VARA Path Test Results Revision.


Tony
 

All:

The plots below show the HF path simulator results for VARA. The ARDOP data shown in each plot was compiled using the same path parameters and added for comparison.

Plot 1 shows the path simulator results for a quiet channel. Throughput is based on file transfer speed only.

                                                                                            Plot 1. 


Plot 2 shows throughput measured over the duration of the entire Winlink session using the same quiet channel path model. The widening in throughput speed between the two modems seems to indicate greater overhead for ARDOP over the course of a Winlink session. The speed gap is less when throughput is based on file-transfer speed only as shown in Plot 1. 

                                                                                Plot 2


Plot 3 shows the on-air (40 meter) throughput data taken from 10 Winlink sessions. Modems were tested twice each session for a total of 20 test runs per modem. Tests were conducted minutes apart with moderate multi-path and strong signals. Throughput is based on file transfer speed - not the total session speed.

The average speed gap between modes was roughly 2:1 which is surprisingly close to the path simulation results shown in plot #1. Variation in throughput speed appeared somewhat greater for VARA which may be due to variations in propagation and the way VARA adapts to those variations.


                                                                                           Plot 3 


Please feel free to send suggestions and comments.

73, Tony -K2MO


Andrew OBrien
 

Congratulations for a fine study Tony. I think I recognize that 330 mile path gateway you used on 40M :) .  Dipole antenna with 50 watts.
Andy K3UK

On Wed, Apr 4, 2018 at 2:08 AM, Tony <DXDX@...> wrote:

All:

The plots below show the HF path simulator results for VARA. The ARDOP data shown in each plot was compiled using the same path parameters and added for comparison.

Plot 1 shows the path simulator results for a quiet channel. Throughput is based on file transfer speed only.

                                                                                            Plot 1. 


Plot 2 shows throughput measured over the duration of the entire Winlink session using the same quiet channel path model. The widening in throughput speed between the two modems seems to indicate greater overhead for ARDOP over the course of a Winlink session. The speed gap is less when throughput is based on file-transfer speed only as shown in Plot 1. 

                                                                                Plot 2


Plot 3 shows the on-air (40 meter) throughput data taken from 10 Winlink sessions. Modems were tested twice each session for a total of 20 test runs per modem. Tests were conducted minutes apart with moderate multi-path and strong signals. Throughput is based on file transfer speed - not the total session speed.

The average speed gap between modes was roughly 2:1 which is surprisingly close to the path simulation results shown in plot #1. Variation in throughput speed appeared somewhat greater for VARA which may be due to variations in propagation and the way VARA adapts to those variations.


                                                                                           Plot 3 


Please feel free to send suggestions and comments.

73, Tony -K2MO



Jouko OH5RM
 

Tony,
Very interesting comparison. Could you run CCIR 520-2 Moderate and Poor conditions comparison?
Jose has published another comparison but I think it is in ideal conditions and maybe using a different simulation SW
https://www.dropbox.com/s/a2ykmf0to9i52no/VARA_ARDOP.gif?dl=0

73 Jouko OH5RM


Tony
 

Andy:

> I think I recognize that 330 mile path

Thanks for being there!

Tony -K2MO

On 4/4/2018 7:04 AM, Andrew OBrien wrote:
Congratulations for a fine study Tony. I think I recognize that 330 mile path gateway you used on 40M :) .  Dipole antenna with 50 watts.
Andy K3UK

On Wed, Apr 4, 2018 at 2:08 AM, Tony <DXDX@...> wrote:

All:

The plots below show the HF path simulator results for VARA. The ARDOP data shown in each plot was compiled using the same path parameters and added for comparison.

Plot 1 shows the path simulator results for a quiet channel. Throughput is based on file transfer speed only.

                                                                                            Plot 1. 


Plot 2 shows throughput measured over the duration of the entire Winlink session using the same quiet channel path model. The widening in throughput speed between the two modems seems to indicate greater overhead for ARDOP over the course of a Winlink session. The speed gap is less when throughput is based on file-transfer speed only as shown in Plot 1. 

