EMRFD Message Archive 7038

Message Date From Subject
7038 2011-12-18 12:30:13 Chris Trask Mixer Musings and the KISS Mixer
I just added this to the group files section:

http://tinyurl.com/c6ufq69

It gives a thumbnail sketch of over a year of experimentation with diode mixers, concluding with the KISS mixer, which uses a single transformer, a single monolithic SPDT switch, and a few passive parts. It has about 3.9dB of conversion loss and approximately +40dBm OIP3 at HF frequencies. If you spend more than US$1 making one, then you've done something wrong. Terribly wrong.


Chris Trask
N7ZWY / WDX3HLB
Senior Member IEEE
http://www.home.earthlink.net/~christrask/
7039 2011-12-18 13:39:44 Edwin Carter Re: Mixer Musings and the KISS Mixer
7040 2011-12-18 15:41:23 Barry N1EU Re: Mixer Musings and the KISS Mixer
*try this:*
*

http://tinyurl.com/88ex235*

73, Barry N1EU

7041 2011-12-18 16:11:11 Edwin Carter Re: Mixer Musings and the KISS Mixer
Thanks it works.
ecarter@windstream.net / WA4YHL



7042 2011-12-18 16:42:30 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> I just added this to the group files section:
>
> http://tinyurl.com/c6ufq69
>
> It gives a thumbnail sketch of over a year of experimentation with
> diode mixers, concluding with the KISS mixer, which uses a single
> transformer, a single monolithic SPDT switch, and a few passive parts.
> It has about 3.9dB of conversion loss and approximately +40dBm OIP3 at
> HF frequencies. If you spend more than US$1 making one, then you've
> done something wrong. Terribly wrong.
>

<>

Okay, that TinyURL worked this morning but not now, as I found on
another list. Just go to the files section and look for "Mixer Musings Rev
0.pdf"

Chris Trask
N7ZWY / WDX3HLB
Senior Member IEEE
http://www.home.earthlink.net/~christrask/
7043 2011-12-18 19:14:19 cbayona.cb Re: Mixer Musings and the KISS Mixer
Doesn't work here with four different browsers.

At 06:11 PM 12/18/2011, you wrote:


>Thanks it works.
>ecarter@windstream.net / WA4YHL
>
>
>
>
7045 2011-12-19 02:27:37 Barry N1EU Re: Mixer Musings and the KISS Mixer
just point your browser to http://groups.yahoo.com/group/emrfd/files/ and
browse to the Mixer Musings file

Barry N1EU

7046 2011-12-19 03:41:36 Tim Hills Re: Mixer Musings and the KISS Mixer
Chris:

Good overview of mixer types and the KISS mixer looks like it would be
useful for space and budget tight applications.

Thanks

Tim Hills
Sioux Falls, SD

7047 2011-12-19 06:47:32 AD7ZU Re: Mixer Musings and the KISS Mixer
Chris,
 
I have been working on a variant of the TLT mixer over the past few weeks so your post is timely..
I was attempting to come up with a version that used a single switch and a single phase 1F clock (ok VFO for the analog thinkers).  My quandary was how to design a single core transformer topology that suppressed the LO, kept the insertion loss low, and kept the impedance flat across 3-30mhz or better 3-60Mhz and of course get a 40db IIP3....  I may try your example. 
 
I do not know if it is even possible to add 2 coupling windings to a 4:1 balun on a single core??
I have also been looking into using a much larger core with fewer windings to avoid saturation.
PA3AKE uses 4 mini circuits T1-6 in his measurements which is a lot of transformers for cheap QRP stuff.
 
I had earlier limited success using 2 spdt TI switches: one in each leg of a transformer for a quadrature to baseband down converter, but the bandwidth was limited to the upper HF (20m - 10m) with the transformer I wound... perhaps a T1-1 or T1-6 would work better?
 
any thoughts??
 
 
Randy
AD7ZU


________________________________
7048 2011-12-19 06:51:40 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> > It gives a thumbnail sketch of over a year of experimentation
> > with diode mixers, concluding with the KISS mixer, which uses
> > a single transformer, a single monolithic SPDT switch, and a
> > few passive parts. It has about 3.9dB of conversion loss and
> > approximately +40dBm OIP3 at HF frequencies. If you spend
> > more than US$1 making one, then you've done something wrong.
> > Terribly wrong.
>
>
> Good overview of mixer types and the KISS mixer looks like
> it would be useful for space and budget tight applications.
>

It's intended primarily for high dynamic range HF receiver front ends. For MW and lower frequencies you can use the FSA4157, which has lower series resistance. Now combine this with a push-pull pair of augmented lossles feedback amplifiers and you've got a really hefty front end.

The H-Mode people use four parts (three transformers and a quad switch). That Russian fellow uses three parts (two transformers and a quad switch). I use two parts (one transformer and an SPDT switch). And we're all getting about the same performance. I can also make a mixer with one part (saturable transformer), but it takes a lot of LO power and the IMD is awful.

