EMRFD Message Archive 3635
Message Date From Subject 3635 2009-10-23 08:56:07 Colin Mixers
Let's see if I have this right. *smile*
A mathematically pure mixer would be an accurate signal multiplier.
That makes sense to me, and so does the associated trig identity of
multiplication.
A few questions come to mind though:
1. Why do so many modern mixer designs use switching designs? It isn't
mathematically perfect and must produce a stack of spurs - the local
oscillator is a square wave wave with a series of odd order harmonics.
The specan pictures I've seen of switching mixer outputs look a mess and
not really much different from an old style single transistor acting as
a mixer, although switching mixers (diode or MOSFET) are usually
balanced and have a good measure of rf-in and local oscillator
cancellation. An Analog Devices data sheet for the AD8343 says that
accurate multipliers tend to be noisy, so switching is the next best
option to permit high IP3 and get low noise.
2. Is the following correct then?
Modern switching FET mixers and balanced diode bridges, aren't really
mixers in the full mathematical multiplication sense, but a compromise
driven by the poor noise of true multipliers and the fact that we have
good enough filters and diplexers etc to deal with all the unwanted
spurs that are generated by them?
3. So, are mathematically accurate multipliers truly that dismal?
4. AD have a couple of multipliers that go well into VHF, has anyone
tested these?
5. If true multipliers are too noisy to be used as receiver mixers, then
the fundamental engineering problem is that the natural occurring
dynamic range of radio signals as received by typical radios is huge.
Does that make sense? I certainly don't doubt Analog Devices
applications advice, but I want to know if my conclusions are sensible
or misguided.
Cheers, Colin VK3KDF3636 2009-10-23 10:14:52 Niels A. Moseley Re: Mixers
Switching mixers are relatively easy to design and offer excellent
linearity, which explains their popularity. In addition, implementing
highly linear wideband analog circuits (such as accurate multipliers) in
modern standard CMOS IC processes is a challenge because of the reduced
voltage headroom and the lower intrinsic gain of the transistors
compared to older process generations.
A balanced switching mixer can be viewed as a perfect multiplier which
is driven by a bi-polar square wave. If you drive a diode ring mixer
with a square wave LO, you effectively have a switching mixer. The same
is true for the NE602 Gilbert cell mixer, which implements an "accurate"
multiplier.
The square wave LO has many more harmonics than a sinusoidal wave. Each
harmonic will down-mix a different RF frequency. The harmonics of a 50%
duty-cycle square wave only has odd harmonics (1,3,5,7 etc..). Thus, RF
signals found at LO, 3LO, 5LO etc. are down-mixed to around DC and will
cause interference. These additional products are termed "harmonic images".
The same holds for up-mixing. Your signals around DC are mixed to LO,
3LO, 5LO, etc.
While there are many harmonic products, they all end up in predictable
places and we can easily deal with them by filtering. The conversion
loss (relative to the fundamental product) is 1/n, where 'n' is the
harmonic.
The first harmonic image is related to the 3rd LO harmonic (there are no
even harmonics) and is supressed by just -20*log(1/3) = 9.5 dB.
Additional RF filtering is used to give more suppression.
Incidentally, the amount of filtering can be reduced by using a parallel
structure of switching mixers, where each mixer is driven by a different
pulse-like waveform. The outputs of the mixers are individually weighted
and summed together. The whole point of this is to have an _effective_
LO waveform that is a staircase approximation of a sinusoid. This has
less harmonics than the single mixer case and thus has reduced harmonic
image energy. In practice, this gives around 40+ dB of suppression
(using 8 switches in 32nm CMOS) of the 3rd harmonic image.
The 40 dB limit is caused by amplitude and phase mismatches within the
analog circuitry. Without using RF filtering, I have been able to push
this figure to over 80 dB, using adaptive signal processing techniques.
See:
http://eprints.eemcs.utwente.nl/15232/01/Moseley_A_400-to-900_MHz_Receiver_with_Dual_domain.pdf
Back to the coal mines..
73,
Niels PA1DSP.
> Hi all,
>
> Let's see if I have this right. *smile*
>
> A mathematically pure mixer would be an accurate signal multiplier.
> That makes sense to me, and so does the associated trig identity of
> multiplication.
>
> A few questions come to mind though:
>
> 1. Why do so many modern mixer designs use switching designs? It isn't
> mathematically perfect and must produce a stack of spurs - the local
> oscillator is a square wave wave with a series of odd order harmonics.
> The specan pictures I've seen of switching mixer outputs look a mess and
> not really much different from an old style single transistor acting as
> a mixer, although switching mixers (diode or MOSFET) are usually
> balanced and have a good measure of rf-in and local oscillator
> cancellation. An Analog Devices data sheet for the AD8343 says that
> accurate multipliers tend to be noisy, so switching is the next best
> option to permit high IP3 and get low noise.
>
> 2. Is the following correct then?
>
> Modern switching FET mixers and balanced diode bridges, aren't really
> mixers in the full mathematical multiplication sense, but a compromise
> driven by the poor noise of true multipliers and the fact that we have
> good enough filters and diplexers etc to deal with all the unwanted
> spurs that are generated by them?
>
> 3. So, are mathematically accurate multipliers truly that dismal?
>
> 4. AD have a couple of multipliers that go well into VHF, has anyone
> tested these?
>
> 5. If true multipliers are too noisy to be used as receiver mixers, then
> the fundamental engineering problem is that the natural occurring
> dynamic range of radio signals as received by typical radios is huge.
> Does that make sense? I certainly don't doubt Analog Devices
> applications advice, but I want to know if my conclusions are sensible
> or misguided.
>
> Cheers, Colin VK3KDF
>
>
>
>
> ------------------------------------
>
> Yahoo! Groups Links
>
>
>
>3637 2009-10-23 11:10:58 Chris Trask Re: Mixers
>3643 2009-10-24 05:26:24 Colin Re: Mixers
Thanks for your reply. You said "switching mixers" and "excellent
linearity" in the one sentence. My confusion is multiplied. *laughs*
How can anything that cuts the top off a waveform be considered linear?
I suspect the rubbery nature of english used to describe technical
matters is a problem when used with mixer complexities.
Just as an example to make talking about this easier, can I propose a
four diode, high level balanced modulator, with input RF of 10MHz, local
oscillator square wave driven of 14MHz and a wanted difference IF of
4MHz:
If you mean that the wanted 4MHz output has a peak to peak rf voltage
that is a simple ratio of the input 10MHz peak to peak rf voltage unless
it exceeds some input overload level, then there is a form of linearity
between the wanted IF out and the RF input.
That doesn't mean there is anything analogue or linear involved in the
local oscillator operation.
Does what I have said above make sense to you? I can grasp the basic
analogue operation of a fet mixer with square wave drive - it is the
same exactly as a fast digital sampling scope. It gives frequency
translation but maintains the linearity of the peak waveform voltage and
so retains any amplitude modulation.
Local oscillator harmonic content makes sense to me although your
information on the harmonic levels is new to me and appreciated. Your
description of parallel mixers with complex different LO waveforms is
also a new to me and a fascinating idea. At first glance it looks like
an extension or development of the Tayloe circuit, or is it another
thing altogether?
Sorry to sound pedantic, but you said: "A balanced switching mixer can
be viewed as a perfect multiplier which is driven by a bi-polar square
wave." If it is a perfect multiplier, then it is an analogue device by
nature - no? Or is it a switching device? I can't see how it can be
anything but one or the other. Don't ask me to visualise photons...
*laughs* Or do I have a conceptual problem here?
I've read descriptions about the Gilbert cell being an analgoue
multiplier, but lots of articles also talk about the upper four
transistors operating in saturation as switches. It looks to me as
though the Gilbert cell can be operated as a analogue multiplier and
also as a switching mixer depending on local oscillator drive level.
Is that what is muddying the waters of language? (More like quicksand
of language. *laughs*) That the Gilbert cell can be a true analogue
multiplier and a switching mixer, primary difference being that the
local oscillator is either low enough for totally linear operation, or
else so large that the upper transistors are in saturation?
