Pages: 1 [2] 3 4 5 6 7 8   Go Down

Author Topic: Sherwood Engineering Receiver Test Data  (Read 2268 times)

KX2T

  • Member
  • Posts: 1545
Re: Sherwood Engineering Receiver Test Data
« Reply #15 on: March 19, 2021, 08:46:27 AM »

What amazes me is Rob has taken years of his OWN TIME, HIS OWN LAB countless hours measuring radio's that should be on his list in his opinions yet you have a bunch of crying wolves saying he should be testing sub $200 dollar dongle tonka toy radio's, really from what Planet are you guys coming from! There are plenty of SWL posting on the web that show comparisons to other like tonka toys, there is no need to gum up a already fine list of radio's for the Amateur's use like Rob's list, it simply shows that if you give someone something for free they turn around and want more free stuff. Be happy for what you have already and try surfing some of the SWL posting you may find these other forums may give you the results your looking for in the sub compact Dongle RX sections but that might take a little WORK for some of you!
 
Logged

N6YWU

  • Posts: 362
    • HomeURL
Re: Sherwood Engineering Receiver Test Data
« Reply #16 on: March 19, 2021, 10:10:34 AM »

yet you have a bunch of crying wolves saying he should be testing sub $200 dollar dongle tonka toy radio's

You sound like the IBM mainframe salesmen who told programmers to ignore sub $2k tonka toy computers like the Apple II.  (AAPL now has a market cap 17X greater than IBM, and app store revenue topped $64B last year).  If you want to expand amateur radio, you have to look at what (toys) potential future and new licensees are buying.  And using (for SOTA, et.al.)
Logged

KF5LJW

  • Posts: 577
    • HomeURL
Re: Sherwood Engineering Receiver Test Data
« Reply #17 on: March 19, 2021, 01:53:47 PM »

What amazes me is Rob has taken years of his OWN TIME, HIS OWN LAB countless hours measuring radio's that should be on his list in his opinions yet you have a bunch of crying wolves saying he should be testing sub $200 dollar dongle tonka toy radio's
I admire his work. I wondered why there were no stand-alone RX in the list? I learned there were a few, but most are tranceivers donated.

What I thought might be going on is he was like you. Fortunately it does not appear to be the case. It would be quite embarrassing if you could buy a receiver that performs as well or better for $200 when you paid $3000 for your tonka toy. SDR is the future like it or not.

Logged

VE3WGO

  • Member
  • Posts: 666
Re: Sherwood Engineering Receiver Test Data
« Reply #18 on: March 19, 2021, 04:00:32 PM »

$200 plus the cost of the pc and display you need to run the SDR.  without the pc, that SDR is dead a doornail.

meanwhile the "$3000 radio" includes a 100W Transmitter, TR switch, likely a touchscreen, and an antenna tuner too, and doesn't need a pc to run it (unless you really want to).

I really would like to see some of these SDrs compared to high end transceivers.  Unless they have high dynamic range analog frontends ahead of the DAC, they won't be anywhere near the top of the list.

Logged

N2DTS

  • Member
  • Posts: 1368
Re: Sherwood Engineering Receiver Test Data
« Reply #19 on: March 19, 2021, 04:52:18 PM »

That is true, even the 7300 and 7610 are not at the top.
If you want really great dynamic range you need to limit signals before the A/D converter or use a really good one.

The direct sample radios have the advantage of being able to display and use the entire spectrum, even running many receivers (slices) on different bands at once.

I wonder how the Flex 6700 does so well, does it have some sort of filter before the A/D converter or just a super A/D converter?
« Last Edit: March 19, 2021, 05:01:28 PM by N2DTS »
Logged

N6YWU

  • Posts: 362
    • HomeURL
Re: Sherwood Engineering Receiver Test Data
« Reply #20 on: March 20, 2021, 09:34:32 AM »

$200 plus the cost of the pc and display you need to run the SDR.  without the pc, that SDR is dead a doornail.

You can run most all of the under $200 SDR receivers from a $35 Raspberry Pi, no expensive PC needed (but almost everyone already has a suitable computer of some sort already).  And you only need to go up to $320 to get a brand new SDR HF transceiver (that also works with a Pi 3B or 4 or Mac or PC or Android mobile or iPhone or iPad, which, again, most people already have at least one of).

Plus this lower priced entry point for these SDRs might get a lot of younger computer savvy potential licensees more interested in amateur radio (hopefully at a high enough rate to offset the current SK/annum rate).

