[minima] Re: Testing KISS Mixer - Some More Data
- From: "Joe Rocci" <joe@xxxxxxxxxx>
- To: <minima@xxxxxxxxxxxxx>
- Date: Thu, 31 Jul 2014 11:22:50 -0400
My guess is that the IMD performance of the ADE-1L+ is worse, as it's usually a
direct function of LO drive. This will be especially important in the xmit mode
where you need as much clean output as you can get in order to minimize high
gain requirements in the transmitter amplifier chain. High gain not only leads
to stability problems, each additional stage adds to current draw and
distortion.
It's kind of academic though, since the Si570 can output about +14 dbm, which
is enough for even some higher-level DBM mixers.
Joe
W3JDR
----- Original Message -----
From: Mvs Sarma
To: minima@xxxxxxxxxxxxx
Sent: Thursday, July 31, 2014 10:41 AM
Subject: [minima] Re: Testing KISS Mixer - Some More Data
Joe, think of ADE-1L+ that works at 3Db LO mixer as against ADE-1thatdemands
7db Lo level.
On Thu, Jul 31, 2014 at 5:16 PM, Joe Rocci <joe@xxxxxxxxxx> wrote:
Steve
Thanks go to you and Thomas for your legwork characterizing a very
interesting circuit. I intend to do my own characterization on this at some
point, but right now I'm pre-occupied with other interesting experiments. Based
on some early simulations on PSpice, I decided that the J310 version of the
KISS mixer is very novel, but doesn't really offer significant performance or
cost improvements over a diode DBM except the potential requirement for less LO
drive power (see below) and a savings of one transformer. However, I do think
it might be a very different story if CMOS bus switches were used instead of
J310's, as described by Trask.
I have only a couple comments about your tests:
1) The signal level you're driving into the mixer seems a bit high at about
-2dbm mixer output. Most level 7 diode mixers are compressing at this point, so
I wonder if that might be the case here. It would be interesting if you could
drop the input drive by 10-20 dB and see if the mixer output tracks
exactly.Overdriving can not only limit the output level, it can make the
undesired products look larger in proportion than they actually are and also
generate additional products. If you do this test, it would be interesting to
report the drive level (or output level) where the output is compressed 1dB.
From here, there are some rules of thumb to estimate IMD performance, if the
DBM experience holds true.
2) I notice some spurious mixing products that aren't part of the expected
performance. I suspect that your signal generator might not be all that
pure...can you look at it on the spectrum analyzer? Also, I'm puzzled by why
some of these products appear on the lower side of the desired output in 76.bmp
and on the upper side in 75.bmp.
Also an editorial comment: You express your LO drive in dbm, but this is
misleading as dbm is a unit of power referenced to 1 milliwatt. In conventional
parlance, it only has meaning if the load impedance is 50 ohms. In this
circuit, you're driving a very high impedance, so all that matters is Pk-Pk
voltage. You could just as well drive the LO port with your generator turned
down 10 dB and changing the LO input transformer turns ratio to end up with the
same Pk-Pk voltage on the gates of the FETs.
Finally, if you want to go another step and possibly help yourself and the
group out, it would be nice if you could perform the same tests with either a
homemade or a purchased DBM. I think there might be a performance improvement.
Joe
W3JDR
----- Original Message ----- From: "Steve VK2SJA" <vk2sja@xxxxxxxxxxx>
To: <minima@xxxxxxxxxxxxx>
Sent: Thursday, July 31, 2014 3:54 AM
Subject: [minima] Re: Testing KISS Mixer - Some More Data
Hi All,
Alright I've had the time to do one additional test. This is still testing
the mixer with the Joe W3JDR alternate bias scheme setup.
I have attached another two S.A. screen grabs.
The first 75.bmp is a complete duplication of the test setup for the
previous test 74.bmp as sent last time. i.e. The test of the alternate
bias sent in the post below. I simply wanted to check for consistency. I
seem to have it. As all results seem to be within 1dBm. The only thing
that I changed was that I manually nudge Marker number 1 down from 14.3 to
14.2Mhz to align with the LO frequency.
To re-cap. Test done with Si570 injected into LO port at +14dBm and a
20Mhz sine wave signal at -10dBm (more about this in a minute) injected
into the IF port of the mixer. That's -10dB at mixer after traversing a
50ohm 6dB attenuator. Simulated TX mode if you will.
The next screen shot 76.bmp is exactly the same test setup. Except that I
have pulled the IF and RF port connectors and reversed them. So now a
20Mhz RF sine wave is on the RF port and the DSA-815 is looking at the IF
port. Normal RX if you will.
The screen shot sums it up. No change in the wanted RF product. Only about
1dB change in the 20Mhz RF signal but about an additional 20dB attenuation
of the unwanted 34.2Mhz LO signal going in this direction.
