G8MNY > TECH 15.09.19 08:54z 202 Lines 9719 Bytes #999 (0) @ WW
BID : 21674_GB7CIP
Subj: Coded Orthogonal Freq Div Mux
Sent: 190915/0834Z @:GB7CIP.#32.GBR.EURO #:21674 [Caterham Surrey GBR]
To : TECH@WW
By G8MNY (Updated Nov 08)
(8 Bit ASCII graphics use code page 437 or 850, Terminal Font)
Not knowing too much about it personally, here is some technical thoughts on
transmitting this mode as I see it.
This Transmission mode is used for broadcasting e.g. DAB (1536 carriers) DRM &
DTV-B Freeview TV (1707 or 6817 carriers now, incompatible 8k after 2013?)
It uses multiple carriers all phase modulated e.g. Quadrature Phase Shift
³ ³ ³ ³ ³ ³ ³ ³ QPSK 90ø /\ /\ /\
³ ³ ³ ³ ³ ³ ³ ³ on each 0øÄÅÄ180ø | | | | | |
ÄÁÄÁÄÁÄÁÄÁÄÁÄÁÄÁÄ>F carrier 270ø \ / \ / \ /
Even carrier spacing Vectors ÄÄÁÄÄÄÄÄÄÁÄÄÄÄÄÄÁÄÄÄÄÄÄ
Each carrier sidebands nearly touch
dependent on symbol rate (baud)
How electrically efficient is this mode? We know it is the best mode for
bandwidth use, as it can use all the channel space right up to the edge
compared to the less information/bandwidth efficiency of a single carrier QPSK
mode such as Digital Sat.
/~ ~\ |~~~~~~~~~~~~~|
| | ³ COFDM ³
_.-" QPSK "-._ ³ CARRIERS ³
Mux Channel Width Mux Channel Width
e.g. 30MHz e.g. 8MHz
But there are problems, as any hams using PSK31 with just 2 carriers will tell
you. The PA linearity is far more important than for speech amplitude modes let
POWER OF 2 CARRIERS (e.g. PSK31)
With a PA say capable of 100W RMS/PEP that is 200W RF crest power, then one
would intuitively think it could handle 2 carriers of 50W, but it CANNOT...
CREST CREST V CREST CREST V CREST CREST V
POWER @ 50ê POWER @ 50ê POWER @ 50ê
200 ´ ..-.. Ã 100 200 ´ Ã 100 200 ´ Ã 100
100 ´ .-~ ~-. Ã 70 100 ´ ..-.. Ã 70 100 ´ Ã 70
50 ´/ 100W PEP \Ã 50 50 ´ .-~ ~-. Ã 50 50 ´ ..-.. Ã 50
25 ´ 100W RMS Ã 35 25 ´/ 50W PEP \Ã 35 25 ´ .-~ ~-. Ã 35
12 ´ Ã 25 12 ´ 50W RMS Ã 25 12 ´/ 25W PEP \Ã 25
6 ´ Ã 17 6 ´ Ã 17 6 ´ 25W RMS Ã 17
As you can see from the above a 100W carrier has a crest of 200W & a peak volt
of 100V in 50ê. These are say the Max PA limits in my example.
Now consider the 50W carrier, it peaks 70V & to add another 50W carrier will
need another 70V, for both you need 140V, but that is more than this PA can do.
Only 2x 25W can be fitted in, that is 50V + 50V is OK. e.g. 50W mean 100W PEP.
Note that with two 25W carriers a typical meter will read something between
50W & 100W due to the slow detector time constants. A true RMS meter (heating
effect type) will read 50W, & a proper peak reading meter 100W (scope display).
With 4 equal carriers, the maximum power is again quartered to 6.5W each, or
25W mean, but still 100W PEP (200W crest) is needed to handle the in phase
peaks of the 4 carriers.
ÄÄÄ´ Â ÃÄÄÄ ³ /³\
Â ³ Á |
³ `\ ³ |
_ ÃÄÄÄ `- Á 100V
/' ³ ³ |
' ÃÄÄÄ Á Á | _
ÃÄÄÄ ³ Â ³ ³ 25V
Á ³ Á \|/ -
42V 50V 17V 100V Peak
17W 25W 2.5W 100W PEP
The carriers can be at any phase angle, on 4 different frequencies & quadrature
phase angle modulated, but at some time the vectors will all add up in phase.
So it can be seen that ten 10V peak carriers can be used on this amp of 100V
max. e.g. 10x 1W or 10W mean. (a 10V Peak carrier in 50ê is 1W)
From this the mean power must go down by the number of carriers, & each carrier
power must go down by that number squared.