So even though they might not be competitive in all situations (multi-multi contest stations, etc.), but it might be useful to know how competitive these SDRs are in other situations.
Logged

KF5LJW

  • Posts: 577
    • HomeURL
Re: Sherwood Engineering Receiver Test Data
« Reply #21 on: March 20, 2021, 11:07:06 AM »

Plus this lower priced entry point for these SDRs might get a lot of younger computer savvy potential licensees more interested in amateur radio (hopefully at a high enough rate to offset the current SK/annum rate).

Glad to see someone finally understands. Without youth, the party is over and time to turn out the lights. With an SDR, you are not limited to any modes of today and tomorrow. The commercial side has already made the move to SDR.

Last thing youth wants to do is listen to two old bald men arguing over a comb on the radio, or that is how I have done it since dirt was discovered.  ;D Radio to youth is no mystery. Today they see radio as module or component of a system used as a link. Its a box on a 1-line diagram going to the Cloud. They are not the least bit interest in talking or texting on a radio (unless it is gaming on the internet) because there are much better methods like a cell phone. Today what interest them is things that require a data or telemetry link. To do things like RC airplanes, drones, helicopters, cars, trucks and my favorite robots, fighting robots. All controlled by a micro-controller or tablet. Problem is ham radio just does not fit due to very limited bandwidth and rules on encoding. What they do have to work with today are modular TX and RX using WiFi, and the 2.4 Ghz and 5 Ghz bands wide enough to broadcast both video and telemetry on a data link.

Ham radio needs to change and modernize embracing new technology like it use to. Today the hobby is way behind and showing its age.
Logged

K6BRN

  • Member
  • Posts: 2231
Re: Sherwood Engineering Receiver Test Data
« Reply #22 on: March 20, 2021, 11:23:30 AM »

Sherwood's list focusses on a few (important) performance metrics, mostly for receivers.  But these metrics by themselves do not indicate the practical capabiity, utility and quality of use.  Case in point - see user reviews for the Yaesu FTDX-3000 and of course the Icom IC-7300.  They're not at the top of Rob's list, but they are at or near the top of user's favorites.

Make no mistake, Rob's list is VERY useful in understanding how a receiver may behave at some of it's corner conditions, and his discoveries, widely published, have driven manufacturers to focus on phase noise and selectivity, among other things, leading to improved products.  But in some areas, this has gone to an extreme that's just a little silly.  Case in point, the Yaesu FTDX-101D, where isolation and tuning of the analog front end seem to be so critical that Yaesu forbids the user from opening it up, even to install optional filters. 

Regarding DSP radios - quality is quality and quality costs.  Do not ever expect a $200  "SDR" to outperform a $1000 radio (analog, digital or hybrid), because some of the most critical components that set performance are in the front end, and good ones are expensive.

In a DSP (SDR) radio, it's the bandpass/preselection filters and ADC.  In an analog radio its the bandpass/preselection filters and 1st mixer.  In both cases, the quality of the frequency reference also directly drives performance - and good ones are expensive.

$200 DSP radios are usually based on commercial chipsets developed for other applications, like digital video broadcasting (DVB).  These chipsets are designed to have ADEQUATE performance in their INTENDED APPLICATION at the LOWEST POSSIBLE COST.  They usually incorporate cheap and relatively low performance front end filters and I/Q mixers to allow cheaper, lower dynamic range ADCs to be built into the devices, with all the resulting I/Q imbalance that implies.  The ADCs are followed by a digital decimator to reduce sample bandwidth to something an attached computer (that you provide) can handle) - audio frequencies.

Actual software post-processing happens at AUDIO bandwidths inside the attached computer, and the lower the processed digital bandwidth the less performance the digital portion of the system will have.

This performance is VERY low in a $100 or $200 dongle.

Yaesu has improved performance of its radios by moving back end digital processing bandwidth from 10's of KHz in the FTDX-5000, 3000, 1200 and FT991 to multiple MHz in the FTDX-101D, making their digital performance much better.  And in the FTDX-101D, they've mated a top-notch and very expensive analog front end to a "high IF sampling" back end, to hit the top of Sherwoods list in performance.

When the cost/performance of the best front end analog mixers is equivalant to the cost/performance of the best available ADCs, then use of an analog front end (mixer) becomes redundant, for equal performance.  At HF frequencies (very low), we are almost there, as shown by the relative performance of the Flex-6600, Icom IC-7610 and Yaesu FTDX-101D.

But further up the frequency bands, that battle is still being fought.

The big attractions of an all digital processing chain is:1.  Flexibility/agility, and 2. Mechanical/elecrical simplicity.  For the latter, simplicity is gained by moving all of the incredible complexity of DSP inside a digital signal processing IC (ASIC, FPGA or HA CPU), making electical/mechanical design and production easier and cheaper.

Have you ever looked inside a direct digital sampling Flex-6600?  Surprise!  It's almost empty, has few shielded compartments and is made mostly of bent metal.  It's pretty inexpensive to make, once development cost has been retired.