These results seem consistent with both predictions and Thomas's own
measurements, in general even if not in precise values as reported here:-
//www.freelists.org/post/minima/How-to-adjust-the-bias-of-the-KISS-mixer-in-the-Minima-Video,18\
(Thomas was using a Rigol Signal generator with different levels I
believe. And perhaps a sine wave LO signal instead of the square wave of
the Si570).
Having seen this first hand I'll ask the same question Thomas did. Should
we be reversing the mixer direction in the Minima so that the higher
unwanted LO level appears at the Crystal Filter in RX mode? Instead of at
the RF port during TX?
Final thing I wish to mention is that I have discovered I have a
measurement error. I don't think that this will be significant in the big
picture but in the interest of full disclosure...
Thinking about my results I became concerned about why I was seeing 5.1dBm
conversion loss when I've seen figures of 6dB and 8dB reported for perfect
component software simulations.
So I went looking. I discovered that -10dBm on my homebrew W7ZOI simple
power meter is being declared as -8.58dBm on the Rigol DSA-815.
So adjusted my 5.1dBm would now be closer to 6.52dBm conversion loss.
Probably still too good to be true but at least moving in the right
direction.
The source or the error? Two things actually.
My homebrew W7ZOI simple power meter based on a AD8307 IC has never been
"properly" calibrated. Something I hope to attend to this weekend.
The other issue is that my signal generator is actually the marker output
of a Sweep Generator. It's an older commercial test set used for aligning
TV sets. The marker output is clean and stable, also adjustable (but not
calibrated) and I was very thankful when someone just gave it to me for
free! Nobody fixes the IF in TV sets anymore it seems. However the sweep
portion is still running and I rather suspect that this RF level is
leaking across and being seen by the power meter. So I'm going to try
modify the unit this weekend so that I can turn the sweep function off.
Anyway, all part of the fun of learning to play the RF game properly.
Over this coming weekend I'll also revert the mixer to the original Minima
design and re-test in both directions. It's going to be a busy weekend :-)
73, Steve
steve,
can you swap the IF and RF ports and remeasure? in both configurations?
as an aside, even these figures are not so bad if u throw in the LPF
into
the RF port.
- f
On Tue, Jul 29, 2014 at 5:32 PM, Steve VK2SJA <vk2sja@xxxxxxxxxxx>
wrote:
Hi All,
Well I finally managed to build a KISS mixer and make some
measurements.
Attached are a couple of Spectrum Analyzer (SA) snapshots and a
picture
of
my mixer.
Test setup:
LO being generated by actual Si570 on Minima logic PCB. Measured
output
into 50ohm W7ZOI simple power meter at +14dBm. This seems in agreement
with a Clifton Laboratories report here:-
http://www.cliftonlaboratories.com/si570_kit_from_k5bcq.htm
IF signal at 20Mhz (approx) being generated by a signal generator (of
sorts) with output of -10dBm as measured *after* a 6dB attenuator pad.
So
-10dBm at the IF port on the mixer.
Spectrum Analyzer (Rigol DSA-815) connected to the RF port.
So we are testing the mixer as if we are in TX mode.
Test 1. KISS Mixer as per original design.
Output levels as seen by SA.
LO ~ -5dBm
IF ~ -50dBm
RF ~ -26dBm
So I make that not quite 20dB of LO suppression in the RF output.
Conversion loss (if my understanding is correct) is something way too
high
at about 16dB.
BTW the two J310 FET's are *very* well matched and selected for high
IDSS.
Observation.
Adjusting the bias on my mixer had very little effect on either the
desired RF output frequency or the LO level. Across the entire voltage
range of the bias only caused about 0.5dB variation in the desired RF
level. While the LO only varied by about 1dB. In the end because it
made
so little difference, I just set it at about mid point at 1.573v and
that's when I took the first SA picture called 73.bmp as attached.
Any ideas why my mixer shows so little variation to bias adjustment
anyone?
Test 2. KISS Mixer modified as per Joe Rocci W3JDR suggested alternate
BIAS arrangement. As per here:
//www.freelists.org/post/minima/Kiss-Mixer-Analysis
(Joe, I hope your still tuned in?).
Nothing in the test setup was altered.
Output levels as seen by SA.
(as per attached 74.bmp)
LO ~ -2.1dBm
IF ~ -39dBm
RF ~ -15.1dBm
LO suppression of only 16dB.
And the conversion loss is now only 5.1dB.
I think I like the look of the output from Joe's version better. As
the
desired RF signal is much higher and the conversion loss much lower.
But
both versions have a very high LO presence in the output.
So now what does it all mean?
Please, nobody assume that just because I happen to own a Spectrum
Analyzer that I know how to use it properly! :-) I'm just a Amateur
having
fun learning not an EE grad.
If anyone has any other ideas for further experiments let me know and
I'll
try my best.
73, Steve.
--
Regards
Sarma
Other related posts:
