What is affected by the lower Master Clock Frequency of the 6300?

  • 2
  • Question
  • Updated 4 years ago
  • (Edited)
What all is affected by the Master Clock Frequency of the 6300 being 1/8 that of the 6500 and the 6700?  I am thinking there may be some of the benefits of the 6500 over the 6300 that I am overlooking.
Photo of Alan - KA4B

Alan - KA4B

  • 111 Posts
  • 11 Reply Likes

Posted 4 years ago

  • 2
Photo of Alan - KA4B

Alan - KA4B

  • 111 Posts
  • 11 Reply Likes
This is something Steve will probably need to address.
Photo of Alan - KA4B

Alan - KA4B

  • 111 Posts
  • 11 Reply Likes
This is a question I am still very interested in.  Obviously having only two slices instead of four reduces the need for clock cycles by half. And reducing the maximum viewable band from 14 MHz to 7 divides the need for clock speed in two again.  It seems like the reduction from 4 to 2 DAX channels and the reduced width from 192 to 96 KHz would be part of the same 4 to 1 reduction in clock speed.  Does limiting the wide band frequency coverage from 72 MHz down to 54 MHz further reduces the need for clock speed by 25%?  What am I missing? I always enjoy your technical answers, Steve.  Can you throw any light on this?
Photo of Gerald - K5SDR

Gerald - K5SDR, Employee

  • 798 Posts
  • 1452 Reply Likes
While Steve can answer on the software side, I can shed some light as well.  To understand this fully, one needs to understand what is called the Nyquist frequency.  That subject requires more space than is available here.  Simply put, the Nyquist frequency is 1/2 of the sampling rate.  Below Nyquist you are "oversampling" and above Nyquist you are under sampling.

In order to prevent images from signals above Nyquist from "folding" back into below Nyquist frequencies, a anti aliasing filter must be employed on the input of the ADC.  To make the math simple lets use an example sampling rate of 100 Msps.  This would mean that the Nyquist frequency is 50 MHz.  Let's say we want to be able to receive a maximum frequency of 45 MHz.  The alias or image frequency will be at 55 MHz, which would fold back to 45 MHz and be heard.  The stop band cut off of the anti aliasing filter at the image frequency determines the image rejection at the folding frequency.  

So to compare the 6500/6700 sampling rate of 245.76 Msps with the 122.88 Msps of the 6300, the Nyquist frequency is doubled.  Nyquist on the 6500/6700 is 122.88 MHz compared to 61.44 MHz on the 6300.  The image frequency on the 6500/6700 receiving at 72 MHz is 173.76 MHz - a spread of over 100 MHz.  That makes it very easy to get excellent stop band rejection in the anti aliasing filter.  On the 6300, the image frequency is 68.88 MHz, which limits the maxumum input frequency as well as limits the anti-aliasing filter stop band filter rejection.  We do have FM broadcast band notch filters on all of the FLEX-6000 series radios to add additional rejection in the 88-108 MHz range where there can be very strong local signals.

Hope that helps.

Gerald
Photo of Steve - N5AC

Steve - N5AC, VP Engineering / CTO

  • 1030 Posts
  • 999 Reply Likes
In most radios, there is a local oscillator that is frequency agile that is tuned based on your request to listen to a certain frequency.  Where this LO is set, depends on your radio.  In a FLEX-5000, the LO would be set to the exact HF frequency you want to tune to because it is a direct conversion radio.  In other radios, the objective is to down-convert a portion of the band and it will be offset from your HF frequency by the IF frequency, etc.  In that first mixer, wherever it happens to be set, mixing occurs.  The mixing is of both wanted and unwanted signals.  The output of the mixer will be the down- or up-converted IF.  If there was a "pure carrier" on the antenna spigot and frequency A and it is converted to frequency B, it will also have acquired the phase noise characteristics of the LO at the LO's frequency.  In other words, if you LO has phase noise of "-130dBc/Hz at a 10kHz offset when tuned to 10MHz." your received signal will have this additional noise on it if you were listening at 10MHz.

In an agile LO radio, your LO's phase noise characteristics almost always (if the clock topology is typical) gets better the lower in frequency you go -- generally 20dB/decade better as you go down and 20dB/decade worse when you go up.  Take the example we used above -- if we now tuned to 1MHz (divide by ten), your phase noise would now be -150dBc/Hz at 10kHz offset (10kHz away from the signal).  If you've ever heard a microwave aficionado say "your phase noise is going to be lousy if you multiply that up!" this is what he's talking about.  An LO designed for HF has to be raised, multiplied, in frequency and as this is done it's phase noise characteristics get worse and worse.

OK so back to the FLEX-6000.  The FLEX-6000 is a direct sampling radio.  There is no mixer, per se, but the ADC does act as a mixer.  In a direct sampling radio, the output of the ADC will have your signals (wanted and unwanted) and they will have acquired the additive phase noise of the LO to the ADC, otherwise known as the sampling clock.  But the phase noise that is imparted on all signals is the phase noise at the sampling frequency...  

Comparing this to a traditional superhet receiver with a synthesized LO, the synthesized radio has an advantage in that the radio is able to divide down the phase noise to get better phase noise at lower frequencies ... Now there are two typical ways to build a LO in a synthesized radio.  One is with a synthesizer and the other is a DDS.  The synthesizer has a PLL which will impart a lot of phase noise and is generally not the best solution for an HF radio.  If you have a good performing HF radio, you generally have a DDS.

