Perseus SDR Receiver Test Results

[Steve - WB3HUZ]

I'm glad to know Flex is a far superior reciever because it transmits....

It's a superior radio. Nice try though.

Scientific Atlanta made boxes that would monitor tens of  MHz in real time back in the 80's. Welcome to the late 20th century.

[WA1GFZ]

Well pretty quiet on 160 so far tonight so ran some quick numbers. First I don't trust my numbers unless they are the same three times. Also rough readings may be off a couple dB.
1. MDS on 75M about -136 with 500 Hz filter iin CW mode.
Noise floor about -140 dBM.
2. Two tone dynamic range two signals 3.870 and 3.890 MHz. 
(3.910. 3.850 spur)       
Spur comes out of the noise floor when both signals hit -80 dBM so pretty crappy around 60 db dynamic range. I notice weird effect so run levels higher.
When both Run up to -27 dBM spur at -121 dBM  (approx 94 dB)
The spur level tracks weird as one generator gets higher than the other. I tried to find the worse case for my results.
3. Then I selected "dither enabled" Spur now comes out of the noise floor at about -65 dBM so about 70 dB dynamic range.
At -30 dBM signals spur is at -120 dBM (about 90 dB dynamic range)
4. Now I also turn on "random enabled" With both signals at -40 dBM the spur is in the noise. I then increased the signals to -25 dBM bringing the spur up to -130 dBM. (dynamic range 105 dB)
Test configuration 2 HP8640B generators with the same HB combiner I've been using for 30years. 6 db pads on all three ports. 
5. 1 final test. Listen to a signal at the MDS with 1 generator then tune the second one up to 8 MHz. Generator 2 generates spurs when set above -14 dBM.  126 dB above the noise floor.
No AGC function yet and the input pad must be out of the circuit with that MDS. Software does not control it yet. The Alex filter module will add filters and step attenuator for AGC with TR control.
Yup, this looks interesting guys.
So Jay, I also notice some weird effects and don't know what these two extra software functions do to performance but haven't had this kind of fun since you checked out my hot rodded SB303 with a pair of URM25s in the ARRL lab in 1977. 

[W1VD]

Frank

Interesting...digesting your data...

With the Perseus I found the MDS and 20 kHz blocking DR measurements just like an analog receiver...once the 'AGC down to the noise floor' issue was corrected.

The two-tone measurement was a bit more 'interesting'. With the dithering off I was seeing two-tone DR in the mid to high 80s...dithering enabled yielded 99 to 100 dB. Since dithering enabled has no effect on MDS or blocking DR I leave it on.

The third order IMD products don't seem to act the same as an analog receiver...where every dB change in input tones makes a 3 dB difference in the IMD product. It appears to me that the IMD product is actually very weakly audible (below the noise floor) at input levels significantly below those that cause the products to be 3 dB above the noise floor. These signals would be inaudible on an analog receiver. Remember that this is with just the signal generators connected to the receiver and no antenna. Once an antenna is connected the noise floor is raised and these 'phantom' IMD products would presumably be inaudible.

Did some tuning around tonight in the 160 meter Stew Perry contest looking for any possible receiver problems but there weren't that many really BIG signals like in the ARRL 160 meter contest a few weeks back. Further on the air tests will have to wait for a more 'big signal' populated contest.       



         

[ab3al]

my commodore amiga 500 kicks all your radio's asses

Huz is right  if hpdsr is so fricken great and flex is so lame why are their so many flexes being heard on the air and
almost no hpdsr's

OH THATS RIGHT.. YOU CANT BUY A FINISHED RIG..

look, I check into their site and they have some good ideas but thats all its been for the last few years ive been
looking.. produce a radio that works at least as well as the flex .. sell a few thousand and then there can be a comparison. Until then your arguments are about as valid as my 2 nephews arguing over who's Transformer can kick the other ones ass.

I dont have a pocket protector
i dont wear highwater plaid pants
and i approve of this message

[W1EUJ]

silly.

[ka3zlr]

Naaa, no thanks.... Still I think, for receiving,

 It's just as much fun to snap on an SX-28...an wait for the Warm sound of Electric Radio...


