.  MSK is what you get without applying any special
filtering to FSK with h=1/2.  It turns out to be the same thing as
OQPSK with a half-sine shape of one symbol duration as the pulse
shape.   Since PSK usually allows a wide variety of pulse shapes,
almost always much longer than one pulse, filtering can be applied
over many symbols to better control the spectal occupancy of the
signal.   Filtering as narrow as 10% excess bandwidth is typical these
days in many applications.   With MSK you can't control the spectral
occupancy, and the main lobe of the frequency response is much wider
than a filtered OQPSK signal.

Hence, MSK is OQPSK with crappy filtering.   It's just my way of
describing it, nothing more.


>But I'm still hoping to tease a working definition of "crappy filtering"
>out of this subthread.  My notion at the moment is that it is filtering 
>that degrades capacity out of proportion to any narrowing of bandwidth.

If you apply any kind of filtering to MSK, it's not MSK any more.
It's spectral occupancy is not great, especially compared to PSK.
This is an impediment to its use in many applications.

Figure 12 here shows the basic idea:

http://ieee802.org/16/tg1/phy/contrib/802161pc-00_11.pdf

CQPSK is similar to MSK in that it is generated with the same type of
modulator as MSK or GMSK, i.e., a VCO with a phase filter.   For MSK
the phase filter is no filter, and GMSK and CQPSK differ only in the
shape of the phase filter.    With MSK the spectrum in FIg. 12 would
have much higher sidelobes, which is pretty much the reason for using
the phase filter.    The distinction with the filtered PSK signal is
clear, and OQPSK with the same filtering would have the spectrum.
The PSK signal shown is only 40% RRC filtering, so something like 10%
would be even narrower.

The width of the CPM spectral skirts causes ICI unless channels are
spread farther apart than they need to be with filter PSK.   Again,
"crappy" is not an inappropriate term in my estimation, but that's
just me.

>>>While these aren't explicit statements about origin, they seem
>>>to downplay that MSK in fact originated from FM; and that OQPSK is 
>>>in fact terminology that came along later.
>
>>You're explicitly stating this as fact again without any evidence.  I
>>think the MSK terminology came along much later.
>
>This could be; I don't know for certain.  I first encountered the MSK
>terminology in one of Oldenwater's papers, which was from the 70's, that 
>I personally ran across in the 80's.  It did not seem like he was 
>introducing a new term.

It was in common use then as it was a pretty active area of
publication at the time.   PSK had been around for a couple of decades
by then, though.

>(Tangentially, the "minimal" in "minimal shift keying" is kind-of
>tenuous, as one can operate FSK with a modulation depth below
>0.5, and even some of the aforementioned 802.15 standards have
>lower modulation depths, such as 0.33.  It isn't necessarily a great idea 
>in general, but people have found reason to do it.  More recently
>very-low-modulation-index FSK has found application.)

Performance drops with h<1/2.    802.15 applications don't really
care, because they WANT limited range.   Limited range is a desirable
characteristic for that application, so there's no reason to not use
something suboptimal if it solves another problem.

>>>>> Designers who were already going to use MSK anyway chose
>>>>> to use the OQPSK terminology in some standards submissions,
>>>>> for what I believe were non-technical reasons.
>
>>You seemed to be stating it as fact and it doesn't align with my
>>understanding of history.   Since I've been working on this stuff
>>since the early 90s there has been the occasional opportunity to see
>
>It is not surprising my sense of the history differs from yours,
>since we are differently-experienced.
>
>>There's an old book, "Digital Phase Modulation", by Anderson, Aulin,
>>and Sundberg, that has a reasonably good treatment of history in the
>>first two chapters. Another book by John Anderson, "Coded Modulation
>>Systems", also has a pretty good treatment of some of the history and
>>chronology.   PSK, mostly BPSK and QPSK, was around in practical use
>>in the 1960s.  "Fast FSK", (aka, MSK), wasn't published until 1972
>>(de Buda), and much of the published research refining it was in the
>>middle and late 1970s.  MSK was actually patented in 1961 (Doelz and
>>Heald) but was apparently not published until much later.  
>
>Thanks for the information and the discussion.
>
>Steve

Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

Eric Jacobsen <eric.jacobsen@ieee.org> wrote:

>On Mon, 5 Oct 2015 06:44:56 +0000 (UTC), spope33@speedymail.org (Steve

>>In my realm FM is any system in which a (real-valued) signal is fed into 
>>an FM modulator.  FSK and its sub-categories happen when you pass a 
>>telegraph (boxcar?) or staircase waveform into an FM modulator.

