Re: HA1YA Dual converter and Afedri SDR

[Leif Asbrink]

Hello Alex, Roger and All,

This is just a note on the clarifications from Alex. Originally
moon-net, but I copy to linrad and Afedri also...

> I want to clarify some things related to this topic:
> 1. Leif own Dual Channel AFEDRI SDR-net x2, when Roger own Dual Channel
> AFE822x SDR-net,
> those device have different hardware the DDC front end chip AFE8220 has
> better ADC parameters , than DDC chip in AFEDRI SDR.
>
> 2. RF Gain can be set to any value from -10dB up to +35db , value
> depend on overall system gain, so there is no any reason to stay with
> 11db or 14dB gain only..
>
> 3. I suppose that sound card Delta 44 (and probably whole PC where this
> card will e used) should be significantly modified to provide dynamic
> range like numbers described by Leif ...
>
> 73!
> Alex

1 and 2.) http://www.ti.com/lit/ds/symlink/afedri8201.pdf gives
the dynamic range of the Afedri8201 as 102 dB in 3 kHz bandwidth.
That means that the noise is at -137 dBc/Hz. In my video:
https://www.youtube.com/watch?v=_30GIKI60Ug
I find the noise at -139 dBc/Hz at a separation of 500 kHz.

I do not see the relevance of your clarification. Regardless
of the front end we should not expect the Afedri noise much
below -137 dBc/Hz.

3.) That assumption is incorrect. I have written in detail
about how to modify the Delta 44 and that the result is
a 3 dB improvement. BUT that is on the bare soundcard. When
the soundcard is used with a WSE converter chain on 144 MHz,
the noise from the soundcard is less than 50% of the total
noise so the 3 dB improvement on the soundcard translates to
maybe 1.5 dB on the 144 MHz system. The Delta44 is since long
obsolete, but there is the Terratec DMX 6 fire USB which
is about 2 dB better than my modified Delta 44.
http://www.sm5bsz.com/linuxdsp/hware/ensemble.htm
I have run the Delta44 in many computers and only the
D5400XS required an adjustment. I had to place a metal foil
on the Delta44 to prevent it from picking up noise from the
video card. The Terratec DMX 6 fire USB would eliminate
such problems completely.

NOTE THE IMPORTANT THING:

Afedri: -139 dBc/Hz (or -137 or something)
WSE: -165 dBc/Hz at wide range (and that is independent of
the soundcard.)
A 26 dB difference - and the Afedri is reported to be much better!!!

From Rogers posting I understand that overload is because of
signals outside the visible range (although I am not 100% sure.)

I do not think Rogers observations have anything to do
with the WSE converters. I think something connected in
front of them misbehaves. The big problem is that such
mis-behaviour would affect any other receiver with the
problematic unit in front.

Possible problems might be:

1) Sideband noise in high Q filters.
2) AM noise in GaAs amplifiers. Some MMICs are particularly bad.
3) Saturation in amplifiers preceeding the WSE system.
Maybe something else.

Those problems might explain why Roger has seen "overload"
on his old system while the new is OK.

All: Take this mail as a warning!! Regardless of your receive system,
"overload" (presumably a noise floor increase) may be caused
by needless problems in something connected in front of your
high quality receiver.

73

Leif







> On 22/05/17 3:32 AM, Leif Asbrink wrote:
> > Hello Roger and All,
> >
> > On Wed, 10 May 2017 11:25:23 -0400
> > Roger Rehr W3SZ <w3s...@gmail.com> wrote:
> >
> > This posting (see at the end of my mail) has made me very sorry.
> >
> > Something - or perhaps even several things have been wrong
> > with your setup all these years - and I never heard any
> > complaints from you. There is no question that the WSE converters
> > provide a much higher dynamic range than your new system.
> > The problems you describe are due to various errors and I can
> > say exactly what has gone wrong, but the errors are really big
> > and should not have gone un-noticed.
> >
> > The performance specification for the WSE units, the entire
> > chain from 144 MHz into the computer through a Delta44 is
> > given here: http://www.sm5bsz.com/linuxdsp/rx144/rx144.htm
> > At the input: 1 dB compression +13 dBm, IIP3 +27 dBm.
> >
> > Close range dynamic range is 143.5 dBc/Hz (5 kHz) and it is
> > limited by the Delta 44 soundcard. Wide range dynamic range is
> > 165.5 dBc/Hz (110 kHz) Look here for details:
> > http://www.sm5bsz.com/linuxdsp/rx144/test144.htm
> >
> > Every single WSE unit has been tested to meet these specifications.
> >
> > In this video: https://www.youtube.com/watch?v=_30GIKI60Ug
> > at 0:48 into the video you can see that the dynamic range of
> > the Netafedri is 139 dBc/Hz at a separation of 500 kHz.
> > Something like 135 dBc/Hz at 100 kHz and at about 30 kHz
> > separation it is at about 125 dBc/Hz. The reason for the poor
> > performance at close range is noise on the supply voltage to
> > the sampling clock. As I have understood from Alex, this problem
> > is not present in all versions of the two-channel NetAfedri.
> > The gain is set to 14 dB in this recording and it could perhaps
> > be possible to get a higher dynamic range by setting gain differently.
> >
> > If you have a look here:
> > https://www.youtube.com/watch?v=gCvYnhzLSfQ
> > you can see that setting the gain differently will not improve
> > the dynamic range of the Afedri significantly. Here I find
> > the noise floor at 100 kHz separation to be -135 dBc/Hz
> >
> > When you connect a HA1YA converter in front of the NetAfedri
> > you add some gain :"RX overall gain: Typ 25dB variable."
> > If I assume your Netafedri has NF=11 dB (as in my decond video
> > above) your noise temp at the Afedri input would be 3335K. You
> > would want it to increase by 15 dB when switching on the
> > transverter which means that the antenna plus transverter noise
> > at about 300K (cold sky) has to be amplified by a factor
> > 333 (=25 dB) which well matches the HA1YA specification.
> >
> > The noise now 15 dB elevated means that saturation is in this
> > configuration 120 dBc/Hz on the antenna input (15 dB degraded.)
> > The noise floor, -173 dBm/Hz with 1 dB NF means that saturation
> > is at -53 dBm on the antenna input. The HA1YA transverter is much
> > better than required, the Afedri would block for out of band
> > signals that would be harmless to the transverter. Should you
> > ever see any problem, a simple filter on the 14 or 28 IF would
> > improve your system dramatically and would be trivial to install.
> >
> > Anyway, you have now claimed that the HA1YA+Afedri that saturates
> > at -53 dBm on the antenna input is significantly better than your
> > old WSE system.
> >
> > For signals outside the visible range the WSE chain saturates at
> > +13 dBm. That is at a NF of 12 dB. For a decent system noise you
> > have to add 26 dB gain. That means that saturation of your
> > WSE should be at -13 dBm, that is 40 dB above your new system.
> >
> > I am curious about your old system now. How was your amplifier
> > chain arranged? Did you perhaps use MMIC amplifiers? They can
> > give horrible levels of AM sideband noise. Here you can see
> > an experiment with a SPF5043:
> > https://www.youtube.com/watch?v=Mf7ApVuQkNY
> > The noise is -128 dBc/Hz at 10 kHz!!!
> >
