Why does the Flex 6600 need band preselectors?

  • 1
  • Question
  • Updated 3 months ago
  • Answered
I’m loving my 6600 - an incredible instrument! As I dig deeper and deeper into the documentation and try to learn the art of SDR signal processing, I recall seeing that the 6600 has band preselectors. I’m guessing that these are some sort of high quality analog filters that sit ahead of the ADC. But I thought the whole point of an SDR was to move this sort of thing entirely into the DSP. For instance, Flex Radio’s VP of Engineering has a brilliant explanation of why an SDR architecture can handle saturation due to strong signals (https://helpdesk.flexradio.com/hc/en-...). If the ADC is fast enough and selective enough to use direct sampling, what is the advantage of using any preselectors at all?

Thanks to anyone who can shed light on this trade off.
Photo of Jeremy Gilbert KC1JZE

Jeremy Gilbert KC1JZE

  • 5 Posts
  • 0 Reply Likes

Posted 3 months ago

  • 1
Photo of Steve - N5AC

Steve - N5AC, VP Engineering / CTO

  • 1031 Posts
  • 1002 Reply Likes
Official Response
Analog to Digital converters (ADCs) are voltage devices.  They essentially measure the instantaneous analog voltage on the input to the ADC hundreds, thousands or millions of times per second and report those findings to the digital device that is connected to them.  Like all integrated circuits, they are designed to operate in a specific voltage range.  Above that range, their measurement accuracy either degrades or, given enough voltage, will fail catastrophically, destroying the part.  

In a radio, the designer understands that the radio may be subjected to different operating environments from an RF field standpoint.  In other words, some RF environments the receiver is in may be benign (no strong signals) whereas other environments may be sufficient to cause issues or damage in the receiver.  Before we explore the rationale for preselectors, let's talk about what damage or performance degradation looks like.

In what I will call a "legacy" receiver (a superheterodyne receiver), a series of mixers convert the frequency of interest to one specific frequency where the actual receiver in the radio operates.  If this receiver is an ADC the manufacturer might call the radio an "SDR," and certainly the radio would have software if takes data from an ADC and then demodulates it, but it does not meet the traditional radio industry definition of an SDR where the radio is "defined" by software and can be changed with a software load.  In legacy radio, the performance characteristics are only partly defined by the ADC -- the larger component of performance is often the limitations in the non-linear mixers in the superheterodyne receiver.  In these radios, protection of these devices from other signals which can pollute the receive passband with reciprocally mixed signals is paramount if you care about performance.  Large out-of-band signals degrade receiver performance in a radio like this and so band filters (preselectors) are an important, and required, component of the radio.

In the direct sampling world, however, the things that degrade performance  are different.  The key things to "avoid" are 1) poor phase noise in the sampling clock and 2) anything that would result in sending too much voltage to the ADC (otherwise known as an overload).  (1) requires careful design and component selection, but isn't really the topic at hand.  For (2) to be true, we just want to limit the total RF input to the radio and ensure it is below the overload threshold.  But, interestingly, in a true direct sampling SDR, additional signals in the receiver are actually BETTER for the receiver.  Because the risk of reciprocal mixing is low if (1) is done properly, the additional signals (especially large ones) linearize the converter.  Understanding this concept requires at least a rudimentary understanding of quantization noise.  The brief explanation is that in the process of converting from analog to digital, any patterns that emerge can cause repetitive frequency components to emerge from the spectrum resulting in undesired noise (which was not an input to the ADC, but is an artifact of conversion).  For a more thorough explanation, there are a lot of good papers on the Internet such as http://www.analog.com/media/en/training-seminars/tutorials/MT-001.pdf.  Statements above from other posters that reference limiting the available signals to a particular band of interest through a preselector are actually not correct.  These extra signals help as long as they don't appreciably overload the converter.

So what do I mean by "appreciably overload the converter?"  There is a point in an ADC where the voltage can cause some degradation of performance.  This is generally referred to as "0dBFS" or 0dB with reference to Full Scale or just Full Scale.  To avoid tripping over anything above this level, most converters are tested at -1dBFS or one full dB below full scale.  But the converter can handle voltages above this level before exhibiting damage.  For example, in the FLEX-6000 Signature Series radios, the radios end up at 0dBFS around +5dBm to +10dBm with no preamp on but only experience damage at something around +25dBm (as I recall).  For good reception, you want to keep the total input below the stated overload level of +5 to +10dBm.  Since the converter is a linear voltage device, it responds using the superposition principle.  This principle says that, in the converter world, the received voltage is the sum of the instantaneous voltage of each individual signal (as a side note, if this wasn't true we could never separate one signal from another in the DSP on the back end and just allow you to listen to a single signal).  While this is easy to explain in pictures, it's a little harder in text.  Briefly, if you are injecting a signal generator into the radio and the power level is above the stated overload value (+5 to +10dBm) then you will experience distortion of some kind in the radio.  But if you have 100 signals all across the bands and they very occasionally sum to be greater than the overload value (since they are not harmonically related in any way, come from different clock or signal sources, this is almost certainly a truism), they will only serve to linearize the converter and reduce quantization noise even if they occasionally overload the converter.  How can this be?

If you are listening to a conversation and someone drops a PA microphone and it produces a momentary loud "bang" in the room, you can probably still understand the conversation.  Similarly, if you overload for a few samples out of the 245,000,000 samples produced by the radio each second, it will not be detectable in the 3kHz bandwidth you are listening in.  It will be too brief to affect anything (and you'd never hear it, unlike my imperfect analogy of a mic drop).  This is why overload is not really a problem as long as it's only occasionally.  If you overload 0.005% of the time (this would be over 10,000 samples a second), you're probably not going to notice it while listening to the radio.

Back to the question at hand: why do we need preselectors?  Well, it's all about the total signal level in the receiver (NOT which bands of signals are in the receiver) and whether it is likely to 1) be a significant amount of overload (from a time perspective) or 2) produce damage.  This is why the preselectors are present.  At a contest station or field day, you may have antennas close together that could couple more than a few mW worth of signal into your receiver.  To prevent this from causing damage, the preselector filters are present.  They absorb the out-of-band energy from a co-located transmitter and keep that energy from the receiver.  If you are not in this kind of environment, they are not necessary.  The FLEX-6300, for example, does not have band preselectors at all.  It does have broadcast band filters, but no amateur band filters.  It would not be suitable for operation at a multi-multi contest station, but for a ham in an environment where he has no strong ham neighbors, etc. it's a great radio.  In the FLEX-6600 (the other end of the spectrum), it has filters that can take power approaching 10W from an out of band signal and filter that before it hits the receiver.  Obviously, this is not typical for amateur radio receivers, but the FLEX-6600 is designed to work at serious contest stations.  It's worth noting that a +10dBm signal which would not hurt any FLEX-6000 receiver is an S9+83dB signal.

Kind of a long-winded answer, but I hope it clarifies why and when the filters are necessary.