>Like a lot of terms in this business, "FM" is overloaded.   In order
>to distinguish the wide variety of analog-FM applications from actual
>FSK, many people reserve "FM" for analog and "FSK" (or something
>appropriate) for digital applications.   

I think that's okay as context-dependent jargon, say within the
scope of a single design project; but can be misleading in a broader 
context (such as comp.dsp).  IMO, etc.

But more importantly thinking of MSK as a type of FM signal can add 
some insight into what you're working with (examples below).

>>2-MSK is OQPSK with filtering, I think we can agree.
>
>Not really.  

It's amazing how we can look at the same set of facts yet come up
with diametrically opposite statements.

If you apply a filter with a half-sine response to an OQPSK
signal whose pulse shape is an impulse (that is, a delta function),
you get MSK.  Therefore MSK is OQPSK with filtering.

(Note, if the OQPSK signal had some pulse shape other then an
impulse, say a rectangular pulse, you'd have to back that out
but you can still specify a filter that can be applied to the
OQPSK signal to obtain an MSK signal...)

>MSK is what you get without applying any special
>filtering to FSK with h=1/2.  

Okay

> It turns out to be the same thing as OQPSK with a half-sine shape 
> of one symbol duration as the pulse shape.   

(In what way is this not ... OQPSK with filtering?)

>Since PSK usually allows a wide variety of pulse shapes,
>almost always much longer than one pulse, filtering can be applied
>over many symbols to better control the spectal occupancy of the
>signal.   Filtering as narrow as 10% excess bandwidth is typical these
>days in many applications.   

I am speculating that we're using different terminology again -- the
pulse shape can be thought of as one filter, the rest of the filtering 
can be though of as a second filter, and I think you're referring to only 
the second of these as the "filtering".  

Whereas to me, they are both filters, and convolved together they are a 
single filter that is applied to unfiltered OQPSK.

> With MSK you can't control the spectral occupancy, and the
> main lobe of the frequency response is much wider
> than a filtered OQPSK signal.

Okay.  Sorta.  There's nothing fundamentally preventing anyone
from applying filtering to an MSK signal in a similar fashion
as is commonly practiced with PSK / QAM signals.  But the results are
not going to work out as well.

The reason it is not going to work out as well is that MSK is a
type of frequency modulation, therefore it has sidelobes (defined by 
Bessel functions) and, those sidelobes carry significant information 
necessary for signal fidelity --  I really do think there's insight to 
be gained by recognizing that MSK is a subcategory of FM.

And, FM being non-linear creates an opportunity in the MSK case to add 
filtering ahead of the modulator and obtain a different (and it turns 
out, desirable) effect distinct from filtering the output of the 
modulator.

>Hence, MSK is OQPSK with crappy filtering.   It's just my way of
>describing it, nothing more.

Got it.

>>But I'm still hoping to tease a working definition of "crappy filtering"
>>out of this subthread.  My notion at the moment is that it is filtering 
>>that degrades capacity out of proportion to any narrowing of bandwidth.

>If you apply any kind of filtering to MSK, it's not MSK any more.

True.  The exact same could be said about PSK.   

>It's spectral occupancy is not great, especially compared to PSK.
>This is an impediment to its use in many applications.

I agree this impediment is there, again due to the sidelobes generic
to frequency modulation.

>Figure 12 here shows the basic idea:
>
>http://ieee802.org/16/tg1/phy/contrib/802161pc-00_11.pdf

Looks like a great paper, I will read it in full at the soonest
opportunity.

>>(Tangentially, the "minimal" in "minimal shift keying" is kind-of
>>tenuous, as one can operate FSK with a modulation depth below
>>0.5, and even some of the aforementioned 802.15 standards have
>>lower modulation depths, such as 0.33.  It isn't necessarily a great idea 
>>in general, but people have found reason to do it.  More recently
>>very-low-modulation-index FSK has found application.)

>Performance drops with h<1/2.    802.15 applications don't really
>care, because they WANT limited range.   Limited range is a desirable
>characteristic for that application, so there's no reason to not use
>something suboptimal if it solves another problem.

(It also points to a likely possibility that the designers
are intending to use a simplistic demodulator such as a frequency
discriminator or zero-crossing counter; since these sort of
gracefully degrade for lower values of h .)

Again, thanks for the discussion.