Best wishes, Colin VK3KDF
Melbourne, Australia3644 2009-10-24 06:40:12 Colin Re: Mixers
Thanks for your reply on this fascinating subject. I'm currently trying
to construct my recent reading into a coherent model, and it hasn't
fully gelled yet for me yet. I try to understand how things work
despite my mathematical limitations. Persistence helps. *grin* One
problem is that english is imprecise for technical work, also experts in
various subjects get used to their own dialects and that makes accurate
translation tricky. I was involved as a technical writer and trainer
for a large digital video project, and had lots of fun translating
engineering computing speak into technician speak. Language is a tricky
problem, plain language isn't simple language. On to the subject.
Much of what you wrote makes sense, and thank you for taking the effort
to write about your experiences. I really appreciate it.
You said: "A balanced switching mixer can be viewed as a perfect
multiplier which is driven by a bi-polar square wave. If you drive a
diode ring mixer with a square wave LO, you effectively have a switching
mixer. The same is true for the NE602 Gilbert cell mixer, which
implements an "accurate" multiplier."
OK, so this confirms what I had asked another writer, Niels earlier
tonight. Gilbert cells can be either analogue multipliers or switching
mixers depending on the local oscillator drive level.
The design outcome for a receiver is likely different for an instrument
or signal generator than a receiver. I imagine that fully linear
analogue multiplication is preferred for instrument applications,
because they are generating signals, noise would be far better
controlled and mixer output spectrum needs to be as low in spurious
signals as possible.
I've only just today read Gilbert's paper on his translinear input
circuit, it will take a few more readings before I fully grasp it. I've
read your net published papers on lossless feedback around the Gilbert
cell mixer and found them fairly easy to grasp. Feedback makes sense to
me, perhaps because I'm just not very mathematical in my way of
understanding things and you've written clearly.
Your comments about reducing active devices is good sense, although
these days, I suspect it would strike many folk as counter-intuitive.
The shortest path between two truths in the real domain passes through
the complex domain. - Jacques Hadamard
Your comments about reducing resistive losses make good sense also.
I've read your paper on lossless feedback with augmentation, and agree
that it makes a mixer more difficult to build. Have you found stability
any trickier with these methods? The use of negative feedback would
suggest better stability, but I've seen some weird things in my time.
I've seen the MC1496 used as product detectors and SSB generators, but
as you say, it is rather frequency limited. What about the AD835 then?
According to the spec sheet it is an accurate (0.1%) 4 quadrant
multiplier and while not cheap, it isn't terribly expensive either.
Typical power consumption is 5V at 16mA. Again, not brilliant, but
worth paying if the output filtering requirements are reduced. Has
anyone attempted to use this part as a receiver mixer? It should be
good up to low VHF.
Best wishes, Colin VK3KDF
Melbourne, Australia3646 2009-10-24 06:52:54 Tim Re: Mixers
> be viewed as a perfect multiplier which is driven by a bi-polar square
> wave." If it is a perfect multiplier, then it is an analogue device by
> nature - no? Or is it a switching device? I can't see how it can be
> anything but one or the other. Don't ask me to visualise photons...
> *laughs* Or do I have a conceptual problem here?
"Linear mixer" is the root of the problem. Multiplication is not a linear process. We often get best performance from our radios by operating one of the two channels in a way that is quite nonlinear.
Look at a receiver mixer with signal in, local oscillator in, and IF out.
When we double the signal in, we expect twice as much IF out. When we put half the signal in, we expect half as much IF out. That's the "linearity" we expect. If the mixer is operating in a region where this linearity isn't good we get intermod and distortion.
At the same time, we usually choose an operating point where the output of the mixer is insensitive to variations in the local oscillator level. Indeed, if you take a +7dbm diode ring mixer, and instead trying running at +4dbm or +10dbm, you see very little change in the IF out level despite the fact that you just varied one of the mixer inputs by a lot.
We choose to run mixers in this way because we find that with most real-world mixer designs, you get the best linearity in the input where you want linearity, by running the other all the way into saturation.
