Monday, 30 January 2017

Power measurements of DVB-T transmitters

Power measurements of DVB-T transmitters- first draft


I have been puzzled for some time about how power of DVB-T amplifiers were measured relative to other modes. This is an important issue in DATV as DVB-T has been criticized as inefficient compared to DVB-S. It would appear that some of the debate comes down to how the power is measured. Most amateurs use simple diode power meters that do not give an accurate reading for DVB-T.

In this post, I outline the theory and practice of measuring DVB-T power correctly with either a thermal power sensor meter or envelope power using a spectrum analyser. It is important that the correct measurements be used otherwise it is comparing apples with oranges. Different power measurements are accepted in amateur radio, vis, CW (peak power, key down) compared to SSB (peak envelope power).

I wonder if another measurement unit for digital modes may be more appropriate, such as the data rate to the DC power input and spectrum width, bits/sec Watt Hz perhaps?

Theory of power measurement of DVB-T transmitters

Others have written this better than I can, so I use a direct quote:

“The output signal of a DVB-T transmitter consists of thousands of carriers modulated in phase and amplitude. Therefore it resembles a Gaussian noise signal. It should be noted, however, that very high peaks of the sum signal are limited due to effects in the process of generating and amplifying the signal. The only simple way to define the power of a COFDM signal like DVB-T is an RMS definition. It is also closely linked to the theoretical system analysis.

As the number of carriers of a given DVB-T system (either 2k or 8k) is constant and all carriers have defined power, the total power of a DVB signal is the sum of all carrier power values. In practice only the total power can be measured. In principle one symbol is insufficient for assessing the power. With thermal power meters the integration time constant is much larger than a symbol period allowing valid measurements”. Implementation guidelines for DVB terrestrial services; Transmission aspects, Digital Video Broadcasting (DVB); RTR/JTC-DVB-304

The actual number of number of carriers for 2k is 1,705 carriers and for 8k, 6,817 carriers.

Practical power measurement of DVB-T amplifiers

Power measurement is a very complex topic. It is covered in many books and application notes available on the web. It is also very important. Try reading about satellite or cell phone power and efficiency!

Simple diode power measurements used by amateurs in SWR meters are designed for measuring CW or SSB, not complex wide-band digital TV signals, including both DVB-T and DVB-S.

Again, I use a direct quote on the practical measurement of DVB-T power:

“Power measurements on DVB-T transmitters: Mean power measurements
In the case of analog transmitters, signal power is determined by measuring the peak power of the sync pulse floor of the modulated CCVS signal. The sync pulse floor is always the reference in analog TV because this signal component must be transmitted without compression or distortion.
In DVB this is different: the “Sync 1 Inversion and Randomization” block of the DVB modulator... ensures constant mean power of the transmitter output signal. In DVB, therefore, it is not the peak power that is measured, based on the crest factor, but the mean output power. Three methods are available:
1. Mean power measurement with Power Meter NRVS and thermal power sensor (FIG 27)
Thermal power sensors supply the most accurate results if there is only one TV channel in the overall spectrum, which is nearly always the case at the DVB-T transmitter. Plus, they can easily be calibrated by performing a highly accurate DC voltage measurement, provided the sensor is capable of DC measurement.
2. Mean power measurement with Spectrum Analyzer FSEx or FSP
A frequency cursor is placed on the lower and another one on the upper frequency of the DVB channel. The spectrum analyzer calculates the power for the band between the cursors (FIG 28). The method provides sufficient accuracy as in DVB-T normally no signals are put on the air in the adjacent channels.
3. Mean power measurement with DVB-T Test Receiver EFA”
“Measurements on MPEG2 and DVB-T signals (4)” in News from Rohde&Schwarz Number 172 (2001/III)

I have left out the detail of test receivers, as most amateurs would not have access to one. A power meter and the spectrum analyser are pictured below.

The simplest way to illustrate the difference in power measurement is by commercial specification of a modern DVB-T LMDOS power transistor. The CW power is 750 W, while the DVB-T power is 150 W, with an efficiency of around 50 percent. The DVB-T power measurement would have been done with a thermal power meter. The CW (key-down) and DVB-T power represents different modes and different ways of measuring power. However, a thermal power meter can measure any type of signal, including DVB-S, and give the RMS value.

Even the CW figure is misleading as Morse is a digital mode with a low duty cycle, the RMS figure would be much lower.

Operating parameters and apparent "efficiency"

The operating parameters of an amplifier will affect apparent "efficiency". The transistor above, uses a PAR (peak-to-average ratio) of only 8 dB. Amateurs have typically used a much higher PAR (30 dB?) in order to reduce the intermodulation skirts. While this will reduce power input as well as output, there has been much debate over the achieved power out compared to the maximum possible power out, usually using the CW output.

As I have discussed in earlier posts, it is possible to use filters to reduce skirts and achieve higher power outputs, comparable to device specifications. I discovered this based on intuition and experience building, modifying and tuning cavity filters for voice FM repeaters. I have subsequently found that they are routinely used in broadcast DVB-T TV.

Using a thermal power meter and operating parameters akin to broadcast TV, the achievable power efficiency can be better understood.

DATV power efficiency; W/W or something else; b/sWHz?