Steve

On Mon, 5 Oct 2015 20:17:57 +0000 (UTC), spope33@speedymail.org (Steve
Pope) wrote:

>Eric Jacobsen <eric.jacobsen@ieee.org> wrote:
>
>>On Mon, 5 Oct 2015 06:44:56 +0000 (UTC), spope33@speedymail.org (Steve
>
>>>In my realm FM is any system in which a (real-valued) signal is fed into 
>>>an FM modulator.  FSK and its sub-categories happen when you pass a 
>>>telegraph (boxcar?) or staircase waveform into an FM modulator.
>
>>Like a lot of terms in this business, "FM" is overloaded.   In order
>>to distinguish the wide variety of analog-FM applications from actual
>>FSK, many people reserve "FM" for analog and "FSK" (or something
>>appropriate) for digital applications.   
>
>I think that's okay as context-dependent jargon, say within the
>scope of a single design project; but can be misleading in a broader 
>context (such as comp.dsp).  IMO, etc.

It comes up all the time when helping to guide clients through
possible solution spaces.   Somebody tries to sell them "FM" for their
link, and it's analog, or analog FM with a Schmitt trigger that they
then call "digital".  Sadly, this happens a lot.  Over time I've
learned, especially in the context of helping people who aren't comm
or dsp professionals, that making the distinction is much more helpful
than not.

>But more importantly thinking of MSK as a type of FM signal can add 
>some insight into what you're working with (examples below).

In my experience thinking of MSK as CPM with no filtering is good for
evaluating how it fits within performance tradeoff spaces.    It's an
oddball in that there are multiple ways of thinking about it, but if
you only think about it as FSK you may miss opportunities to make a
much more flexible system.

>>>2-MSK is OQPSK with filtering, I think we can agree.
>>
>>Not really.  
>
>It's amazing how we can look at the same set of facts yet come up
>with diametrically opposite statements.
>
>If you apply a filter with a half-sine response to an OQPSK
>signal whose pulse shape is an impulse (that is, a delta function),
>you get MSK.  Therefore MSK is OQPSK with filtering.

MSK is OQPSK *only* in the case where the OQPSK pulse shape is a
half-sine.   OQPSK with any other pulse shape is not MSK.   So I would
disagree with your description above.

>(Note, if the OQPSK signal had some pulse shape other then an
>impulse, say a rectangular pulse, you'd have to back that out
>but you can still specify a filter that can be applied to the
>OQPSK signal to obtain an MSK signal...)

The OQPSK pulse shape must be a half-sine for MSK and OQPSK to be
equivalent.

>>MSK is what you get without applying any special
>>filtering to FSK with h=1/2.  
>
>Okay
>
>> It turns out to be the same thing as OQPSK with a half-sine shape 
>> of one symbol duration as the pulse shape.   
>
>(In what way is this not ... OQPSK with filtering?)

It is OQPSK with one specific, crappy filter, i.e., a half-sine pulse
shape.  

>>Since PSK usually allows a wide variety of pulse shapes,
>>almost always much longer than one pulse, filtering can be applied
>>over many symbols to better control the spectal occupancy of the
>>signal.   Filtering as narrow as 10% excess bandwidth is typical these
>>days in many applications.   
>
>I am speculating that we're using different terminology again -- the
>pulse shape can be thought of as one filter, the rest of the filtering 
>can be though of as a second filter, and I think you're referring to only 
>the second of these as the "filtering".  
>
>Whereas to me, they are both filters, and convolved together they are a 
>single filter that is applied to unfiltered OQPSK.

The pulse shape is the shape of the pulse that comes out of the
transmitter, however it got there.   Usually for PSK this is done with
a single pulse-shaping filter in the modulator.   There's nothing
stopping anyone from using multiple, cascaded filters to do it,
though.  In my experience this is not commonly done.

FSK, MSK, or other CPM modulators generally do not have an equivalent
of a pulse-shaping filter in the IQ signal path.   CPM often has a
phase filter, e.g., the Gaussian phase-shaping filter that makes MSK
into GMSK.   The CPM phase filter determines the transmitted pulse
shape, but it is arrived at very differently than the pulse-shaping
filter in a typical PSK or QAM or other single-carrier modulator.

>> With MSK you can't control the spectral occupancy, and the
>> main lobe of the frequency response is much wider
>> than a filtered OQPSK signal.
>
>Okay.  Sorta.  There's nothing fundamentally preventing anyone
>from applying filtering to an MSK signal in a similar fashion
>as is commonly practiced with PSK / QAM signals.  But the results are
>not going to work out as well.

