Duty Cycle

With the return of sporadic E season on 6 meters my thoughts turn to transmitter duty cycle. That might seem odd unless you know that I almost exclusively operate digital modes on 6 meters. I have to monitor my use of the amplifiers more carefully than I do on CW and SSB. Most hams know that it is recommended to operate transmitters and amplifiers at lower power on digital modes, yet the understanding of why is perhaps lacking. Failure to reduce power has not gone well for many.

Despite so much written on the subject it seems worth another article on the subject. Will it help? I don't really know. Repetition has its own benefits and perhaps putting it all in one place will help a few readers. Or not. I'll try anyway.

Here are a couple of samples. On top is audio from a recorded wav file from one of my contest messages. There is no compression in the recording so the duty cycle is low; compression is added during transmission. On the bottom is a sample of CW keying. The duty cycle of the transmitted signal is slightly lower due to shaping: gradual rise and fall times to prevent key clicks.

Duty cycle for both CW and SSB can be exquisitely calculated, if you wish. That isn't necessary since we don't need that degree of precision. CW is approximately 50% when you average over dots and dashes, inter-character and inter-word spaces, and of course ordinary brief pauses. SSB can be less than 20% though few of us operate that way. Instead we equalize and compress the audio so that the duty cycle of SSB is also approximately 50%. In contrast, digital modes like FT8 and RTTY are 100%.

That is (literally) only half the story. We interleave receiving and transmitting. Assuming typical communication, conversation or contesting, we do both approximately equally. Therefore the transmit duty cycle is closer to 25% for CW and SSB and 50% for digital. FT8 transmit cycles are really only about 42% and if you CQ a lot without replies, the duty cycle for CW and SSB may be closer to 35%. Again, we can mostly ignore these nuances for the purpose of this discussion.

One problem with the basic definition of duty cycle is that it is a moving average that changes from moment to moment. For example, although a single phone QSO may have a 25% duty cycle, when averaged over, say, 1 hour of time in front of the radio, the duty cycle can be much less since you spend more time listening than in QSO.

Digital is similar. An FT8 transmission interval is 15 seconds, where you receive for 15 seconds and transmit for 15 seconds (actually only about 12.6 seconds). Over the course of a QSO the duty cycle is 50%, double that for CW and SSB. If you behave like a robot the 1 hour average may be the same, and therefore far higher than for CW and SSB outside of contests. That is one difference between digital and modes like CW and SSB where robots are, at least so far, absent. 

That 100% duty cycle is only when averaged within the bounds of a single transmission interval. Said another way, the instantaneous duty cycle during an FT8 transmission is 100%.

Up to this point I very much doubt that I've said anything that the majority of readers don't already know. From here I want to combine the idea of a moving average with transmitter and amplifier operating parameters. The connection between them is heat, more specifically heat transfer.

This is me operating on 6 meters FT8 with an Acom 1200S solid state amplifier. Notice the power level. I rarely go above 650 watts on FT8 since the temperature soon rises towards 70° C, the nominal limit. When the shack is warm the power must be kept below 500 watts. The reason is that input air is warmer and is less able to remove heat from the amplifier. On CW and SSB I regularly operate at a full kilowatt, where the lower duty cycle keeps the amp within its temperature limit.

In this stylized diagram we can see how the temperature rises during alternating transmit and receive cycles, light gray for FT* and dark for CW and SSB. Average power output is the same for both. Although the reality is a more complex, the diagram communicates the important ideas. Line width shows the variation based on ambient temperature: the hotter the air going into the amp the less effective the cooling system.

At first the temperature rise is sharp, with the slope gradually declining as the amp heats up. The slope declines (the lines really ought to be curves!) because of the increasing temperature difference between the incoming coolant fluid (ambient air) and the surface of the heat sink.

The red line is the temperature at which the amp protection trips. Operating with lower average power is required on FT8 to keep the temperature within the acceptable range. However, as you can see, it is possible to operate at higher power for a short period. Robot operators and contesters need to be more mindful of the long term average, and therefore the duty cycle.

Cooling effectiveness is a function of coolant heat capacity, ambient temperature, airflow volume, and the surface area of the material to be cooled. As is the case for any heat pump, the ability of the coolant to draw heat from the material increases with the temperature differential. Cool air cools better, and the hotter the material the more heat can be drawn off by the coolant.

At right is a picture I took of my Acom 1500 output air vent showing the integrated heat sink (cooling fins) of the 4CX1000A, the chimney directing air through the cooling fins and the exhaust air temperature sensor (top centre). 