In contrast, the IF sampling, hybrid analog/digital FTDX-101D is packed to the brim with RF-tight cavities, custom castings, hundreds of components and needs a lot of "touch labor" to complete.  It's an expensive radio to build, no doubt about it.  And its very critical internal tuning WILL drift over time, degrading performance - That's a future problem that owners have not realized yet.

So - should new hams play with cheap "SDR radios"?  Yep.  It's like a teaser.  It may take them off into development of things other than radios (like hobbyist NVAs and SAs) and will likely lead them to more serious radios later on.  Lot's of good MAY come from this (some already has).  But if and only if they are technically capable AND have some good training and mentoring.

Otherwise they'll hit the same dead end that uninformed amateur SDR enthusiasts have hit when assuming a DSP system with 3 KHz of processed bandwidth can become a "perfect radio".  It can't, and they still don't know why.

Hopefully, young people will explore this field and go into it professionally - because there is just as much (and at present much more) money to be made in digital signal processing (for communications, control systems and remote sensing) as there is in robotics.  In fact, advanced robotics relies on DSP to perform.

Best Regards (and sorry for the very long essay!)

Brian - K6BRN
« Last Edit: March 20, 2021, 11:29:40 AM by K6BRN »
Logged

N6YWU

  • Posts: 362
    • HomeURL
Re: Sherwood Engineering Receiver Test Data
« Reply #23 on: March 20, 2021, 12:27:32 PM »

The ADCs are followed by a digital decimator to reduce sample bandwidth to something an attached computer (that you provide) can handle) - audio frequencies.

Actual software post-processing happens at AUDIO bandwidths inside the attached computer, and the lower the processed digital bandwidth the less performance the digital portion of the system will have.

Actually, the opposite.  With an RTL-SDR, you can DSP process up to 2 MHz of bandwidth on a Raspberry Pi, even more with an SDRPlay and a faster computer.  With an HF+ you can process up to 768kHz, and with an HL2 up to 4 slices of 384 kHz of bandwidth each.  All of these allow a suitable computer to simultaneously demodulate, skim, or decode digital signals across a much wider bandwidth than with just transceiver SSB audio.  Custom gateware on an HL2 allows FT8 skimming on up to 8 (or 10?) HF bands *simultaneously*.  How many of the bigger transceivers can do that.

As for performance, it depends on which performance criteria in what situations.  The pure RF sensitivity of some of these units (RSPdx and HF+ Discovery), absent strong interferers, might be quite competitive with far more expensive boxes.  But that needs to be tested.
Logged

N6YWU

  • Posts: 362
    • HomeURL
Re: Sherwood Engineering Receiver Test Data
« Reply #24 on: March 20, 2021, 01:03:01 PM »

The ADCs are followed by a digital decimator to reduce sample bandwidth to something an attached computer (that you provide) can handle) - audio frequencies.

Actual software post-processing happens at AUDIO bandwidths inside the attached computer, and the lower the processed digital bandwidth the less performance the digital portion of the system will have.

For receiving narrow-band signals, a high-speed ADC in conjunction with proper decimation to a lower bandwidth actually improves a few performance metrics due to DSP processing gain.
Logged

W6RZ

  • Member
  • Posts: 449
Re: Sherwood Engineering Receiver Test Data
« Reply #25 on: March 20, 2021, 02:48:56 PM »

I wonder how the Flex 6700 does so well, does it have some sort of filter before the A/D converter or just a super A/D converter?

There are only two ADC's used by direct sampling radios from Icom and Flex. The LTC2208 and the AD9467. That's it. And they have very similar performance.

The reason that radios test all over the place on the Sherwood list is due to semiconductor device to device variability. Just look at radios with more than one sample on the list. The IC-R8600 is simultaneously at the top of the list at position 4 and at position 20. Same goes for Flex products. The 6700 is at both position 2 and 22 (below the IC-7300).
Logged

W6RZ

  • Member
  • Posts: 449
Re: Sherwood Engineering Receiver Test Data
« Reply #26 on: March 20, 2021, 02:53:06 PM »

I'm tempted to Amazon Rob Sherwood an Airspy HF+, just to shut up the naysayers. Based on testing from others, it would probably place next to the FTdx-3000.
Logged

K6BRN

  • Member
  • Posts: 2231
Re: Sherwood Engineering Receiver Test Data
« Reply #27 on: March 20, 2021, 03:06:14 PM »

Hi Ron:

"For receiving narrow-band signals, a high-speed ADC in conjunction with proper decimation to a lower bandwidth actually improves a few performance metrics due to DSP processing gain"

Oversampling, filtering and integration (another type of filtering) improves dynamic range BEFORE decimation.  Performing these steps afterwards simply throws away usable bandwidth.