73
Jack.

[WA1GFZ]

Jay,
I just got a message that my attenuator relay might have a problem making it unable to select the pad....not that I will ever need it.
I try some close in dynamic range numbers today.

[W1VD]

What's the A/D in Mercury? QS1R? Been studying this morning - think I'm ready for the next 'night class' with Stu. Look for you on 75 if you're around this afternoon...or 160 tonight. 

[W1VD]

Some additional measurements...

Phase Noise (500 Hz bandwidth)

Freq. Offset	Phase Noise
500 Hz	-138 dBc/Hz
1 kHz	-141 dBc/Hz
2 kHz	-143 dBc/Hz
5kHz	-147 dBc/Hz
10 kHz	-150 dBc/Hz
20 kHz	-153 dBc/Hz
50 kHz	> -153 dBc/Hz
100 kHz	> -153 dBc/Hz

Signal source is a crystal test oscillator.

Two plots below showing spurious free dynamic range - close in and further out. Perseus claims 110 dB...measured > 112 dB.

[WA1GFZ]

Jay,
This Perseus may be cleaner than the R390A.
My RA6830 phase noise goes through -100 dBC on the display at about 2 KHz.
I have not tried this yet on Mercury. BTW Both mercury and QSR1 use the 2208 A/D. Your's is one model older 80 MHz. 14 Bits.
Today I tried close in dynamic range at 5 KHz spacing. It was easily 100 dB but every few seconds I got an audio transient sneak through that kicks the S meter up. I had to drop the level down a lot to make it stop. I reported to the software guys. I reported this to the mother ship.
Yup, Stu could write a book for us dumb hardware guys.
 

[KF1Z]

,,,,,,,, every few seconds I got an audio transient sneak through that kicks the S meter up. I had to drop the level down a lot to make it stop.

Could be an FPGA timing issue?

The software dudes will know...

Had something that sounds similar here...
Turned out to be some timing timing problem within the fpga code...

The hpsdr is using the Cyclone III  fpga, right?

[WA1GFZ]

I think that is the FPGA. I got an email from one of the designers with some things he wants me to try. He also said my MDS & dynamic range numbers agree with his.

Jay, I assume your phase noise numbers include the BW correction factor of 27 dB??

[N3DRB The Derb]

the best radio is the one you have.

[W1VD]

Yes...corrected for the 500 Hz BW...27 dB.

Those phase noise numbers are truly something to behold.

The Perseus software could use a tweak for better cw reception. When set to fast, the AGC tends to overreact a bit just at the filter edges when encountering a BIG signal.  No problem on medium or slow. Minor tweak I'm sure...or a setting I haven't discovered yet.

You're right on that Derb.   

     

[WA1GFZ]

All this stuff in software makes me feel helpless...software pukes have taken over. I wish I had notheing to do so I could learn new skills.
My uncle Joe makes some great wine....maybe we need to send Stu some.

[W3RSW]

send me some 

Happy New Year you guys.
Derb's right.

My Ocean Hopper was the best ever !

...and one of my QS1R's obvious merits is listening to bluegrass on 980khz, one click away to 3733, or 3885..  all with pre-set bandwidths, mod type, band span, etc.  Of couse an entry level computer is helpful.
As far as receiver only, hookup wire zoo, T/R lashups, software learning curves up to dumb level for operating and somewhat smarter for developing... um, I mean, we are hams, right?

"I find no need to trash that which is not on my side of the greener pasture."  If I had an SX -28 I would treasure it.

[WA1GFZ]

Mercury Phase Noise
 I found the 5 MHz reference coming out of my Racal RA6830 about the cleanest in my shack. HP8640B had some spurs out around 20 KHz that was messing up the numbers.
I drove Mercury with about -22 dBM with the preamp on. It was in cw mode with 500 Hz. bandwidth. 500 Hz bandwidth correction factor is -27 dB so it was added to the levels I measured. It looks quite good and didn't notice any difference changing to Penny as a clock source.
The Final numbers
5.0000 MHz           0dB
4.9999                  -27.3
4.9998                  -33.7
4.9997                  -62.4
4.9996                  -126
4.9995                 -129
4.999                   -139
4.998                   -141
4.997                   -142
4.995                   -144
4.990                   -145
4.980                   -144
I don't think anyone can complain with these numbers.
Jay, The 6830 reference is a 20 MHz VCXO phase locked to a TCXO so your source may be a bit cleaner.

[W1VD]

Frank

Your numbers look good - no doubt an xtal would be better. Neglected to mention earlier...I used a signal level within a dB of full scale (adc clip) for the phase noise test and the spurious spectrum displays.

Rick

Have an SX-28A and it's my favorite BA AM receiver. The 14 kHz wide i-f bandwidth, bass 'in' activated and pp 6V6's make for quite the listening experience. The AVC curve allows about 10 dB output change for signals from 10 uV to 1000 uV before it levels off. I find this to be desirable as it adds another dimension to the listening experience (hearing signals fade up and down rather than just seeing it on an s meter)...along with the usual selective fading, etc. Modern receivers with super flat AGC don't sound nearly as good to me. 

Glad I don't have to pick just one receiver - that would be tough. I enjoy them all! 

   

[AB2EZ]

Frank
Rick
Jay

With either type of SDR , once we get some mastery of the software, we should be able to emulate any classic receiver we want to emulate. As we all know, the existing SDR software emulates a diode detector in regular AM reception mode.

We should be able to emulate, for example, the AVC (a.k.a. AGC) behavior of the SX28A that Jay described. In fact, that should be almost trivial to do if you have even basic abilities with one of the flavors of SDR software (which I don't yet have).

Perhaps someone will create a high level/graphical programming language for SDR's that can be compiled to produce the executable software.

Separate issue (food for thought... I don't think there is an obvious answer):

Which would be better for receiving signals (separate from the issue of which would be better for a broadband spectrum display):

a) An SDR that samples the r.f. at 196.608 MSPS using an A/D having 16 bits of accuracy (possibly including a preamp and/or a pre-selector/roofing filter)?

b) An SDR with a state-of-the-art front end (like the IC-7800) that samples the i.f. at 192 kSPS, using an A/D having 24 (or more) bits of accuracy.

My conjecture: If the sampling clock (case a) has the same phase noise limitations as the local oscillator clock (case b) ... both assumed to be state of the art... you will get better results with the i.f. based approach for ham radio applications.

I believe that if you use a preselector in case a), then the larger sampling rate  196.608 /.192 = 1024 corresponds (after decimation etc.) to 5 extra bits of A/D precision in terms of performance. I.e. 1 bit of extra precision improves the SNR by 6 dB, while a factor of 2 in sampling rate (when used properly) improves the performance by 3dB. Thus each additional bit of accuracy corresponds (in performance improvement) to a factor of 4 in sampling rate. A factor of 1024 in sampling rate is (writing it a different way) a factor of 4**5.   

Best regards
Stu

[Steve - WB3HUZ]

Right on bro. All this crap about mine's better and yours is not modern is silly. 99.9% of amateur radio is old news. Being ignorant of this fact get many hams in trouble. Get over it and move on.

the best radio is the one you have.

[WA1GFZ]

Stu,
I thought the IF approach was better since the A/D sees limited bandwidth but you have to live with synthesizer phase noise effecting close in dynamic range. I have not tried Mecury on my BB Racal front end yet. The software guys are sending me beta drops to test. I think in the long run a high q preselector would be the way to go once the software works.
Jay, I was a bit below saturation also but I didn't run calibration at 5 MHz. fc

[Steve - WB3HUZ]

Which would be better for receiving signals (separate from the issue of which would be better for a broadband spectrum display):

a) An SDR that samples the r.f. at 196.608 MSPS using an A/D having 16 bits of accuracy (possibly including a preamp and/or a pre-selector/roofing filter)?

b) An SDR with a state-of-the-art front end (like the IC-7800) that samples the i.f. at 192 kSPS, using an A/D having 24 (or more) bits of accuracy.

Stu:

Receiving what type of signals? Typical narrowband ham radio signals? Then option B. Wideband, non-ham signals? Then A.

The K3 shows what a good front-end and narrow roofing filters can do for narrow, closely-spaced signals. And it is really only scratching the surface.