In contrast, consider the picture that I pulled from the internet of an LDMOS device being bonded to an amplifier heat sink. We have more options on where to place the temperature sensor. It can be a thermocouple in the device, on the heat sink or a sensor placed in the airflow as is done for tube amps.

With respect to heat transfer, tubes like this have the advantage. Bonding of device elements to the heat sink is entirely integrated. The designer's job is to ensure that the air flow is properly routed and of sufficient volume to meet the cooling requirements. 

Glass envelope tubes are more challenging since there is a combination of conducted heat (bottom pins and top anode) and radiated heat from the interior metal components, especially the anode. Typically the air flow is from the bottom, to cool the pins and their glass seals, then around the glass envelope. Air flow is directed using vents and chimneys. The temperate sensor should always be at the exhaust.

Solid state devices are acutely sensitive to good thermal bonding since there is so much heat concentrated in a small volume and with limited surface area to conduct the heat away. For example, for 1000 watts of RF at 60% efficiency, the heat produced is 670 watts. That's a lot to transport over a few square centimeters of heat sink contact area! Once you get the heat over that barrier, cooling the heat sink is relatively easy.  

It is no surprise that LDMOS longevity is highly dependent on transferring that thermal load to the heat sink. Thermal protection must trigger quickly and reliably to protect the devices. The MTTF chart is from an old Freescale presentation on LDMOS. The device mentioned is likely obsolete now, however newer devices, like any semiconductor, will have similar thermal characteristics.

This brings us to the question of what we're actually measuring and where we're measuring the temperature.

All methods of thermal protection are by proxy. That is, we're measuring temperature at some remove from the locations where the heat is generated and the points of greatest criticality. Indeed, most amplifiers have more than one measurement system to detect thermal problems. For example, in a tetrode like the 4CX1000A the control grid has almost zero ability to dissipate heat. Exhaust air temperature won't detect that. It is necessary to monitor grid current and quickly shut down the amp when current indicates excess power dissipation (by P = I²R). 

Despite the sensitivity of the grids, overall tube temperature can be very high. It is common for me to measure a temperature of 90° C when continuously running ~1000 watts of FT8 on 6 meters during warm July days when the house air conditioning is off. That's enough to brew tea yet the thermal protection trigger is even higher.

LDMOS devices can quickly fail if the semiconductor junction temperatures exceed their limits. Unless there is a thermocouple built into the device we are limited to proxy measurements outside the device. 

Ideally it should be on the metal body or, if that isn't possible, on the heat sink near the LDMOS. Exhaust air measurement may be too far removed from where the heat is generated. Unlike a tube, a brief internal temperature spike may be unrecoverable. Remember that when you absolutely must work that new one.

Proxy measurements are dependent on good design and construction practices so that the thermal transfer from the interior to the heat sink is predictive of the junction temperatures. Unfortunately that is not always the case and the devices don't last for long. The screenshot is from a video by W8JI demonstrating poor thermal bonding of a FET in a late model Ameritron amplifier.

To quote the aforementioned Freescale presentation: LDMOS device thermal resistance benefits from having a backside source that is thermally and electrically bonded to the package flange, which in turn is directly mounted to the heat sink. Metal to metal contact is best. I've built power supplies and other projects where the power transistor requires a thin insulator between it and the heat sink, with thin coatings of toxic conductive paste to minimize thermal resistance. There is also the hazard of capacitance between the transistor case and the heat sink in RF applications.

It should be clear by now that thermal protection circuits on amplifiers, tube or solid state, require a physically removed sensor that measures by proxy, which demands excellent construction so that the proxy measurements are predictive of temperatures at the critical points. Otherwise we should expect regular and expensive repairs. 

In this context, duty cycle is really only one factor among many, and not necessarily the most important. Indeed, it can be expensively misleading. 

Increasing airflow (bigger and noisier fans) can only be effective if thermal transfer from device to heat sink meets the design specification. In this respect, solid state devices are more difficult to reliably cool than tubes. 

Although competent amplifier designers take these factors into account and incorporate thermal protection, failures can still occur. A little common sense on our part can pay dividends:

• Buy from companies with a reputation for good build quality, that stand behind their products and don't practice blame shifting when failures occur
• Device ratings and duty cycle matter less, often far less, than amplifier design and construction 
• Install equipment so that airflow is unconstrained even if the fans are annoying; many modern solid state amps can be operated remotely if it's a problem 
• More protection circuitry is better than less, despite the annoyance of false alarms 
• Understand that nothing is forever: failures will occur, even in expensive equipment 

I intend to follow my own advice when I buy my next amplifier, probably this summer.