> > The HA1YA transverter uses PGA103+ which is much better, but
> > limits the dynamic range to something like 147 dBc/Hz. That
> > is by far better than the Afedri, so not an issue for your system,
> > but it could be the limitation for the HA1YA transverter depending on
> > the quality of its local oscillator and perhaps other things.
> >
> > Dear Roger, your observation that WSE overloads while HA1YA+Afedri
> > does not does not make any sense at all. Something has been seriously
> > wrong with the front end all these years.
> >
> > You write:" I have found with the new system that the limitation
> > on my making contacts is other stations not hearing me when I
> > copy them fully, even though I am putting out 1500 watts here."
> >
> > This is a REALLY serious problem that should have been detected
> > very long ago. There could be three entirely different explanations
> > (maybe more??)
> >
> > 1) Errors in the front end have caused an elevated noise floor
> > causing S/N to be much below what it should have been.
> >
> > 2) To my knowledge the only proper drive routine for the Delta 44
> > soundcard under Windows is for 32 bit Win XP. Timing errors could
> > cause poor detection in digital modes.
> >
> > 3) The correction for the actual sampling clock which is of course
> > not excactly 96 kHz has been incorrect.
> >
> >
> > Dear Roger, I am very sorry learning that you have lost so many
> > QSOs in the past due to set-up errors with the WSE system.
> >
> > There is no magic involved whatsoever, I would very much like to
> > see investigations in which you compare the performance of
> > both systems. I would guess you have antenna mounted LNAs
> > preceeding the HA1YA transverter and by use of 3 dB splitters you
> > could produce identical signals to two separate systems.
> >
> > It is easy with Linrad to evaluate extremely small S/N differences
> > and I think it will not be difficult to find the exact reason
> > for the poor performance you have suffered from all these years.
> >
> > 73
> >
> > Leif
> >
> >
> >
> >
> >
> >
> >> Hi All,
> >>
> >> Last October we had a discussion on Moon-Net about the theoretical pros
> >> and cons of a system consisting of the Afedri AFE822x SDR-Net ( Dual
> >> Channel ) and an HA1YA dual channel receive downconverter. This
> >> discussion was engendered by the incredibly good results that Conrad
> >> PA5Y obtained with this system.
> >>
> >> Since then I purchased an Afedri AFE822x SDR-Net ( Dual Channel ) and
> >> an HA1YA dual channel receive downconverter and I have used them for
> >> several weeks.
> >>
> >> I had planned to do an A/B test comparing them to my "old" WSE receiver
> >> that I used with Linrad.
> >>
> >> But my results with the Afedri AFE822x SDR-Net ( Dual Channel ) and an
> >> HA1YA dual channel receive downconverter have been so good that I have
> >> retired/removed the WSE from my system and I no longer plan to do an A/B
> >> comparison because it is clear that the new receive combination meets my
> >> needs and I have had NO problems here with overload with this new system.
> >>
> >> I have found with the new system that the limitation on my making
> >> contacts is other stations not hearing me when I copy them fully, even
> >> though I am putting out 1500 watts here.
> >>
> >> By monitoring the JT65 EME chat page I am able to see who is on the air,
> >> and I also find that the performance of the new system in this regard is
> >> superb and surpasses my experience with the old equipment.
> >>
> >> Previously it was often the case that I would copy only one side of a
> >> QSO, and now I generally see both sides of any QSO that is completely
> >> successfully by others, and frequently (but not always) see both sides
> >> of the QSO even when one of the QSO partners is not copying the other so
> >> that the QSO attempt is unsuccessful.
> >>
> >> I had previously installed cavity-input preamps because of problems with
> >> strong out-of-band signals causing a rise in the noise floor and