[AB2EZ]

Steve
et. al.

I am assuming that for most ham radio applications, we are primarily interested in the following

I. Receiving a signal, in the presence of noise, selective fading, and interference... whose total (two-sided) rf bandwidth is less than 50 kHz. This would include: the signal itself, plus portions of the offending noise and interference. This assumes that we might wish to measure the noise and interference in a wider band of frequencies than those occupied by the signal, to estimate their characteristics, in order to perform adaptive noise and interference cancellation.

Example 1: An interfering signal, 20 kHz off frequency, is producing sideband components that are within +/- 5 kHz of the center frequency of the desired signal. In this example, these sideband components are not distortion products... just high frequency components of the audio/modulating signal being used by the (strong) interfering signal that is centered 20kHz away from the desired signal. We may want our SDR to be able to receive both the desired signal and the interfering signal simultaneously in order to perform a software-based spatial-temporal interference cancellation algorithm.

Example 2: To adaptively cancel a man made noise signal (e.g. power line noise) it may be useful to simultaneously receive both the desired signal as well as a wider-bandwidth slice of the interfering signal. This would by particularly true if there was relatively little other interference.

II. Panadapter displays of various types for various purposes

I am assuming that wideband (e.g. 1-30 MHz) panadapter displays, as used by ham radio hobbyists, are mostly decorative, and not worth the trouble of implementation using a swept local oscillator.

The relevant technology is moving forward very rapidly, and we have varying degrees of "attachment" to our traditional receivers (a.k.a. boat anchors) and our traditional engineering knowledge and instincts. Therefore, I am not surprised that the opinions regarding SDR technologies are quite diverse. Decades ago, the same reactions occurred among engineers during the transition from analog switching technoloy to digital switching technology (circa 1975-1985), during the transition from circuit switched voice to packet switched voice (still underway), during the emergence of cellular telephone services (circa 1985-1995), and during the transition from the PSTN to the Internet (mostly completed).

I remain neutral (personally) in this regard. I enjoy using the 50kHz i.f. output of my beautiful Drake 1A (circa 1957) in combination with my (circa 2008) SDR-based synchronous detector. With the Power SDR panadpater display, I can easily move the i.f. of the Drake 1A to the clearest (in terms of interference) side of the carrier, and actually receive decent sounding AM. It is also fun to "see" the shape of the Drake 1A's filter by observing the displayed noise floor.

At the same time, I am (personally) very much interested in the theoretical limitations of varous approaches, the practical limitations of various approaches, and the prospective implementation of adaptive automatic spatial-temporal (MIMO) noise and interference cancellation... including the possibility of automatically cancelling out certain obnoxious AMer's whenever my receiver detects their presence 

Best regards
Stu

[Steve - WB3HUZ]

Most of what you are proposing doesn't really fall into either of your A or B scenarios. Rather, as you noted, it's about multiple antenna and/or receiving channels. This means adaptive antenna systems. If done correctly, what receiver topology follows may not be that important.

Adaptive interference mitigation has been around for a long time but most of it has been at frequencies other than HF. If you can find them do some reading on work by William Gardner on cyclostationarity.

[AB2EZ]

Steve

What I propose is as follows (simplest case of two antennas used to adaptively cancel out one interfering signal)

a. Two antennas far enough apart and/or with differing polarizations and/or with different gain patterns... to provide to two "independent" received signals (particularly with respect to the amplitude and phase relationship between the desired signal and the interfering signal)

b. Two analog receiver front ends (rf => i.f.)

c. Two A/D converters

d. Adaptive interference processing comes to life as an algorithm executing on a digital signal processor using the multiple (two in this case) digitized i.f signals


I have chaired several patent committees in large companies, and I have served as a consultant to venture capital companies considering investing in various concepts... and I believe that 99% of the things I have reviewed over the last 30 years are old ideas whose time has (or has not) come.

The issue (in my opinion) is not whether any of these ideas are new to the world... the issue is whether the combination of the available technology and the current needs of prospective users of the applications have come together to make this an opportune time to bring these old ideas to life in real implementations.

Stu