If it changes the pulse shape, it is no longer MSK.   If it changes
the phase trajectory, it is no longer MSK.

>The reason it is not going to work out as well is that MSK is a
>type of frequency modulation,

Or it is CPM.   Or it is OQPSK with a half-sine pulse shape.

>...therefore it has sidelobes (defined by 
>Bessel functions) and, those sidelobes carry significant information 
>necessary for signal fidelity --  I really do think there's insight to 
>be gained by recognizing that MSK is a subcategory of FM.

You don't have to think of it that way to get the full benefit of MSK.
In fact, if you miss that it is the unfiltered case of CPM you may
miss that you can do the phase detection with a Viterbi demodulator,
which is the ML detector for CPM.

Or, you may miss that you can demodulate it like OQPSK and put a
soft-decision capacity-approaching code on it and do even better.

Limiting the conceptual view to FSK with h = 1/2 is...limiting.

>And, FM being non-linear creates an opportunity in the MSK case to add 
>filtering ahead of the modulator and obtain a different (and it turns 
>out, desirable) effect distinct from filtering the output of the 
>modulator.

If you filter the phase trajectory by filtering prior to the
modulator, it is no longer MSK.   e.g., If you apply a Gaussian filter
to the phase trajectory, it is GMSK.

>>Hence, MSK is OQPSK with crappy filtering.   It's just my way of
>>describing it, nothing more.
>
>Got it.
>
>>>But I'm still hoping to tease a working definition of "crappy filtering"
>>>out of this subthread.  My notion at the moment is that it is filtering 
>>>that degrades capacity out of proportion to any narrowing of bandwidth.
>
>>If you apply any kind of filtering to MSK, it's not MSK any more.
>
>True.  The exact same could be said about PSK.   

Definitively not true.   When specifying PSK you have to specify what
filter you use so that you can put a matching pulse filter on each end
of the link.   Any matching pair can be used if you want best
performance, but even when you use non-matching or a poorly matching
pair, it is still PSK.   You can transmit unfiltered, NRZ pulses with
PSK and receive it with integrate-and-dump filters, which are the
matching filters for NRZ.   That's stilll PSK.   So, filtered or
unfiltered or any kind of filter you want, it's still PSK.

MSK is an oddball because it is a specific limiting case, and it is
defined as such.   Changing it moves it out of the definition of what
MSK is.    If you add a phase filter it is no longer MSK, it is then
defined by the phase filter (TFM, GMSK, CQPSK, etc., etc, there are a
lot of them, and they're generally all MSK with phase filters added).
If you change the pulse shape by some means other than the phase
filter, which would be a really odd thing to do, it is also no longer
MSK.  

>>It's spectral occupancy is not great, especially compared to PSK.
>>This is an impediment to its use in many applications.
>
>I agree this impediment is there, again due to the sidelobes generic
>to frequency modulation.
>
>>Figure 12 here shows the basic idea:
>>
>>http://ieee802.org/16/tg1/phy/contrib/802161pc-00_11.pdf
>
>Looks like a great paper, I will read it in full at the soonest
>opportunity.

I really only wanted to point out the difference in spectrum in Fig.
12.

>>>(Tangentially, the "minimal" in "minimal shift keying" is kind-of
>>>tenuous, as one can operate FSK with a modulation depth below
>>>0.5, and even some of the aforementioned 802.15 standards have
>>>lower modulation depths, such as 0.33.  It isn't necessarily a great idea 
>>>in general, but people have found reason to do it.  More recently
>>>very-low-modulation-index FSK has found application.)
>
>>Performance drops with h<1/2.    802.15 applications don't really
>>care, because they WANT limited range.   Limited range is a desirable
>>characteristic for that application, so there's no reason to not use
>>something suboptimal if it solves another problem.
>
>(It also points to a likely possibility that the designers
>are intending to use a simplistic demodulator such as a frequency
>discriminator or zero-crossing counter; since these sort of
>gracefully degrade for lower values of h .)

>Again, thanks for the discussion.
>
>Steve

Eric Jacobsen
Anchor Hill Communications
http://www.anchorhill.com

Eric Jacobsen <eric.jacobsen@ieee.org> wrote:

>On Mon, 5 Oct 2015 20:17:57 +0000 (UTC), spope33@speedymail.org (Steve

>>But more importantly thinking of MSK as a type of FM signal can add 
>>some insight into what you're working with (examples below).

>In my experience thinking of MSK as CPM with no filtering is good for
>evaluating how it fits within performance tradeoff spaces.    

Thanks, an interesting perspective.