"With an RTL-SDR, you can DSP process up to 2 MHz of bandwidth on a Raspberry Pi"

So, this would require a minimum of 4 MSPS of processong bandwidth (NOT I/O bandwidth), divided by the number of CPU cycles it takes to execute a single usable usable filter response, like a 16-64 pole FIR filter with DECENT dynamic range (at least 16 bits mantissa), or a somewhat shorter IIR filter (with ringing).

1.5 GHz/10 clock cycles per machine instruction cycle/~1000 arithmentic+move&store cycles per useful DSP filter function = about 150 Khz bandwidth.  Clever algorithm minimization could do better, perhaps - but you would still be very CPU constrained.

You could compromise and get this bandwidth up by 10x by dramatically reducing dynamic range (likely to need an off-board AGC) or use a very short and crappy IRR filter (maybe 4x gain, there), but still, not much really useful processing bandwidth can be done in what is essentially a serial CPU.  Which is why FPGAs and ASICs (or Harvard architecture TMS CPUs) are used for these functions.

The problem is that really useful DSP "radios" with very good, agile filters have very long, very complex processing strings, normally executed in parallel.  A Raspbery Pi cant do that - it's more limited to "Hey, that's a cool demo!" class of processing at very low RF bandwidths.  But in the audio range, it has enough horsepower to do a respectable job of noise reduction or audio filtering, for example.

Here's a simple challenge...

Build and demonstrate a direct sampling tuner/digital demodulator for the USA AM broadcast band using just a Raspberry Pi, one ADC and a speaker.  Kudos if you do this using just a "chopper" (single digital input) for sampling -  this should be very possible on the lower BCB.  The catch is that you have to sample the BCB signal at t(at least) he RF Nyquist rate and need to digitally filter the result to 5 KHz (audio) bandwidth and drive the speaker.  NO SHORTCUTS, like sub-sampling (you may appropriately decimate after filtering, though) or using the speaker as your final RX "clean-up" filter.

You can't really afford two mixers and their associated filters - so direct conversion is your best bet, and it plays into the RPI's sweet spot at audio frequencies.  BCBs are often quite stong, so getting just a few bits out of the chopper via oversampling could help. too.

Nice "maker" article HERE on the topic, BTW:  https://hackaday.com/2021/03/16/the-raspberry-pi-pico-as-an-sdr-receiver/

The author thinks the useful range of the Raspberyy Pi is about 500 KHz.  Maybe.  Depends on what you're doing.

For the DSPs I am paid to design and deliver, it's a non-starter, for a number of reasons.  There are much, much, much better processing choices.

For experimenting with in an amateur radio applications, its a nice gadget that a lot of fun toys have resulted from.  And amateur radio is all about experimenting - with no downside, really.  But don't expect to build a high performance direct sampling HF receiver from just a Raspberry Pi.  Not going to happen.  Not today.  Need to wait a few years.

Best Regards,


Brian - K6BRN



Logged

W6RZ

  • Member
  • Posts: 449
Re: Sherwood Engineering Receiver Test Data
« Reply #28 on: March 20, 2021, 03:27:30 PM »

You'd be surprised what a RPi4 can actually do. I can transmit a fully compliant DVB-T2 signal at 5 MHz of bandwidth and 19 Mbps of data throughput with a Beagleboard-X15 (another small Linux system with about the same performance as a RPi4).

Logged

K6BRN

  • Member
  • Posts: 2231
Re: Sherwood Engineering Receiver Test Data
« Reply #29 on: March 20, 2021, 03:29:12 PM »

Hi Ron (W6RZ):

We've debated before on similar topics....  I seem to remember you've been involved in the "maker" business - lots of fun.

"I'm tempted to Amazon Rob Sherwood an Airspy HF+, just to shut up the naysayers. Based on testing from others, it would probably place next to the FTdx-3000."

By all means, please do this.  I'd really, really enjoy the results.

Just for chuckles, we can compare it to a Flex-6600, too.  The Flex crew includes some very good DSP designers.

Until you've had to design, deliver and service a commercial/industrial/military DSP product that actually has to perform well, all the time, it's very hard to understand how much development, simulation, test, heritage IP and finesse goes into it.  Which is why really good algorithm and ASIC/FPGA designers are paid so very much to do it.

If you've got the "chops" and a history of success, the jobs are there for the asking.  And sometimes they follow YOU, depending on earned reputation.  There's a very small pool of really capable, proven talent and it needs to be grown.  Which is why I highly recommend that young engineers explore this workspace in college and in their early career.

Brian - K6BRN



Logged
Pages: 1 [2] 3 4 5 6 7 8   Go Up