He who designs wih the fewest parts wins. :{b

Chris
7049 2011-12-19 07:40:54 cbayona Re: Mixer Musings and the KISS Mixer
There are many FETs that are faster, was this one chosen for break
before make qualities? I did some test a couple of years ago and
breaking before making made a big difference in noise of the mixer at
20M and up.

At 07:10 AM 12/19/2011, Chris Trask wrote:
> >
> > > It gives a thumbnail sketch of over a year of experimentation
> > > with diode mixers, concluding with the KISS mixer, which uses
> > > a single transformer, a single monolithic SPDT switch, and a
> > > few passive parts. It has about 3.9dB of conversion loss and
> > > approximately +40dBm OIP3 at HF frequencies. If you spend
> > > more than US$1 making one, then you've done something wrong.
> > > Terribly wrong.
> >
> >
> > Good overview of mixer types and the KISS mixer looks like
> > it would be useful for space and budget tight applications.
> >
>
> It's intended primarily for high dynamic range HF receiver
> front ends. For MW and lower frequencies you can use the FSA4157,
> which has lower series resistance. Now combine this with a
> push-pull pair of augmented lossles feedback amplifiers and you've
> got a really hefty front end.
>
> The H-Mode people use four parts (three transformers and a
> quad switch). That Russian fellow uses three parts (two
> transformers and a quad switch). I use two parts (one transformer
> and an SPDT switch). And we're all getting about the same
> performance. I can also make a mixer with one part (saturable
> transformer), but it takes a lot of LO power and the IMD is awful.
>
> He who designs wih the fewest parts wins. :{b
>
>Chris
>

--
Cecil
k5nwa
http://www.softrockradio.org/
http://thepartsplace.k5nwa.com/
http://parts.softrockradio.org/

Never take life seriously. Nobody gets out alive anyway.
7050 2011-12-19 10:06:24 Gian Re: Mixer Musings and the KISS Mixer
Hi Chris,

Congratulations for your document.

If you look in Files "I7SWX 1T DOUBLE BALANCED FST3125 MIXER2 squarer.pdf" You will see the KISS mixer version using FST3125, I uploaded in Aug 2007 and it was published in RadCom, Technical Topics, June 2004 issue.
Also the mixer was used as a second mixer in G3XJP Pick-A-Star Transceiver.

73 and Seas
7053 2011-12-19 19:59:31 Chris Trask Re: Mixer Musings and the KISS Mixer

> I have been working on a variant of the TLT mixer over the past
> few weeks so your post is timely..
> I was attempting to come up with a version that used a single
> switch and a single phase 1F clock (ok VFO for the analog
> thinkers).  My quandary was how to design a single core
> transformer topology that suppressed the LO, kept the insertion
> loss low, and kept the impedance flat across 3-30mhz or better
> 3-60Mhz and of course get a 40db IIP3....  I may try your
> example. 
>

Using these monolithic SPDT switches takes care of the LO isolation problem. Looking at Fig. 9 shows you how the LO isolation is accomplished by conducting equal and opposite LO voltages across matched pairs of diodes. The quality of the isolation is dependent on the matching of the diodes and the balance of the transformer.


> I do not know if it is even possible to add 2 coupling windings
> to a 4:1 balun on a single core??
>

That could get messy. The best coupling (low insertion loss, wide bandwidth) is obtained by twisting two or three wires together. The transformer core then serves primarily to extend the low frequency performance. You can come up with quite a few usable configurations with that.

>
> I have also been looking into using a much larger core with
> fewer windings to avoid saturation.
>

I'm going to look into using larger cores after I've throughly tested the one that I have on the bench. Chances are that I'm pressing against the limits of the HP141T spectrum analyzer.

>
> PA3AKE uses 4 mini circuits T1-6 in his measurements which
> is a lot of transformers for cheap QRP stuff.
>

You should avoid using the Mini-Circuits transformers. The cores are far too small for good IMD performance at these signal levels. I resorted to using larger Fair-Rite -61 cores in my wideband loop antenna amplifier to get better IMD performance.


> I had earlier limited success using 2 spdt TI switches: one
> in each leg of a transformer for a quadrature to baseband
> down converter, but the bandwidth was limited to the upper
> HF (20m - 10m) with the transformer I wound... perhaps a
> T1-1 or T1-6 would work better?

> any thoughts??
>

I'm a firm advocate in rolling your own transformers as you can easily get far better performance than the Mini-Circuits products. They are convenient, but poorly suited for high IMD designs due to the small core size.
 