(Tangentially, 2-MSK can be either continuous phase, or non-continuous 
phase.  However, only the continuous phase sub-category of 2-MSK is 
equivalent to OQPSK with the right filtering (or the right pulse 
shape, take your pick).  So we're talking about a form of modulation
that is both MSK and CPM.)

> It's an oddball in that there are multiple ways of thinking
> about it, but if you only think about it as FSK you may miss
> opportunities to make a much more flexible system.

I'm not suggesting excluding other perspectives, just that
the perspective of MSK (or FSK) being a form of frequency modulation
is useful.

>>>>2-MSK is OQPSK with filtering, I think we can agree.

>>>Not really.  

>>It's amazing how we can look at the same set of facts yet come up
>>with diametrically opposite statements.

>>If you apply a filter with a half-sine response to an OQPSK
>>signal whose pulse shape is an impulse (that is, a delta function),
>>you get MSK.  Therefore MSK is OQPSK with filtering.

>MSK is OQPSK *only* in the case where the OQPSK pulse shape is a
>half-sine.   

Correct

OQPSK with any other pulse shape is not MSK.   

Correct

> So I would disagree with your description above.

Ah.  The disconnect here is as follows:

English often is ambiguous between whether a statement involves
the general case or a specific case.

"MSK is OQPSK with filtering" might mean, "there exists a filter such 
that applying that filter to OQPSK results in MSK".

Or it might (much less likely, in my opinion) mean "for any filter, 
applying the filter to OQPSK results in MSK".

I meant the first of these two, whereas the second is clearly false 
(trivial example, suppose the filter is H(t) = 0).

But I'll allow that my statement was technically ambiguous.

I hope this clears at least that much up.

>>> With MSK you can't control the spectral occupancy, and the
>>> main lobe of the frequency response is much wider
>>> than a filtered OQPSK signal.

>>Okay.  Sorta.  There's nothing fundamentally preventing anyone
>>from applying filtering to an MSK signal in a similar fashion
>>as is commonly practiced with PSK / QAM signals.  But the results are
>>not going to work out as well.

>If it changes the pulse shape, it is no longer MSK.   If it changes
>the phase trajectory, it is no longer MSK.

>>The reason it is not going to work out as well is that MSK is a
>>type of frequency modulation,

>Or it is CPM.   Or it is OQPSK with a half-sine pulse shape.

>>[...] MSK is a type of frequency modulation, therefore it has
>>sidelobes (defined by Bessel functions) and, those sidelobes
>>carry significant information necessary for signal fidelity --
>>I really do think there's insight to be gained by recognizing
>>that MSK is a subcategory of FM.

> You don't have to think of it that way to get the full benefit of MSK.

You don't *have* to think of it that way, but thinking of it that way
provides some immediate insight into the problem...that there
are information-containing sidelobes (perhaps more correctly, sidebands) 
as there are in any form of frequency modulation.

>Limiting the conceptual view to FSK with h = 1/2 is...limiting.

I am not suggesting this, just to be clear.  It is most effective
to have multiple conceptual views on this problem (and many
other problems).

>>And, FM being non-linear creates an opportunity in the MSK case to add 
>>filtering ahead of the modulator and obtain a different (and it turns 
>>out, desirable) effect distinct from filtering the output of the 
>>modulator.

>If you filter the phase trajectory by filtering prior to the
>modulator, it is no longer MSK.   e.g., If you apply a Gaussian filter
>to the phase trajectory, it is GMSK.

(See my comment upthread about whether GMSK is MSK.)

>>>If you apply any kind of filtering to MSK, it's not MSK any more.

>> True.  The exact same could be said about PSK.   

> Definitively not true. 

(I think we might have to disagree on this.)

> [...] So, filtered or unfiltered or any kind of filter you want, it's 
> still PSK.

> MSK is an oddball because it is a specific limiting case, and it is
> defined as such.   Changing it moves it out of the definition of what
> MSK is.    

I will counter by saying that, in the general case, filtering 
a phase-modulated signal will add amplitude modulation, thereby moving it 
outside of the definition of being purely phase modulated.

As an example, apply the filter 1 + D to QPSK.  The magnitude of the 
resulting symbols might be 0, sqrt(2), or 2.  The signal is no longer
purely phase modulated, it is also amplitude modulated.  You
end up with a QAM constellation with nine points.

In general, filtering a modulated signal will distort it such
that it might no longer meet the constraints that define
the underlying modulation.  MSK and PSK both equally fall into this 
category.  

Steve