Back again to the FLEX-6000.  Because the phase noise of the oscillator is so important, the base phase noise of the oscillator in a direct sampling radio is critical.  But it is much easier to get the phase noise of a single frequency oscillator better than an adjustable frequency oscillator.  In the FLEX-6500/6700, we use an oscillator with very low phase noise and then we divide it by 4 which gives us another 12dB improvement in phase noise.  This oscillator is very expensive and was carefully selected to be best-in-class.  If you want to build an inexpensive direct sampling radio and want to cut costs, this is a place you can do it, but you have to be careful about how you do it.

In the FLEX-6300, we wanted a cost leader that also performed very well.  Oscillator technology has continued to advance and we were able to find a lower cost oscillator that did not have to be divided down in the FLEX-6300 and by choosing the lower sampling frequency we also get a phase noise advantage.  We do lose the processing advantage you get from oversampling, however.  In our lab tests and in experiences with customers that own both radios, our belief is that you generally will never see this difference in normal operation.  The FLEX-6700 is a "spared no expense I want the best" and the FLEX-6300 is a "I want a top performing direct sampling radio."  

There some other fallouts.  Namely we have reduced panadapter bandwidth available in the FLEX-6300 so you see 7MHz at a time rather than 14MHz at a time.  Also because Nyquist is at 61.44MHz in the FLEX-6300, we dropped the 4m band which is at 70MHz and we had to limit the receiver to 54MHz to get the Nyquist filter to "fit."  In case you're not familiar with the Nyquist filter, here's the deal: at the Nyquist frequency (half the sampling frequency), signals below the Nqyuist fold up and signals above fold down.  So in the FLEX-6300, Nyquist is at 61.44MHz so a signal at 60MHz will also appear at 61.44 - 60 + 61.44 = 62.88MHz.  And a signal at 63MHz would fold to 61.44 - 63 + 61.44 = 59.88MHz.  For a ham radio, our chief concern is that we have sufficiently filtered the mirror frequency to 54MHz and lower.  54MHz at the top of the receive band folds to 68.88MHz so our filter specs are that we don't want really any insertion loss at 54MHz and we want a LOT at 68.88MHz.  This is a sharp filter so getting this filter right was one of our challenges and the reason that we do not receive above 54MHz in the FLEX-6300.  So yes, the drop from 72 to 54MHz allowed us to use a lower sampling frequency.

I know you mentioned the clock-cycle vs. capabilities discussion.  For the record, some things follow this rule and some do not.  I think most people would be surprised by many of the things we do in the FPGA to optimize how the part works.  It really is a bit of a strange world -- you have all of these extremely powerful resources at your fingers, but each optimization to squeeze and get more out of the part requires effort and can increase complexity.  So there are tradeoffs in the investment you make, how maintainable the system is and what you've built it to do.  There are all kinds of tricks to share resources for slow-moving pieces, use different clock rates at different points in the part, etc. so we look at all these things when we decide how to implement.

The net-net is that we do a lot of "what if" scenarios trying different oscillator designs and sampling topologies to make sure we build the best receiver we can in the price class we are targeting.  The FLEX-6300 is an awesome radio (if we do say so ourselves) and I know it will stand the test of time and performance testing.  If you are contemplating which radio to buy in the family, the FLEX-6000 Family Datasheet  shows the differences in each radio.
(Edited)
Photo of Alan - KA4B

Alan - KA4B

  • 111 Posts
  • 11 Reply Likes
Thank you, Gerald and Steve.  I am wondering if these differences will effect the quality of the NR or NB when they are improved.  That is, will the performance of these features be as good on the 6300 as on the more expensive units?
(Edited)
Photo of Gerald - K5SDR

Gerald - K5SDR, Employee

  • 798 Posts
  • 1452 Reply Likes
ADC sampling rate has no effect on NR and NB.
Photo of AE0MW

AE0MW

  • 103 Posts
  • 27 Reply Likes
This was a great thread, thanks for the in depth responses guys!

I have one question though, after reading Steve's response a few times to understand it better, what is the limitation that prevents 2m from working on the 6500 series? I understand it's quite limited even on the 6700, I'm just curious as it appears from the datasheet the only difference between 6500/6700 is the additional SCU, no changes in clock frequency or FPGAs.
Photo of Gerald - K5SDR

Gerald - K5SDR, Employee

  • 798 Posts
  • 1452 Reply Likes
To reduce the cost of the FLEX-6500, we do not have the additional filtering and added low noise preamp that is needed for 2m.  The receiver performance is not limited at all on 2m on the 6700.  The exciter is limited to +10 dBm as a transverter IF but can drive a DEMI amplifier as a number of customers are doing.
Photo of AE0MW

AE0MW

  • 103 Posts
  • 27 Reply Likes
Thanks for the explanation Gerald!


Photo of Alan - KA4B

Alan - KA4B

  • 111 Posts
  • 11 Reply Likes
Gerald and Steve, I have not experienced any other amateur radio company whose leaders interact with their customers and prospects at the level that you and some of the others at Flex do.  Your willingness to answer questions in such detail, address concerns and discuss new features so freely is truly a breath of fresh air. This forum was a brilliant idea. Thank you again!
(Edited)
Photo of Gerald - K5SDR

Gerald - K5SDR, Employee

  • 798 Posts
  • 1452 Reply Likes
Thanks for the feedback.  I just wish there were more hours in the day so that I could stay in closer touch with our customers than I do.  ;>)