                                                                                Plot 2


Plot 3 shows the on-air (40 meter) throughput data taken from 10 Winlink sessions. Modems were tested twice each session for a total of 20 test runs per modem. Tests were conducted minutes apart with moderate multi-path and strong signals. Throughput is based on file transfer speed - not the total session speed.

The average speed gap between modes was roughly 2:1 which is surprisingly close to the path simulation results shown in plot #1. Variation in throughput speed appeared somewhat greater for VARA which may be due to variations in propagation and the way VARA adapts to those variations.


                                                                                           Plot 3 


Please feel free to send suggestions and comments.

73, Tony -K2MO




Graham
 

Tony,

One point ,  or  perhaps  two )

The  path simulators , due to  processing , introduce additional time  overheads ,  Jose's  testing ,  determined ,  ''PathSim add about 1.1 second delay in the communication processing'', not relevant in non-arq  systems  

There exists a  modified version of VARA , with the WAIT timer  extended to 1.5 seconds. to eliminate the  error introduced  by the  test circuit 
which Jose  can send  you .

Contrary expectations ,  narrow b/w  and  robustness  are diametrically opposed , Its noted  , where  as  ARDOP  coding changes the  on-air component to MMFSK at lower  rates  VARA  reamain's  OFDM  at  all  speeds  , as noted by  Jouko,  in poor conditions the  OFDM  coding provides, higher  data  rates and link-survival , at  50 bauds  OFDM modem is  providing  lower  s/n  than  conventions  2 tone  rtty , whilst offering  resistance to selective  path and in band qrm. 

Two  year's  back , this  advantage was not  expected and perhaps  remains  contrary to  expectations , but testing  and  on-air experience is showing this to be the  case ..even with the  increase in the  power density  , provided by the  change  from  ofdm to  mmfsk , the design refinements  in the  vara modem have  eliminated the  advantage .  providing a  constant   rms/duty cycle at all  speeds .

73-Graham
G0NBD


Tony
 

On 4/4/2018 12:23 PM, Jouko OH5RM wrote:
Tony, Very interesting comparison. Could you run CCIR 520-2 Moderate and Poor conditions comparison?
Jose has published another comparison but I think it is in ideal conditions and maybe using a different simulation SW
https://www.dropbox.com/s/a2ykmf0to9i52no/VARA_ARDOP.gif?dl=0 73 Jouko OH5RM

Jouko:

It's a time consuming process so it may take a while. In order to get accurate results, the modem needs to make multiple runs through the simulator for 10 minutes or more each run. The process is then repeated for each s/n level. The previous tests took days to complete!

Jose's simulations appear to show higher than expected throughput which normally happens when pumping too much audio from the modem into the simulator. It's matter of calibration.

Tony -K2MO 


Jouko OH5RM
 

Tony,
Yes I see testing is very time consuming. And looking closer at Jose's message I found he has made the tests I was asking for. 
I hope we here in OH could make some more on band 80m tests in coming weeks. 

Jouko OH5RM


Tony
 

On 4/4/2018 2:17 PM, Graham wrote:
narrow b/w and robustness  are diametrically opposed

I think there are limitations to this though Graham. Our transmitters
are peak-power limited so we can't simply increase the power to overcome
the relatively poor PAPR (Peak to Average Power Ratio) inherent in
wide-band, multi-carrier modems without causing distortion.

Getting the signal high enough above the noise to realize the
performance of such modems seems challenging and it appears to be the
reason why Pactor engineers chose to included narrow-band protocols with
fewer carriers and improved PAPR.

Jose takes a different approach where the PAPR appears to remain
constant. There must be a benefit to this and it must be in the
technique used to improve PAPR. Not sure if that's easier or harder to
do with ODFM modems?

Tony -K2MO


Graham
 

Hi Tony,

Using the standard metering, its  quite difficult to tell, as to the  actual power being  radiated. 

Testing with the  ic7100 , with  single tone  drive  level and the  vara  b/w representative  test tone .. 

Set to  max carrier , at the point of  ALC , 100 watts  on power meter [ steady  carrier] , ID [ supply] indicates    14 amps  @ 13.~  volts 
Changing over to the  Vara modem , test  signal , the power meter, external ave/pk  ,  set to  'average'   indicates  25 watts , but if  set to  peak detect / hold , then the  meter increases  to   50 watts  indicated  ... that  would suggest the  peak  to average is  3 dB ? but the  metering  still  shows  below the  actual , 

Taking a  close look at the  lcd-ID , that  is  peaking  very close and occasionally to  the  same  point  as the  single carrier .. From that, its reasonable to assume the  Tx is  producing the  same peak power  out  , possibly  at a slight  reduction , depending  on the  available overheads ,  the  mode  was  developed on a ft897d 

Power/B/W  is seemingly  at  odds  with  convention,  eg, Multi-psk , help files , indicate  2 tone  50 baud rtty  to require  -5.5 dB , where  as the  modem is  functioning to  -12 dB s/n with x 4 the  data  rate ,  Yes its true, the  higher the  available power , the  modem will  adapt to make best  use , but in use , its notable that  quite low powers are providing  reliable transfers, -12 dB is in the  psk31 zone ,

 

 looking at the  plot, the level  is  raised  app 15 dB across the   2.4 KHz  b/w , compared to back ground  noise . all  of this  is  harvested by the  DSP ,  one of the  reasons b/w limitation is  undesirable . 

A lot of  features of the  modem  are not  immediately  obvious ,  take  adaptation to  path  length  and or latency  in the  equipment ,  the  modem is  capable of  establishing  links  with  differing  latency  at  each end of the  path as  well as  adapting  for changes of  path  at the  start  and  during  the  transfer . there being  no  need  for a  hard  switched ,  long/short path. or equipment limitations** 

**NB- Work  continues  on SDR tx/rx  latency   issues , also  testing of the 'FM  VHF'  version  of the  modem now continues 

By maintaining  a fixed  'on air carrier rate'  all  speeds  of the  modem are  provide  with the same  level  of  robustness , where  as deploying , fast  single or  serial carriers  , the 'required' path properties are severely constricted , to  ideal  paths , the  ability to eliminate in band  qrm is  also  maintained by the  retention of the  ofdm modem .

link establishment /  handshaking is  accomplished  using  mmfsk  at  reduced  drive , this  reduces the  power  required in standby 'RX'  to a absolute minimal level   and ensure the  connection  is  maintained in deep qsb,  but as noted , the  main OFDM modem maintains  transfer to -12 B , so  no  data payload is  transferred below this  level , being arq , unrecoverable data loss , is  addressed by repeat requests ,  to date  levels  of -12 db s/n  related  to  psk31 , 


As for implementation ,  that  remains something of  mystery , to  date  it took two  over years  of  development to  reach the this  stage , Initially , it was a  multiple mode  concept , to accommodate the  high speed ofdm  mode,  but , now , all  data  transfer is via the  ofdm  modem ,  real  data  from  deployment  has been quickly  assimilated into new  versions....

On paper the  modem exceeds P4 , however , to date , no tests  have been undertaken , 

VHF  FM is  next .. 

73-Graham
G0NBD



 


Tony
 

On 4/8/2018 12:57 PM, Graham wrote:

> Using the standard metering, its  quite difficult to tell, as to the  actual power being  radiated.

Graham:

I'm sure Jose would know the answer to that? What's interesting is that VARA and other multi-carrier modems seem to achieve higher than expected output. In theory, the average power should be reduced to a few watts to prevent clipping, but that doesn't seem to be the case.

I'm told that part of the reason is because the distortion products of a multi-tone modem are nowhere near as bad as a two tone scenario so the transmitter can see moderate levels of ALC with minimal distortion. 

Whatever the case may be, VARA is one fast modem.

Thanks Graham,

Tony -K2MO


Tony
 

On 4/8/2018 12:57 PM, Graham wrote:
the  main OFDM modem maintains transfer to -12 B , so  no data payload
is  transferred below this  level , being arq , unrecoverable data
loss , is  addressed by repeat requests ,  to date  levels  of -12 db
s/n  related  to  psk31 ,
Graham:

Assuming a 3kHz AWGN channel, are you saying the modem is capable of
decoding down to -12db s/n?

Tony -K2MO


Graham
 

Yes Tony 

One of the  issues was the  peak ratio , its a little vague , but ofdm   coding  seems to  produce  rogue  spikes , 
which may or may not be reduced  and or clipped , as the  vara system is  running  at speed  and is  quite 
'tame'  can only  assume that the  peak  ratio is  actually to  design , rather than  unintended consequence 

Some one  with a  decent spectrum  analyser , with a  'maths' package could  determine , as to  what  the  levels 
are  exactly .. using  conventional  metering , its only  really  possible to  confirm 'similarity'  

Jose is  recommending a level of 25 watts ,  but thats  derived  from the   ft897 ,  that holds with  the  power meter  in  average , minimal
smoothing ,  but  when switched to   peak detect + hold , [ I used  op amp's to provide , precision diode peak detect 
and  sample & hold  , with timed hang ] so  the  meter  is  indicating actual  'peak' values , it shows  x2  at 50 watts 

But this is a  instantaneous 'dc'  reading , analysis of the  spectrum is  needed  to  determine the  power ..


Graham
 

Tony 

In the  last tests , the  modem  maintained  traffic  flow  to  -12 dB s/n
 
NB: There is  not an actual cut off  level ,  more of a combination of  factors ,  that 
serve to degrade performance to  the  point that  the system is  no longer
viable , as can be seen as the  path is  deteriorated , then the required
s/n increases , essentially  to  provide a  better  sample of  recovered 
data , for the  decoding  process  to evaluate .

The more block's  that are successfully  recovered 'single pass' ,
equates to  less  ARQ  action
which in turn , increases the data  throughput / time 

73-Graham




Rick Muething
 

Grahm, Tony, All,

It is a basic fact of engineering that OFDM modes have higher crest factors (or PAPR ...Peak to Average Power Ratio if you prefer).  This is a necessary evil when you use multiple carriers. There is a little you can do with schemes like clipping or soft compression which can reduce the PAPR some but these also generate some "damage" to the constellations so they have to be used sparingly.

THE most important thing for most digital modes is to operate the transmitter linearly.  ARDOP for example has a two tone test feature that allows the user to send a two tone signal at the same peak drive amplitude as the OFDM modes.  If the two tone signal doesn't "flat top" or create excessive ALC  (which can create distortion and intermod) the OFDM waveform will not either. Unfortunately relying just on a wattmeter reading is not always helpful...is it a peak or averaging wattmeter (or something in between)???  On ARDOP the PAPR runs from about 2.2 to about 8 depending on the modulation mode and number of carriers (a pure sine wave has a PAPR of 2.00, a two-tone test signal has a PAPR of about 2.8)

73,
Rick  KN6KB


Tony
 

Rick:

Quick question if I may:

It's my understanding that if you double the number of say PSK carriers
from 1 to 2, you double the throughput speed. In order to maintain the
same error rate as the single carrier, you need four-times the
transmitter power.

Apply that square law to say a 12 carrier modem vs. single carrier, and
you need 144 times the power to maintain the same error-free throughput.

Assuming this is correct, the average power of the 12 carrier modem
would be less than 1 watt and that begs the question: how can 64 carrier
modem like MT63 have a PAPR of something like 10:1. Can soft compression
and or other techniques reduce the PAPR that much?

Thanks,

Tony -K2MO

On 4/9/2018 1:04 PM, Rick Muething wrote:
Grahm, Tony, All,

It is a basic fact of engineering that OFDM modes have higher crest
factors (or PAPR ...Peak to Average Power Ratio if you prefer).  This
is a necessary evil when you use multiple carriers. There is a little
you can do with schemes like clipping or soft compression which can
reduce the PAPR some but these also generate some "damage" to the
constellations so they have to be used sparingly.

THE most important thing for most digital modes is to operate the
transmitter linearly.  ARDOP for example has a two tone test feature
that allows the user to send a two tone signal at the same peak drive
amplitude as the OFDM modes.  If the two tone signal doesn't "flat
top" or create excessive ALC  (which can create distortion and
intermod) the OFDM waveform will not either. Unfortunately relying
just on a wattmeter reading is not always helpful...is it a peak or
averaging wattmeter (or something in between)???  On ARDOP the PAPR
runs from about 2.2 to about 8 depending on the modulation mode and
number of carriers (a pure sine wave has a PAPR of 2.00, a two-tone
test signal has a PAPR of about 2.8)

73,
Rick  KN6KB






Graham
 

Ok  Rick, 

This may not be quite as relevant  as it  first seem's , speech is complex waveform and the  TX  ALC   is  designed to  cope  with  this ,  so it  follows , the  more  constant  ofdm signal ,  noting , the  vara  on air  carrier  rates remain the  same , should  not  present  too  much of a  problem ,  Jose is  recommending  'Long'  hang  AGC  for the  Rx

One point on  tx/rx , is that  probably  very  few equipments are  designed  for  full  break in operation ,  where the  last  known  , RX  agc  level  is  retained  during  TX , as for  the  TX  alc  ,  advancing the  drive to the  point of  alc is  probably , as good  as  it  gets , similarly 'Long' rx agc hang  .. 

I did  say, the  best  could be  hoped  for  is  'similarity'  in the  readings , looking back ,  test  versions  with   differing  papr  where  undergoing  air test  , one  year  back , over the  UK<>EA  1000  mile  path , so the  issue  was  given considerable 'air time' 

A two tone test  is  as  you  point out a  valid  test , the  reasoning  for the  2.4 Khz b/w representative tune tone , [ from the  modem] is one  ensuring  compliance  in terms  of  bandwidth , many  stations  appear to  be  using  interfaces that  limit  b/w  , which in turn will  prevent  maximum  performance, better  over  wide , than  under b/w   , again, averaging  metering  indicated's   25 % of single tone ,  but  as  previously noted , a precision  peak detector is showing  50% ....  but once  familiarized , this should not  present a  problem .

73-Graham
G0NBD


Graham
 

Tony 

Its in the  name , OFDM , the  carriers  in effect overlap , the  rest is down to the  data  pre  and  post  coding 
in the  carrier  group , exists data  carriers and  reference carriers, part of the  balance , is  ref/data numbers
image shows ,  a sort of  pass  band  ripple 

 from  memory MT63 is a comb waveform ? , with  spaced  carriers , but  very effective in its day  

73-G

Related image


Rick Muething
 

Tony,

Well it is not quite linear with respect to scaling since the carriers are all of different frequencies. They don't all add to a maximum together.  I can give you a couple of examples from ARDOP:

1. on a Single carrier of say 4PSK (2 bits/symbol) the PAPR is about 2.2

2. On that same type 4PSK modulation a 10 carrier mode has (with a small bit of soft compression) a PAPR of about 8 so you are getting 10 x the throughput but the energy/bit goes down by a factor of about 35. The one carrier mode on a SSB transmitter might have a average power out of say  45 Watts or 45 watts/carrier. The 10 carrier mode would have an average power (assuming the same radio with the same PEP rating) of about 12.5 watts or 1.25 watts/carrier. So that means for the same bit error rate you would need  about a 15.5 dB stronger signal.   On OFDM you really would like a transmitter with a high peak power capability even through the average power will be much less.

This is why you see the throughput fall off rapidly when using a large number of multi carriers in weak conditions.  With proper FEC coding you can get throughput in a pure white gaussian channel (which of course we NEVER have on HF) that approaches the Shannon limit but on HF it will always be less... often significantly less.

So the bottom line is while it is nice to be able to claim a high max speed with a particular protocol you almost never have enough power or well behaved multipath to ever achieve this.  Things like rapid adaptability (changing to slower but more robust modes (which usually also includes fewer carriers or narrower bandwidths) is important in maximizing net throughput.  There are some tricks and tradoffs like using repeated carriers (which fail decode) to help try and reconstruct (with other of the earlier same failed carriers) to achieve some improvement but this kind of averaging is not usually that productive.

The math on this is  very well documented and proven.  For example communicating over the extreme distance of the Voyager spacecraft with only its modest 10 watt transmitter and perhaps a 25-30 db transmitting dish Antenna is amazing.  BUT....the cost is enormous (200 ft dish receiving antennas cooled amplifiers to near absolute zero and bit rates of something like 10 bits/second). These communications are withing a few tenths of a dB of the Shannon limit but this is all on microwave frequencies (no multipath)....not over HF!

This is why single carrier modes (like P4) are interesting (much better PAPR and therefor more efficient transmitter utiliztion) but you need high symbol rates and large constellations for this and that means very complex channel compensation (sometimes called equalization) techniques which means LOTS of very fast and very complex math to "chase" the changing HF channel.

As is typical in engineering ....as much as we wish it weren't true there is no free lunch!

73,

Rick

On 4/9/2018 4:07 PM, Tony wrote:
Rick:

Quick question if I may:

It's my understanding that if you double the number of say PSK carriers
from 1 to 2, you double the throughput speed. In order to maintain the
same error rate as the single carrier, you need four-times the
transmitter power.

Apply that square law to say a 12 carrier modem vs. single carrier, and
you need 144 times the power to maintain the same error-free throughput.

Assuming this is correct, the average power of the 12 carrier modem
would be less than 1 watt and that begs the question: how can 64 carrier
modem like MT63 have a PAPR of something like 10:1. Can soft compression
and or other techniques reduce the PAPR that much?

Thanks,

Tony -K2MO


On 4/9/2018 1:04 PM, Rick Muething wrote:
Grahm, Tony, All,

It is a basic fact of engineering that OFDM modes have higher crest
factors (or PAPR ...Peak to Average Power Ratio if you prefer). This
is a necessary evil when you use multiple carriers. There is a little
you can do with schemes like clipping or soft compression which can
reduce the PAPR some but these also generate some "damage" to the
constellations so they have to be used sparingly.

THE most important thing for most digital modes is to operate the
transmitter linearly.  ARDOP for example has a two tone test feature
that allows the user to send a two tone signal at the same peak drive
amplitude as the OFDM modes.  If the two tone signal doesn't "flat
top" or create excessive ALC  (which can create distortion and
intermod) the OFDM waveform will not either. Unfortunately relying
just on a wattmeter reading is not always helpful...is it a peak or
averaging wattmeter (or something in between)???  On ARDOP the PAPR
runs from about 2.2 to about 8 depending on the modulation mode and
number of carriers (a pure sine wave has a PAPR of 2.00, a two-tone
test signal has a PAPR of about 2.8)

73,
Rick  KN6KB









Tony
 

Rick:

Thank you for the analogy: saving your email for future reference.
Clearly the PAPR and the resulting watts-per-carrier needs to be know
before you can compare modems. I have a few more questions if you don't
mind.

1. What would be the PAPR of say a 20 carrier 4PSK modem? Assuming it's
8, the same as the 10 carrier 4PSK example you gave, one would need 72
times the power to maintain the same throughput as the single carrier
4PSK modem. That's an 18.5db drop which seems drastic if the
signal-to-noise is marginal to begin with let alone the effect of
multi-path.
.
2. How much did soft compression improve the PAPR on the 10 carrier 4PSK
modem?

Thanks Rick,

Tony -K2MO

On 4/9/2018 8:50 PM, Rick Muething wrote:
Tony,

Well it is not quite linear with respect to scaling since the carriers
are all of different frequencies. They don't all add to a maximum
together.  I can give you a couple of examples from ARDOP:

1. on a Single carrier of say 4PSK (2 bits/symbol) the PAPR is about 2.2

2. On that same type 4PSK modulation a 10 carrier mode has (with a
small bit of soft compression) a PAPR of about 8 so you are getting 10
x the throughput but the energy/bit goes down by a factor of about 35.
The one carrier mode on a SSB transmitter might have a average power
out of say  45 Watts or 45 watts/carrier. The 10 carrier mode would
have an average power (assuming the same radio with the same PEP
rating) of about 12.5 watts or 1.25 watts/carrier. So that means for
the same bit error rate you would need  about a 15.5 dB stronger
signal.   On OFDM you really would like a transmitter with a high peak
power capability even through the average power will be much less.

This is why you see the throughput fall off rapidly when using a large
number of multi carriers in weak conditions.  With proper FEC coding
you can get throughput in a pure white gaussian channel (which of
course we NEVER have on HF) that approaches the Shannon limit but on
HF it will always be less... often significantly less.

So the bottom line is while it is nice to be able to claim a high max
speed with a particular protocol you almost never have enough power or
well behaved multipath to ever achieve this.  Things like rapid
adaptability (changing to slower but more robust modes (which usually
also includes fewer carriers or narrower bandwidths) is important in
maximizing net throughput.  There are some tricks and tradoffs like
using repeated carriers (which fail decode) to help try and
reconstruct (with other of the earlier same failed carriers) to
achieve some improvement but this kind of averaging is not usually
that productive.

The math on this is  very well documented and proven.  For example
communicating over the extreme distance of the Voyager spacecraft with
only its modest 10 watt transmitter and perhaps a 25-30 db
transmitting dish Antenna is amazing.  BUT....the cost is enormous
(200 ft dish receiving antennas cooled amplifiers to near absolute
zero and bit rates of something like 10 bits/second). These
communications are withing a few tenths of a dB of the Shannon limit
but this is all on microwave frequencies (no multipath)....not over HF!

This is why single carrier modes (like P4) are interesting (much
better PAPR and therefor more efficient transmitter utiliztion) but
you need high symbol rates and large constellations for this and that
means very complex channel compensation (sometimes called
equalization) techniques which means LOTS of very fast and very
complex math to "chase" the changing HF channel.

As is typical in engineering ....as much as we wish it weren't true
there is no free lunch!

73,

Rick




On 4/9/2018 4:07 PM, Tony wrote:
Rick:

Quick question if I may:

It's my understanding that if you double the number of say PSK carriers
from 1 to 2, you double the throughput speed. In order to maintain the
same error rate as the single carrier, you need four-times the
transmitter power.

Apply that square law to say a 12 carrier modem vs. single carrier, and
you need 144 times the power to maintain the same error-free throughput.

Assuming this is correct, the average power of the 12 carrier modem
would be less than 1 watt and that begs the question: how can 64 carrier
modem like MT63 have a PAPR of something like 10:1. Can soft compression
and or other techniques reduce the PAPR that much?

Thanks,

Tony -K2MO


On 4/9/2018 1:04 PM, Rick Muething wrote:
Grahm, Tony, All,

It is a basic fact of engineering that OFDM modes have higher crest
factors (or PAPR ...Peak to Average Power Ratio if you prefer). This
is a necessary evil when you use multiple carriers. There is a little
you can do with schemes like clipping or soft compression which can
reduce the PAPR some but these also generate some "damage" to the
constellations so they have to be used sparingly.

THE most important thing for most digital modes is to operate the
transmitter linearly.  ARDOP for example has a two tone test feature
that allows the user to send a two tone signal at the same peak drive
amplitude as the OFDM modes.  If the two tone signal doesn't "flat
top" or create excessive ALC  (which can create distortion and
intermod) the OFDM waveform will not either. Unfortunately relying
just on a wattmeter reading is not always helpful...is it a peak or
averaging wattmeter (or something in between)???  On ARDOP the PAPR
runs from about 2.2 to about 8 depending on the modulation mode and
number of carriers (a pure sine wave has a PAPR of 2.00, a two-tone
test signal has a PAPR of about 2.8)

73,
Rick  KN6KB