To a point, power efficiency, no matter how defined or measured, is probably not what we really want to know. Power in, data rate, spectrum band-width and probably cost are what we really want. So a new unit of efficiency is data rate (bits/second) divided by power in (Watts) and band-width (Hz); b/s W Hz! There is probably a unit name for it, but I don't know what it is. I am serious and will discuss it in another post with modes and numbers. 

Let's do it!

So far has been "book learning". The next step for me is to do it. I had sufficient interest in the DVB-T power puzzle to buy a used thermal power meter and a spectrum analyser capable of measuring envelope power. New power meters and sensors are very expensive (~$10K), but I imported a used HP 438A power meter and HP 8482H thermocouple thermal sensor for about $500. The spectrum analyser is a Siglent SSA3021X. Calibrated dummy loads and attenuators are also needed.

I also have a SWR power meter, a cheap AD8307 power meter and access to a Bird power meter, all of which use diode sensors. I should use a HP diode sensor to suit my meter, but I can't afford it!

I certainly don't advocate everyone buying this gear, but I want to investigate the puzzle.

Testing power output from digital signal power amplifiers is not simple. My bench tests will be the subject of another post.


to be done- see introduction

Commercial DVB-T Amplifiers and filters and implications for DATV

Commercial DVB-T Amplifiers and filters and implications for DATV (first draft)


Commercial solid-state DVB-T amplifiers use about 250 W "pallet" amplifiers; a pair of  amplifiers, each using a pair of LMDOS transistors in a single package. The circuit boards are mounted on a piece of thick copper then on a heat sink. To get higher power, 1 to 50 KW, many pallet amplifiers are used in parallel with a system of splitters and combiners. The amplifier output passes through a series of filters, to stay within a standardized spectrum mask to limit adjacent channel interference. To conclude I note some implications for DATV.

Pallet amplifers

A typical pallet amplifier is pictured below (bought on eBay). The gold rectangle on the left splits the input to a pair of Doherty amplifiers, while the one on the right combines the amplified signal to the output. The combiner uses a type of circulator to dump RF to a dummy load, the black rectangles on either side of the combiner, that doesn't go to the output, so it does not go to the other amplifier.

The white rectangles are the LMDOS transisters in this case BLF888A. Each contains a matched pair of Mosfets. Each amplifier is a modified form of class B, with a pot to adjust the quiescent current for each to improve linearity. Each amplifier has impedance matching on the input and output. A fairly conventional RF amplifier widely used in amateur radio.

LMDOS Transistor

A snip from the data sheet for a modern form of the same transistor is below. It is specifically designed for broadcast TV, with three versions for the three TV bands and runs on 50 V.

Of significant interest is the 150 W at 50 percent efficiency on DVB-T. This would appear to be an anomaly as it is rated as a 750 W device. The anomaly comes down to how the power is measured, a long-standing misunderstanding as to how DVB-T power is measured. I will discuss this in another post, but the 750 W is for single CW carrier, compared to the RMS value of the nearly 8000 carriers in the envelope of an 8 MHz DVB-T channel.

The PAR (peak-to-average ratio) value of 8 dB is the operating parameter of the amplifier, allowing only 8 dB of Crest factor before clipping. Crest factor can be as high as 38 dB. The amplifier deliberately clips the crests to 8 dB. The reason for this is that the Crest Factor is statistical. With a large number of pallet amplifiers, the crests add up potentially to give extremely high peaks, with voltages that would damage RF equipment after the amplifier. A PAR of only 8 dB will create significant intermodulation skirts that need to be filtered out.

Complete transmitter

The block diagram of a high power amplifier using multiple pallet amplifiers and drivers. in this case there are 8 pallets per block and 8 blocks, giving 64 pallets in all. The output would be about 64 x about 250 W or probably 15 kW total.

There is a filter to remove the shoulders or skirts to an acceptable level. The filters use up to 8 cavity filters of different types, I believe including some notch filters per my 70 cm amplifier filter.

There is also a harmonics filter as cavity filters are resonant at odd harmonics of the main signal. Power and cooling are major issues.

TV stations run with at least one reserve amplifier, sometime another at a different site, such as at another channels tower.

Spectrum masks

Spectrum masks are mandated for commercial TV stations to limit adjacent channel interference, as shown below.

The spectrum masks are achieved using the cavity filters discussed in the last section. There are two masks, Critical masks with a 50 dB limit for skirt for stations that have adjacent channels, as is common in metropolitan areas. Uncritical masks are less strict at 40 dB for skirts, and are used for isolated stations common in rural areas where the interference is less of a problem.

Measuring sideband emissions of T-DAB and DVB-T transmitters for monitoring purposes Rec. ITU-R SM.1792 1, RECOMMENDATION ITU-R SM.1792

Implications for DATV

There are a number of implications for DATV particularly those using DVB-T.

DVB-T amplifier efficiency is higher than commonly understood because of the different way it is measured, just like the difference between measurement of CW and SSB amplifiers.

Commercial DVB-T amplifiers are driven much harder than most DATV amplifiers, with low PAR to clip crests, but create bigger shoulders or skirts. This can be done with DATV amplifiers, provided filters are used.

Filters to remove intermodulation skirts or shoulder are mandatory. They could be used in DATV to reduce interference and to get more power. They are probably necessary for multiple lower bandwidth channels, 2 MHz, adjacent in a band.