Chris
7054 2011-12-19 19:59:31 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> Congratulations for your document.
>
> If you look in Files "I7SWX 1T DOUBLE BALANCED FST3125 MIXER2 squarer.pdf"
You
> will see the KISS mixer version using FST3125, I uploaded in Aug 2007 and
it
> was published in RadCom, Technical Topics, June 2004 issue.
> Also the mixer was used as a second mixer in G3XJP Pick-A-Star
Transceiver.
>

Since I started passing this around last week I've been sent a number of
URLs and PDF documents showing similar attempts at simplifying the approach.
Cecil, K5NWA pointed out that the "break before make" nature of the FSA3157
results in noise problems above 20m. I'm wondering if using the FST3125/6
and applying some form of timing correction as I did in Fig. 11 and Fig. 12
could be accomplished?

Chris Trask
N7ZWY / WDX3HLB
Senior Member IEEE
http://www.home.earthlink.net/~christrask/
7055 2011-12-19 19:59:31 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> > The H-Mode people use four parts (three transformers and a
> > quad switch). That Russian fellow uses three parts (two
> > transformers and a quad switch). I use two parts (one transformer
> > and an SPDT switch). And we're all getting about the same
> > performance. I can also make a mixer with one part (saturable
> > transformer), but it takes a lot of LO power and the IMD is awful.
> >
> > He who designs wih the fewest parts wins. :{b
>
> There are many FETs that are faster, was this one chosen for break
> before make qualities? I did some test a couple of years ago and
> breaking before making made a big difference in noise of the mixer at
> 20M and up.
>

It was chosen to minimize the parts count while keeping the performance
at HF frequencies high.

Chris
7056 2011-12-19 20:00:16 Ashhar Farhan Re: Mixer Musings and the KISS Mixer
Chris,
Fantastic stuff! Why don't you post on your website too? The yahoo
groups urls 'float' and they are accessible only to group members who
are logged in. This means, it won't be indexed by the search engines.
- farhan

On 12/19/11, cbayona <CBayona@cbayona.com> wrote:
> There are many FETs that are faster, was this one chosen for break
> before make qualities? I did some test a couple of years ago and
> breaking before making made a big difference in noise of the mixer at
> 20M and up.
>
> At 07:10 AM 12/19/2011, Chris Trask wrote:
>> >
>> > > It gives a thumbnail sketch of over a year of experimentation
>> > > with diode mixers, concluding with the KISS mixer, which uses
>> > > a single transformer, a single monolithic SPDT switch, and a
>> > > few passive parts. It has about 3.9dB of conversion loss and
>> > > approximately +40dBm OIP3 at HF frequencies. If you spend
>> > > more than US$1 making one, then you've done something wrong.
>> > > Terribly wrong.
>> >
>> >
>> > Good overview of mixer types and the KISS mixer looks like
>> > it would be useful for space and budget tight applications.
>> >
>>
>> It's intended primarily for high dynamic range HF receiver
>> front ends. For MW and lower frequencies you can use the FSA4157,
>> which has lower series resistance. Now combine this with a
>> push-pull pair of augmented lossles feedback amplifiers and you've
>> got a really hefty front end.
>>
>> The H-Mode people use four parts (three transformers and a
>> quad switch). That Russian fellow uses three parts (two
>> transformers and a quad switch). I use two parts (one transformer
>> and an SPDT switch). And we're all getting about the same
>> performance. I can also make a mixer with one part (saturable
>> transformer), but it takes a lot of LO power and the IMD is awful.
>>
>> He who designs wih the fewest parts wins. :{b
>>
>>Chris
>>
>
> --
> Cecil
> k5nwa

>
> Never take life seriously. Nobody gets out alive anyway.
>
>

--
Sent from my mobile device
7057 2011-12-19 20:39:22 William Carver Re: Mixer Musings and the KISS Mixer
> You should avoid using the Mini-Circuits transformers. The cores are
> far too small for good IMD performance at these signal levels. I
> resorted to using larger Fair-Rite -61 cores in my wideband loop
> antenna amplifier to get better IMD performance.
,.............,
> I'm a firm advocate in rolling your own transformers as you can easily
> get far better performance than the Mini-Circuits products. They are
> convenient, but poorly suited for high IMD designs due to the small
> core size.
>
> Chris

WRONG. Given that the cores are tiny it is indeed a puzzle how MCL
transformers produce very high intercepts. But they do. They are NOT
"poorly suited for high IMD designs". Your firm advocacy is based on
what?

As I'm sure you (Chris) are aware, PA3AKE has constructed and performed
extensive measurements H-mode mixers constructed with seven different
MOSFET switches and 13 different transformers from 1.8 MHz to 50 MHz.
One of those transformers used FT37-43 cores, orders of magnitude bigger
than the MCL cores.

http://martein.home.xs4all.nl/pa3ake/hmode/index.html

The FSA3157 and MCL ADTT1-6 transformer intercept peaked at +48.7 dBm at
3.5 MHz, a low of +40.6 dBm on 12m, and still +39.2 dBm at 50 MHz!