PA MAX PEP PA MAX PEP
MEAN POWER = -------------- And each CARRIER POWER = -------------- 2
No of CARRIERS (No of Carriers)
For DAB for example with 1536 carriers, means a 1kW amp could only do 651mW
mean power, but still peaking 1kW PEP, & each carrier would only have a power
420uW, YES MICRO WATTS for no clipping!
For DTV with 6817 carriers spread over 8MHz, a 33kW PEP amp could only do 4.85W
mean power, & each carrier would be 710uW, but still 33kW PEP when all the
carriers occasionally are in phase! This may only be for the odd RF half cycle,
but distortion will occur if these levels are exceeded!
So a very large Tx aerial gain & large PA are need to get this Tx mode to give
a respectable ERP.
This actually is a similar case to the old telecomms "SSB coax systems" where
1000s of phone channels were put down one amplified coax. Again large amps were
needed & AGC systems to protect them from peak overdrive, as more phone calls
were made, a reduced signal to noise occurred on each channel.
With PA mixing due to non linearity the intermodulation product just GROW the
spectrum like this with an amp run to the 1dB (10%) compression point..
³ COFDM ³
-20dB? ³ CARRIERS ³ -20dB?
-45dB? ³ 2nd order ³ in channel ³ 2nd order ³ -45dB?
____________³ product ³ distortion ³ product ³___________
>-70dB? 2 Channels Adjacent Channel >-70dB?
away QRM channel QRM e.g.8MHz
It is very difficult to run a high power RF amplifier efficiently & very
linearly at the same time!
I have been told the broadcast PAs are not run at a linear drive level at all,
but driven until the spectrum growth/data error rate due to peaks hard
clipping is troublesome! E.g. as the drive level is increased, 1st the out of
channel mix products increase (2nd 3rd 4th etc.) then eventually there is
enough inband mix products to cause the Rx bit error rate to go up.
Of course added RF filters either in PA stage or after a broadband PA, may
reduce the apparent growth, but will not affect unwanted in channel distortion
sidebands present in the passband from causing errors.
³ | |
³ | |
³ | |
³ _./ \._
With COFDM linear phase or amplitude flatness of the filter is not too
important (this why it is fairly immune to signal path ghosts), as long as each
or most carriers can be decoded in their own narrow bandwidth, al` IcGC
However signal to noise is important, especially from its own cross modulation
QRM. RF channel filtering only cleans up adjacent channel mixing products, not
in channel products.
Added losses & possible filter flash over must be a problem with peak powers
1000s of times greater than the mean if this approach is used.
AVOIDING IN PHASE
Because there is a serious PA overload problem if all or even most of the
carries add up in phase, the coder software may be designed to alter/restrict
the data that could cause this, but data options & data bandwidth loss or a
time overhead will result if the method is used.
It should be clear that amplifying all the carriers in 1 PA is not a wise way
of going about multiple carrier transmission if there is to be no clipping.
Ideally each carrier could have a linear PA (needed for QPSK) & a loss free RF
directional coupler/adder used. But the short term cost of this complexity is
excessive, & the system flexibility (no of carriers) will be hardware fixed.
Comment from Andy GM7HUD:-
In reality EER is used with non-linear amplifiers for real QRO solutions. In
addition the I & Q components can be low level amplified to clipping & then
filtered before that signal is passed to the linear PA stages. This can gain an
extra 2dB in performance. Filtering the components before the main PA will
results in a higher BER at the receiver due to in-band self QRM, but the coding
used by the modulation scheme is normally robust enough to cope with this.
OLD ANALOGUE TV PAs
For main QRO stations these use a separate PA for the FM sound & a Linear PA
for the Video. The video PA is also high level (current) modulated with line
syncs to improve efficiency. The travelling wave type PA tube (several types)
often have annular target anode rings of different HT voltages that greatly
improving overall efficiency.
Using the TV PA efficiency definition of PEP/MEAN DC, efficiency of over 100%
are common on the average picture content (not a black screen)! So you can get
33kW PEP from a 33kW UHF amp & it needs < 33kW of DC!
With this design of QRO PA, the main problem being the regrowth of unwanted
lower sideband of the TV VSB signal. This is normally dealt with by a water
cooled suckout filter in the PA tank circuit. As the sound carriers are added
in with directional couplers after the linear PA there is no mixing problems.
ANALOGUE & DIGITAL TV Co-existing
With a clean QRO analogue Tx the adjacent channel emissions are generally very
low something like > -60dBc. This has allowed the QRP digital TV signals to
exist on adjacent QRO analogue channels providing the Rx is linear. The reverse
is not possible, e.g. QRO Digital with QRP analogue adjacent!
Why Don't U send an Interesting Bul?
73 De John, G8MNY @ GB7CIP
Read previous mail | Read next mail