> >> production of spurious signals within the EME passband. Even with these
> >> cavity-input-preamps, I had some residual problems with overload with
> >> the WSE receiver, but I have seen NO such problems with the new receive
> >> combination.
> >>
> >> I have never before gone for a period of several weeks without problems
> >> with overload rearing their ugly head, and I have never so consistently
> >> experienced such excellent receive performance at my current QTH, and so
> >> I am extremely happy with this new receive gear. At this point I feel
> >> that this is the most that I have gotten for my money in buying ham
> >> radio gear ever.
> >>
> >> I personally believe that the "I copy down to -xx dB" statements that
> >> folks make are pretty much meaningless, but as I know someone will ask,
> >> I will here give the information that I copy down to -30 dB (WSJT scale)
> >> with the new system.
> >>
> >> I apologize for not doing an A/B test as I had promised some of you.
> >> But I have too many uncompleted projects that need my time to spend time
> >> on such a project when I have seen such outstanding performance day
> >> after day with the new system. I do not need to jump out of an airplane
> >> with and then without a parachute to prove that I would do better to use
> >> a parachute (if you try that experiment do not use the reverse order or
> >> your experimental data will be incomplete).
> >>
> >> I know that others in a different RF environment (like Mike, W9IP
> >> reported) may have overload problems with this (or any) system, but here
> >> I do not have such problems with this system.
> >>
> >> Before trying this system, I was a skeptic regarding the likely
> >> performance of this Afedri AFE822x SDR-Net ( Dual Channel ) and an
> >> HA1YA dual channel receive downconverter system in my RF environment.
> >> After using the system for several weeks, I have been converted and am
> >> extremely impressed with the Afedri-HA1YA combination and will not go
> >> back to my previous system. The only reason I would ever sell either
> >> component of this new system would be if I were quitting 2M EME, or if
> >> either Gabi or Alex were coming out with an even better version of their
> >> hardware. In that regard, Gabi is about to release a
> >> dual-receive-channel 144 MHz-28 MHz transverter so if you are thinking
> >> of a purchase, take a look at that first. These are just about to be
> >> released.
> >>
> >> The "rest" of my system has not changed. It still consists of 2x2 array
> >> of M2 2MXP32 with on-the-tower cavity input preamps for both H and V.
> >> 1500 watts from LZ2US amp, fed by K3 which is only used for transmit and
> >> not for receive. Linrad, MAP65, WSJT for signal processing as described
> >> here:
> >> http://www.nitehawk.com/w3sz/LinradMAP65.htm
> >>
> >> Information on the Afedri is here:
> >> http://www.afedri-sdr.com/index.php/new-afe822x-sdr-net-dual-channel
> >>
> >> Information on the HA1YA dual-channel receive downconverter is here:
> >> http://www.ha1ya.hu/htmkepek/dual_rx.htm
> >>
> >> Information on the new HA1YA dual-channel-receive transverter is here:
> >> http://www.ha1ya.hu/htmkepek/me2xp_spec.htm
> >>
> >> Finally, thanks to Conrad PA5Y for making us aware of this system by
> >> posting his results last autumn!
> >>
> >> Have a great day all, and 73,
> >>
> >> Roger
> >> W3SZ
> >>
> >> On 10/29/2016 11:40 AM, Roger Rehr W3SZ wrote:
> >>> Hi All,
> >>>
> >>> I want to thank Alex and Alex and Conrad and everyone else for an
> >>> excellent discussion from which I learned a lot, as I am sure many did.
> >>>
> > _______________________________________________
> > Moon-Net posting and subscription instructions are at http://www.nlsa.com/nets/moon-net-help.html
> >
> >
> > ---
> > This email has been checked for viruses by AVG.
> > http://www.avg.com
> >
> >
>
> _______________________________________________
> Moon-Net posting and subscription instructions are at http://www.nlsa.com/nets/moon-net-help.html
>

[David Anderson]

The perplexing thing though, is that Roger appears to use exactly the same preamps etc in front of both systems, correct me if I am wrong.

Roger, I know you were originally planning on doing an A/B test and decided that it was pointless when you saw those immediate improvements with the new system, however in view of this I think there is even more reason to do such tests, because it does not make any sense left like this. I am sure all 'sides' would welcome such a test, though we shouldn't make them seem like adversaries, rather experimenters with a common goal.

73

David GM4JJJ

[Ermanno F.]

Hello to all.

I've been using WSE receivers with Delta-44 for years, and the system has
always been the
best available. Since 2015 using also Afedri Dual Channel AFE822x with the
Javornik transverter.
I quickly found the input of Afedri very noisy and the definitive solution was
to bypass
the internal VGA. I added between the transverter and Afedri a bandpass filter
and a low-noise pre-amplifier.
The calculated by Leif are definitely accurate and the WSE dynamics is higher
and better than Afedri; you notice it
immediately when there are strong signals in the range. Instead when the band
is calm
with Afedri/Javornik i receive signals that I can not receive with WSE. Use at
the same time
in parallel the 2 systems with the same antenna and preamps and parameters for
the best distribution of gain. I had also prepared a comparative excel sheet
in 2015
but I can not find. However, I have retrieved an image (attach) where you can
see a low signal
better decoded with Afedri. I do not know the reason but the practical this is
so.
I noticed that with JT65 signals at the limit, their decoding depends greatly
on stability
 frequency of the receiver (will this be the reason?) The WSE are very stable
once they are
reached the temperature and if you try to receive low signals before about 30
minutes of power ON
do not decode anything. This is my small contribution only from the practical
point of view of reception
without entering the merit of instrumental measures.
Help you figure out why this happens?


73 Ermanno / ik7ezn


>----Messaggio originale----
>Da: "Leif Asbrink" <le...@sm5bsz.com>
>Data: 29/05/2017 2.48
>A: <moon...@mailman.pe1itr.com>, <lin...@googlegroups.com>,
<afed...@yahoogroups.co.uk>
>Ogg: Re: [Moon-Net] HA1YA Dual converter and Afedri SDR

[Leif Asbrink]

Hello Alex,

> Again, AFE822x does not use AFEDRI8201 DDC chip, it has inside other
> *dual channel* AFE8220-Q1 DDC chip ...
Interesting. The data sheet specifies 105 dBc in a 3 kHz bandwidth:
http://www.ti.com/lit/ds/symlink/afe8220-q1.pdf
That means -140 dBc/Hz which is a single dB above the
-139 dBc/Hz that I find on my unit (and I looked for reciprocal
mixing so there should be a small difference in this direction.)

> In my experiments i have never met a Sound Card (including 24-bit) that
> really can provide dynamic range we can expect from the bits numbers...
> In same time I have never owned / tested Delta-44...
Sure. That is totally impossible!!
The (modified) Delta44 can be classified as a 18 bit card - and
the two lowest bits mostly contain noise.

Most soundcards are bad, they have excess noise or they have
noisy reference voltages or noisy sampling clocks.

There are some good cards however:
http://www.sm5bsz.com/linuxdsp/hware/ensemble.htm
The top five on the list. Only two of them have 4 channels.

Now, with the WSE units the quality of the soundcard does
not affect the wide range dynamic range because there are
filters in the RX2500 that prevent signals more than 45 kHz
from the center to reach the soundcard. Fig. 2 here:
http://www.sm5bsz.com/linuxdsp/rxiq/antiali.htm
Suppression is 30 dB or more.

The wideband dynamic range of the WSE system is about 25
dB better than the wideband dynamic range of any Afedri
unit. Rogers report that the Afedri works better for
him in terms of no "overload" needs an explanation. Please,
Roger, in what way did overload show up with your old system?

73

Leif





>
> 73!
> Alex

[Leif Asbrink]

Hello Ermanno,

There are several things to worry about. I would split them into
two classes:

A) Effective S/N (with effects of blocking, reciprocal mixing and
system noise figure included.)

B) Decoding efficiency of digital data.

Roger reports problems of both cathegories. I have big problems
with understanding what is behind the "A" problems.

The "B" problems is another thing:
The WSE system was designed to allow it to be phase-locked to
a 10 MHz standard, but with the low interest I met with the project
I decided to skip this add-on. As a result, the WSE units need a very
stable temperature and a long warm-up time to be perfectly stable
for digital modes. To some extent that is due to the decoder because
the decoder could search for signals that drift in frequency. (I have
a feeling it already does, but it could then search a wider range of
frequency drifts.) Your observation that about a 30 minutes warm-up
time is needed is not surprising at all.

The "B" problem may also be caused by soundcard errors. As far as
I know, the only reliable driver for the Delta44 is the Win XP 32 bit
driver. It is fine under Win XP 32 bit, and once I was running it
sucessfully under 32-bit Windows 10 installed in compatibility mode.
(I have not been able to reproduce that.) All other combinations
of Windows versions and Delta44 drivers cause glitches and surely
they constitute errors that may degrade detection. There are many
other soundcards than the Delta 44. Dynamic range would be a bit lower
for the visible range (±45 kHz) but the wide range dynamic range would
not be affected.

-----------------------------------------------------------------------

I do not believe in black magic. Any observed performance difference
that is statistically significant must have an explanation. Not knowing
what is going on makes me frustrated. Discrepances of 25 dB (3000 times
in power) are just un-thinkable....

In EME, the normal situation is that the noise floor is totally
dominated by the antenna+LNA. That can be certified by switching off
the DC supply to the LNA (while pointing to quiet sky) Noise should
go down by at least 15 dB, maybe 17 and then one knows that the system
NF is the NF of the LNA within 0.1 dB or so. (When comparing more
receivers on one antenna, one should of course observe at least 15 dB
on all of them.) This means that in EME, any "A" type problem has
to be caused by undesired signals of some kind. Not likely to
have any effect, but there is LO leakage at 149 MHz at -50 dBm from the
RX144. Feeding such small power into the output of a LNA should have
no effect, but the signal is there (2nd LO harmonic.)

In recent times I found something completely new to me. AM noise
in GaAs FETs. I think that is actually the practical dynamic range
limitation for most high performance EME stations today although
nobody is normally looking for it. It seems everybody believes that
sideband noise is always phase noise.

Roger reports something unexpected. VERY unexpected. There is something
here to understand. I can not imagine that Roger is the only one to have
suffered from performance degradation due to whatever unknown phenomenon
present. I find it unlikely that the "A" type problem is related to the WSE
system. Presumably there is something here for everybody to be aware of.

73

Leif

[Joe Taylor]

Hi Leif, Ermanno, Roger, and all,

I do not have an Afedri receiver, so can offer no useful comment on its
performance. But I have used the WSE receivers with (unmodified)
Delta-44 for some 13 years, and I consider the system performance to be
extremely good. There are plenty of strong signals in my area, both in
and outside but near the 2-meter band. I have never had overload or any
other dynamic-range problems.

As for Ermanno's reported case(s?) where Javornik/Afedri/Linrad/MAP65
decoded signals that WSE/Delta44/Linrad/MAP65 failed to decode:

Yes, frequency stability is important for decoding the weakest JT65
signals. I have not had problems with the WSE receiver in this regard,
perhaps because I never turn mine off unless I am away from home for
several days or longer. To reduce heat buildup in the closed WSE boxes,
I have inserted wooden spacers between each of the stacked units; and
somewhat against Leif's advice, I run mine with power supplies set at
+/- 12 V DC instead of the recommended +/- 15 V. No doubt this reduces
the dynamic range by a small amount -- perhaps 2 dB? -- but in practice
there has been no observable deleterious effect.)

Anyway, Ermanno, I believe your screen shots do *NOT* provide evidence
that Javornik/Afedri/Linrad/MAP65 has outperformed
WSE/Delta44/Linrad/MAP65, at least in this particular case.

The entries under "KV" show that all decodes on the Afedri screen were
multi-period average decodes. It's impossible to tell from the screen
shots exactly how your two setups were configured and used, and
multi-sequence averaging in MAP65 is somewhat complicated. However,
it's clear that all of these decodes are suspect. The reported "DF"
values jump around irregularly: -117, -178, -230, -125, -250, -111, -184
Hz. Moreover, your decodes show DG0JMB calling CQ in both even and odd
UTC minutes! Surely the real DG0JMB signal was not frequency-jumping
like this, and he was not calling CQ in both sequences.

Bottom line: I don't know exactly what was going on in your example.
But your WSE/Delta44/Linrad/MAP65 results look perfectly trustworthy,
and your Javornik/Afedri/Linrad/MAP65 results certainly do not.

-- 73, Joe, K1JT