The same switches, but homebrew FT37-43 cored homebrew transformers
peaked at +47.7 dBm, also on 3.5 MHz, but were below +40 dBm at 18 MHz
and higher.

I can picture a response saying that the homebrew transformers were not
optimized, or there was some error in the measurements. Assertions of
that kind are easy to make, but unjustified. Martein has spent years,
etching hundreds of boards, and made his measurements using a close to
state-of-the-art Rohde and Schwartz spectrum analyzer which he purchased
specifically for these measurements.

Bill - W7AAZ
7058 2011-12-20 02:34:06 ha5rxz Re: Mixer Musings and the KISS Mixer
All agreed here, the transformers from Minicircuits are very good and I can't fault them, both in commercial and amateur service.

Peter HA5RXZ

7059 2011-12-20 05:03:53 Chris Trask Re: Mixer Musings and the KISS Mixer
>
>Chris,
>Fantastic stuff! Why don't you post on your website too? The yahoo
>groups urls 'float' and they are accessible only to group members who
>are logged in. This means, it won't be indexed by the search engines.
>

I'm going to see if I can make some room on my web page as this hot linking to Yahoo! group files is not suitable, plus the search engines such as Google can't access them.

Chris
7060 2011-12-20 05:10:48 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> > You should avoid using the Mini-Circuits transformers. The
> > cores are far too small for good IMD performance at these
> > signal levels. I resorted to using larger Fair-Rite -61
> > cores in my wideband loop antenna amplifier to get better
> > IMD performance.
> > ,.............,
> > I'm a firm advocate in rolling your own transformers as you
> > can easily get far better performance than the Mini-Circuits
> > products. They are convenient, but poorly suited for high
> > IMD designs due to the small core size.
> >
>
>
> WRONG. Given that the cores are tiny it is indeed a puzzle how
> MCL transformers produce very high intercepts. But they do.
> They are NOT "poorly suited for high IMD designs". Your firm
> advocacy is based on what?
>

A considerable amount of first-hand experience. I've had more than one design where using MCL transformers had lower IMD performance than using homebrewed ones. I get lower insertion loss and much better bandwidth as I am not limited to the small size that their transfromers are confined to. At the same time, I can construct mine so that they can use the exact same PCB footprint, so my PCB designs will accomodate both.

Chris
7061 2011-12-20 05:37:51 AD7ZU Re: Mixer Musings and the KISS Mixer
just curious..
does Minicircuits use a different core mix (a secret recipe) in their parts than is generally available?
I have dissected one and didn't see anything that I wouldn't be able to build.
or have they just made careful measurements and optimized their designs?
 
The TLT topology is appealing because it requires only a 1F single phase low power LO, though improved symmetry might be possible using a 2x clock such that switching always occurs on the same clock edge, taking the LO duty cycle out of the equation.  The QEX article on TLT mixers did note some issues with unterminated windings, however the remedy added significant complexity... and did not show results which differed significantly from the simpler implementation as built and measured by PA3AKE.  I am also curious to know if the slight degradation noted by PA3AKE is due to switching asymmetry or something else at higher frequencies?   This is all just curiosity on my part.  For QRP projects any mixer that requires 1F single phase LO, has conversion loss < 6db, IP3 > 30db from 3mhz - 30mhz, and >50db IF rejection is probably good enough... if it could be built on a single core that would be icing
7062 2011-12-20 07:01:49 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> just curious..
> does Minicircuits use a different core mix (a secret recipe)
> in their parts than is generally available?
> I have dissected one and didn't see anything that I wouldn't
> be able to build.
> or have they just made careful measurements and optimized
> their designs?
>

Really can't say. Looking at their really tiny unencapsulated transformers I don't see anything out of the ordinary there either. For that footprint I use the 2843002302 or 2861002302 core and #36 bifilar or trifilar wire.


> The TLT topology is appealing because it requires only a 1F
> single phase low power LO, though improved symmetry might be
> possible using a 2x clock such that switching always occurs
> on the same clock edge, taking the LO duty cycle out of the
> equation. 
>

Then you could use the FST3306, which has switching times that are all but symmetrical.

>
> The QEX article on TLT mixers did note some issues with
> unterminated windings, however the remedy added significant
> complexity... and did not show results which differed
> significantly from the simpler implementation as built and
> measured by PA3AKE. 
>

I don't think the unterminated windings is a serious issue. It already exists with diode ring and FET ring mixers. If you look at my split-ring mixer functional diagram (Fig. 14) you'll see that what takes place is that the transformer goes from 1:1 inverting to 1:1 noninverting. Terminating the open winding would probably ruin the conversion loss as half the power would be dissipated in the termination, plus it would ruin the return losses.

>
> I am also curious to know if the slight degradation noted
> by PA3AKE is due to switching asymmetry or something else
> at higher frequencies?  
>

Cecil, K5NWA mentioned earlier that there is a "dead zone" with the FSA3157 that may be the cause of increased noise above 20m. I'm going to do the same timing test as I did for the uncompensated split-ring mixer (Fig. 9 and Fig. 10) to see how serious it is. I'll include that test with the next revision.

BTW: The paper has the same photos for Fig. 10 and Fig. 13, so you cannot see the correction due to the added pair of diodes.

>
> This is all just curiosity on my part.  For QRP projects
> any mixer that requires 1F single phase LO, has conversion
> loss < 6db, IP3 > 30db from 3mhz - 30mhz, and >50db IF
> rejection is probably good enough... if it could be built
> on a single core that would be icing on the cake.
>

There is a serious goal of achieving +50dBm OIP3, but so far people have only been able to achieve slightly better than +40dBm on a regular basis. The switching devices are the key issue, with transformer core nonlinearities being second. Changing the topology (ring, H-Mode, split-ring, etc.) has no effect. I've looked at using a simplified feedforward technique, but it has problems of it's own.

Something will come to mind sooner or later.

Chris
7063 2011-12-20 13:35:06 William Carver Re: Mixer Musings and the KISS Mixer
On Tue, 2011-12-20 at 05:37 -0800, AD7ZU wrote:
>
> just curious..
> does Minicircuits use a different core mix (a secret recipe) in their
> parts than is generally available? I have dissected one and didn't see
> anything that I wouldn't be able to build. or have they just made
> careful measurements and optimized their designs?

All good questions. Inspecting the "open" SMD transformers doesn't show
me anything: they actually look kind of clumsily wound in some cases.

Although their volume must be very high, I wouldn't think they are
making their own cores. But their business depends strictly on
performance and it appears they have left no stone unturned to get it.
Whether the sinter their own cores, or are simply in bed with
proprietary mixes from a core company I don't know, but they're doing
something right.

> The TLT topology is appealing because it requires only a 1F single
> phase low power LO, though improved symmetry might be possible using a
> 2x clock such that switching always occurs on the same clock edge,
> taking the LO duty cycle out of the equation. The QEX article on TLT
> mixers did note some issues with unterminated windings, however the
> remedy added significant complexity... and did not show results which
> differed significantly from the simpler implementation as built and
> measured by PA3AKE. I am also curious to know if the
> slight degradation noted by PA3AKE is due to switching asymmetry or
> something else at higher frequencies? This is all just curiosity on
> my part. For QRP projects any mixer that requires 1F single phase LO,
> has conversion loss < 6db, IP3 > 30db from 3mhz - 30mhz, and >50db IF
> rejection is probably good enough... if it could be built on a single
> core that would be icing on the cake.

I share your curiosity. It is a complex puzzle indeed.

A fixed time asymmetry in the switch will become an increasing phase
angle with increasing frequency. So for mixer application a switch
designed to be symmetrical is desirable. The FSA3157 has a CMOS (Vcc/2)
threshold instead of TTL (1.4V), and I've measured some and found the
threshold was within a few percent of Vcc/2.

When driven by logic having equal/symmetrical rise and fall times the
switch input seems as close to perfect as one could imagine. Martein,
Chris and I are in agreement that the FSA3157 is 'da switch.

DC balance at the LO input doesn't mean the switch itself is perfectly
symmetrical so Martein's mixers all have DACs to allow per band
adjustment if you want. Personally (1) that somehow feels like cheating
and (2) IP3 is now so high that intermod isn't much of an issue, so I
bias the switch input at Vcc/2 with a single trimpot for "tweeking".

When the H-mode was first published few people could measure high
intercepts so Colin suggested nulling RF-IF feedthrough to get close to
the best intercept. To a first approximation that's OK. Better than
nothing, especially when he was driving the gates of individual MOSFETS.

But Martein, leaving no stone unturned, measured RF-IF and LO-IF
isolation as well as intercept, and found that their best numbers did
not coincide at one bias point. And he found some MCL transformers have
better isolation that others.

This additional data suggests to me that we are at a point where
asymmetries are small, and one can compensate for slight transformer
asymmetry and/or slight capacitive imbalance with a slight shift in
switch symmetry.

LO-IF isolation is serious to me because without isolation DDS LO spurs
feed right into the IF strip. Especially with older DDS chips the spurs
were, to my mind, more of a problem than IP3 because there were so
damned many spurs! Yet if much switch asymmetry is invoked to produce
best LO-IF isolation the mixer will produce a signal at the IMD
frequencies (2*f1-f2 for example).

So what you want is not to be compensating asymmetries, but you want
everything to be as symmetrical as possible to start with. I look at
Martein's monumental work and picked the transformer that has the best
combination of intercept and LO-IF isolation. ADDT1-1 produces 40 dBm
IP3 even at 6m (OK, 39.7!) and far better LO isolation than other
transformers. The price paid with this 1:1 transformer is more switch
loss, so the conversion loss is 0.4 dB higher than with 1:2
transformers. That was a no-brainer concession.

I feel like a mountain climber who has reached the summit of Mount
Everest by following in the footsteps of a very talented and hardworking
guide. There's still an aura of mystery about the remaining IMD and
feedthrough, but from a practical standpoint the receiver front end is
near-perfect.

Bill- W7AAZ
7066 2011-12-20 14:02:58 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> >
> > > Good overview of mixer types and the KISS mixer looks like
> > > it would be useful for space and budget tight applications.
> > >
> >
> > It's intended primarily for high dynamic range HF receiver
> > front ends. For MW and lower frequencies you can use the FSA4157,
> > which has lower series resistance. Now combine this with a
> > push-pull pair of augmented lossles feedback amplifiers and you've
> > got a really hefty front end.
> >
> > The H-Mode people use four parts (three transformers and a
> > quad switch). That Russian fellow uses three parts (two
> > transformers and a quad switch). I use two parts (one transformer
> > and an SPDT switch). And we're all getting about the same
> > performance. I can also make a mixer with one part (saturable
> > transformer), but it takes a lot of LO power and the IMD is awful.
> >
> > He who designs wih the fewest parts wins. :{b
>
> There are many FETs that are faster, was this one chosen for break
> before make qualities? I did some test a couple of years ago and
> breaking before making made a big difference in noise of the mixer at
> 20M and up.
>

I just tested the ON/OFF time discrepancy of the FSA3157 and found that
the break-before-make time is approximately 18-21nSec. This was while using
the LO pulse shaper circuit so as to get the rise and fall times as fast as
practical.

Chris
7067 2011-12-20 14:40:24 William Carver Re: Mixer Musings and the KISS Mixer
> I just tested the ON/OFF time discrepancy of the FSA3157 and found
> that
> the break-before-make time is approximately 18-21nSec. This was while
> using
> the LO pulse shaper circuit so as to get the rise and fall times as
> fast as
> practical.

Does the make-before-break subtract that same abount on both edges? That
is, do you get two 18-20 nS narrower closures, but of equal length? Or
does the M/S ratio change?

Make before break means all the switches are off. I'd expect an increase
in conversion loss from that, but not "noise". What say ye?

W7AAZ
7068 2011-12-20 15:07:17 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> > I just tested the ON/OFF time discrepancy of the FSA3157 and found
> > that the break-before-make time is approximately 18-21nSec. This
> > was while using the LO pulse shaper circuit so as to get the rise
> > and fall times as fast as practical.
>
> Does the make-before-break subtract that same abount on both edges? That
> is, do you get two 18-20 nS narrower closures, but of equal length? Or
> does the M/S ratio change?
>
> Make before break means all the switches are off. I'd expect an increase
> in conversion loss from that, but not "noise". What say ye?
>

Same for both edges. The datasheet says that it should only be 0.5nSec,
but that's for a 50 ohm source and load with a shunt capacitance. This has
a series inductance from the transformer leakage inductance.

I'm going to do some conversion loss tests tomorrow.

Chris
7069 2011-12-20 16:22:47 William Carver Re: Mixer Musings and the KISS Mixer
On Tue, 2011-12-20 at 16:06 -0700, Chris Trask wrote:
> >
> > > I just tested the ON/OFF time discrepancy of the FSA3157 and found
> > > that the break-before-make time is approximately 18-21nSec. This
> > > was while using the LO pulse shaper circuit so as to get the rise
> > > and fall times as fast as practical.

> Same for both edges. The datasheet says that it should only be 0.5nSec,
> but that's for a 50 ohm source and load with a shunt capacitance. This has
> a series inductance from the transformer leakage inductance.

Now THAT is curious: series inductance increases make-before-break? A
current spike coming OUT of the switch pin when the internal "break"
occurs, extending the time to the "make" threshold?

Wow, I wouldn't even be wary enough to spot that. I would think they'd
have a note about that in the data sheet.
7070 2011-12-20 18:02:16 cbayona Re: Mixer Musings and the KISS Mixer
In a conventional transformer the core uses the
magnetic flux to couple the signal and must be
large, if it's a transmission line transformer
then the core is not used to couple the power and the core can be much smaller.

At 07:37 AM 12/20/2011, you wrote:
>just curious..
>does Minicircuits use a different core mix (a
>secret recipe) in their parts than is generally available?
>I have dissected one and didn't see anything
>that I wouldn't be able to build.
>or have they just made careful measurements and optimized their designs?

>The TLTÂ topology is appealing because it
>requires only a 1F single phase low power LO,
>though improved symmetry might be possible using
>a 2x clock such that switching always occurs on
>the same clock edge, taking the LO duty cycle
>out of the equation. The QEX article on TLT
>mixers did note some issues with unterminated
>windings, however the remedy added significant
>complexity... and did not show results which
>differed significantly from the simpler
>implementation as built and measured by
>PA3AKE. I am also curious to know if the
>slight degradation noted by PA3AKE is due to
>switching asymmetry or something else at higher
>frequencies?  This is all just curiosity on
>my part. For QRP projects any mixer that
>requires 1F single phase LO, has conversion loss
>< 6db, IP3 > 30db from 3mhz - 30mhz, and >50dbÂ
>IF rejection is probably good enough... if it
>could be built on a single core that would be icing
7071 2011-12-20 18:02:44 cbayona Re: Mixer Musings and the KISS Mixer
I should have been more specific, I compared
regular make before break FST3253, FST3125
switches with the same switches with adjusted
timing. These switches are very commonly used in
SDR radios as the mixer and suffer poor
performance at the higher frequencies. They have
asymmetrical switching characteristics where they
turn on twice as fast as they turn off.The
adjusted timing was set to where it broke the
connection just slightly before the opposite
switch turned on, that had much lower noise
compared to not having timing adjustments which
had brief periods where opposite switches were both on.

It was not a very scientific test but good enough
for what I wanted, the input was connected to a
dummy load and I measure the output noise of the
switch with and without adjustments to the
timing. I did not test to see if a larger dead
zone made it better or worst, I was a test to
compared if the opposite switches were both
turned on versus having a tiny dead period so the
switches never were turned on at the same time.

The period to prevent on overlap was about .5ns
where neither switch were on, the test frequency
was the 10M band, having the switches not overlap
made a big difference, the hardware to change the
timing cost less than a dollar.

I would be interested to know what happens where
the dead period get to be much larger.


> Cecil, K5NWA mentioned earlier that there
> is a "dead zone" with the FSA3157 that may be
> the cause of increased noise above 20m. I'm
> going to do the same timing test as I did for
> the uncompensated split-ring mixer (Fig. 9 and
> Fig. 10) to see how serious it is. I'll
> include that test with the next revision.
>
> BTW: The paper has the same photos for
> Fig. 10 and Fig. 13, so you cannot see the
> correction due to the added pair of diodes.
>
> >
> > This is all just curiosity on my part. For QRP projects
> > any mixer that requires 1F single phase LO, has conversion
> > loss < 6db, IP3 > 30db from 3mhz - 30mhz, and >50db IF
> > rejection is probably good enough... if it could be built
> > on a single core that would be icing on the cake.
> >
>
> There is a serious goal of achieving
> +50dBm OIP3, but so far people have only been
> able to achieve slightly better than +40dBm on
> a regular basis. The switching devices are the
> key issue, with transformer core nonlinearities
> being second. Changing the topology (ring,
> H-Mode, split-ring, etc.) has no effect. I've
> looked at using a simplified feedforward
> technique, but it has problems of it's own.
>
> Something will come to mind sooner or later.
>
>Chris
>
>
>
>------------------------------------
>
>Yahoo! Groups Links
>
>
>
--
Cecil
k5nwa
7072 2011-12-20 19:29:04 kb1gmx Re: Mixer Musings and the KISS Mixer
7074 2011-12-20 20:32:14 William Carver Re: Mixer Musings and the KISS Mixer
Martein shows intercept, conversion loss, RF-IF and LO-IF isolation on
HF for all combinations of transformers and switches. And in some cases
he took data for 6m and 4m (70 MHz European band).

Yes, performance degrades. It takes time for a switch to operate and at
50 MHz the time for a half cycle is only 10 nS. But the FSA3157 DOES
operate at 50 and 70 MHz, and although the numbers aren't as high as 40
meters they are still respectable.

For example, FSA3157 with MCL ADT1-6T transformers the 6m input
intercept is +42 dBm, conversion loss 5.52 dB. The performance factors
that suffer and RF-IF and LO-IF isolation.

W7AAZ


7075 2011-12-20 21:01:45 AD7ZU Re: Mixer Musings and the KISS Mixer
 
Allison may be on to something. I know the data sheets for the switches always specify a very low capacitance load as part of the timing specs and found there is no mention of inductive loads or sources .. after all these things are made to switch high speed digital signals .. which to my knowledge never use inductors and as a matter of design go to great lengths to eliminate any inductance with leadless SMT and tight PCB layouts.  Then I ran across this:  http://www.potatosemi.com/%c2%a0 and interesting discussion under the "High Frequency Noise Cancellation Technology" selection that confirms some of what Allison noted.
 
50db of opposite sideband rejection requires an I/Q demodulator to have < 1degree of phase error if everything else is perfect.  So..  at 10m each of 4 switches operates over 180 degrees of a 120mhz clock = 4.15ns period. To maintain < 1 degree of phase error at that frequency requires extremely tight symmetry like .02ns. Then there is the amplitude balance of 0.1db  for 50db of opposite sideband rejection hence the crosstalk between switches must also be very low. That's probably why there is so much discussion about quadrature error correction in the SDR circles these days... and also why Analog Devices now "features" quadrature error correction in some ADCs... which tells us that Analog Devices cannot get the phase and amplitude errors out of their own monolithic I/Q modulator / demodulator (probably fet switched) products...there I feel better now.
 
The initial reas
7076 2011-12-20 23:25:00 William Carver Re: Mixer Musings and the KISS Mixer
> The initial reason I became interested in mixer designs was that I
> have not been able to get the phase and amplitude of I/Q sampling
> circuits to track over any sizable frequency range. .. then it struck
> me why not just downconvert using a mixer and avoid having near UHF
> quadrature VFO signals and dial in an I/Q demodulator at a single
> low IF ..like 5mhz or so...just like the superhet guys have been doing
> for many decades.. just think I could eliminate the I/Q demod and
> DSP altogether if I added an IF amp and a crystal filter..
>
>
> Randy


Your heterodyne idea sounds Super. Unfortunately good quartz costs more
than silicon, which is a significant part of the push behind DSP.
7077 2011-12-21 05:11:39 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> This has been my base problem with most all the mixers using
> the various CMOS switches. Most all fail badly by time you
> get to 6M. That suggests the dynamics of the switching is
> still not fast enough.
>
> It would be interesting to look at and apply leaded or better
> SMT MOSFETs or JFETs as they often are not encumbered with
> the monolithinc process constraints.
>

I've tried various FETs and bipolars in an effort to make better mixers, and the results were less than satisfactory. I then returned to experimenting with diodes to see if there was a topology that would work better than a ring mixer, and the results were that no matter what the topology the performance for a given diode remained pretty much the same.

Noticing the basic function of the split-ring mixer led to using the FSA3157, and now I want to see what the mimitations are. I know a priori that it will be limited to HF frequencies or at best 6m. The simplicity of just using two parts is attractive.

Chris
7078 2011-12-21 05:16:47 Chris Trask Re: Mixer Musings and the KISS Mixer
>
> Martein shows intercept, conversion loss, RF-IF and LO-IF
> isolation on HF for all combinations of transformers and
> switches. And in some cases he took data for 6m and 4m
> (70 MHz European band).
>
> Yes, performance degrades. It takes time for a switch to
> operate and at 50 MHz the time for a half cycle is only
> 10 nS. But the FSA3157 DOES operate at 50 and 70 MHz, and
> although the numbers aren't as high as 40 meters they are
> still respectable.
>
> For example, FSA3157 with MCL ADT1-6T transformers the 6m
> input intercept is +42 dBm, conversion loss 5.52 dB. The
> performance factors that suffer and RF-IF and LO-IF
> isolation.
>

Yes, I've seen his performance data listings. He's very through and provides a good baseline for comparing other forms.

Chris
7079 2011-12-21 07:31:35 kb1gmx Re: Mixer Musings and the KISS Mixer
7080 2011-12-21 07:39:47 kb1gmx Re: Mixer Musings and the KISS Mixer
7081 2011-12-21 07:47:44 Chris Trask Re: Mixer Musings and the KISS Mixer
I've been pounding the internet highway pavement this morning looking for other switches to try. Not much out there, but Pericom has an SPST NO (PI5A121) and SPST NC (PI5A122) switch in the SOT-23-5 package that may be worth looking into. With these, the ON/OFF overlap can be controlled and you still have the single LO input. Neither Digi-Key nor Mouser have them in stock, so I've requested samples of all three.

They also have an SPDT switch (PI5A124) in the SOT-23-6 package, which I've also requested sample of.

Chris
7084 2011-12-21 15:25:20 AD7ZU Re: Mixer Musings and the KISS Mixer
I must say all the discussion on mixer topology has been an enlightening read. seems before long every possible wrinkle will be tried, measured, tested, analyzed, critiqued, cussed and discussed, re-engineered, optimized, and shared on EMRFD...I learn new stuff every day here!
 
might also have a look at http://www.potatosemi.com/%c2%a0 I am going to order a few of their spdt and 4:1 variations along with some FSA3157s.
 
The Potato chip technology suppresses switching noise, operates at 3.3v, and the on/off symmetry spec is better than others I have seen.  the downside is that the max ron = 15ohms  at 3.3v. 
 
Then build a similar mixer with a single transformer and single spdt switch. ..if that works then do a double balanced version.  I have a superhet QRP 20m kit I use for experiments.
 
 
Randy
AD7ZU 


________________________________
7085 2011-12-22 04:43:58 Russell Shaw Re: Mixer Musings and the KISS Mixer
7087 2011-12-22 05:55:35 Chris Trask Re: Mixer Musings and the KISS Mixer