> I've read descriptions about the Gilbert cell being an analgoueEven if you try very hard to make and operate a Gilbert cell multiplier in a region where it is linear in both input channels, you discover that by design one of the input channels (the "upper four transistors") is inherently less linear and the other is inherently more linear.
> multiplier, but lots of articles also talk about the upper four
> transistors operating in saturation as switches. It looks to me as
> though the Gilbert cell can be operated as a analogue multiplier and
> also as a switching mixer depending on local oscillator drive level.
>
> Is that what is muddying the waters of language? (More like quicksand
> of language. *laughs*) That the Gilbert cell can be a true analogue
> multiplier and a switching mixer, primary difference being that the
> local oscillator is either low enough for totally linear operation, or
> else so large that the upper transistors are in saturation?
The channel most susceptible to crossover distortion is the channel that you want to saturate in mixer use.
Tim N3QE3649 2009-10-24 07:28:18 Chris Trask Re: Mixers
> You said: "A balanced switching mixer can be viewed as a perfect
> multiplier which is driven by a bi-polar square wave. If you drive
> a diode ring mixer with a square wave LO, you effectively have a
> switching mixer. The same is true for the NE602 Gilbert cell mixer,
> which implements an "accurate" multiplier."
>
> OK, so this confirms what I had asked another writer, Niels earlier
> tonight. Gilbert cells can be either analogue multipliers or
> switching mixers depending on the local oscillator drive level.
>
>Gilbert's translinear innovation was his one true real addition to the art of circuit design. He did NOT invent the Gilbert (sic) cell, and many of us now refer to the six-transistor tree as a "commutative mixer", which was first suggested by Prof. Robert Clark of UC Berkely. The transistor form of the mixer first appeared in the Review of Scientific Instruments a few years before Gilbert filed his patent, and is open claim that he was not aware of a certain patent by Heck was 100% bogus as he had used it as a reference in his initial publication in the IEEE Journal of Solid-State Circuits years before his patent filing date.
>I've only just today read Gilbert's paper on his translinear input
>circuit, it will take a few more readings before I fully grasp it. I've
>read your net published papers on lossless feedback around the Gilbert
>cell mixer and found them fairly easy to grasp. Feedback makes sense to
>me, perhaps because I'm just not very mathematical in my way of
>understanding things and you've written clearly.
>
>Looking at the wealth of patents for mixers that try to improve the basic Gilbert Cell, you get the impression that correcting the basic topology is hopeless. Active devices produce distortion and noise, and passive devices produce noise.
> Your comments about reducing active devices is good sense, although
> these days, I suspect it would strike many folk as counter-intuitive.
>
>The stability factor actually improves when using augmentation, at least with the passive method.
>I've read your paper on lossless feedback with augmentation, and agree
>that it makes a mixer more difficult to build. Have you found stability
>any trickier with these methods? The use of negative feedback would
>suggest better stability, but I've seen some weird things in my time.
>
>The input equivalent noise density of 50nV/sqHz is not all that great. You would typically want to see this spec in the single digits for serious receiver work.
> What about the AD835 then? According to the spec sheet it is an
> accurate (0.1%) 4 quadrant multiplier and while not cheap, it
> isn't terribly expensive either.
>
We simply plod on, improving things one small step at a time.
Chris
,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_3650 2009-10-24 07:54:29 Niels A. Moseley Re: Mixers
> Thanks for your reply. You said "switching mixers" and "excellentHere, "linearity" refers to the case where a varying signal is applied
> linearity" in the one sentence. My confusion is multiplied. *laughs*
to one input port of the mixer (for example, the RF port) and a constant
signal is applied to the other input port (the LO port). The signal at
the output of the mixer should be a linear function of the varying signal.
The mixer should also be linear if the RF port signal is constant and
the LO port signal is varied.
A (perfect) switching mixer can be considered to be time-variant linear
circuit as it switches between to linear modes: output = input and
output = -input.
As a counter example, consider a mixer built from a single FET with a
(non-switching) sinusoidal LO, which can be roughly described by the
following equation:
output = alpha*(LO + RF)^2 + additional terms.
,where alpha depends on the FET device characteristics.
expanding the above gives:
output = alpha*(2*LO*RF + LO^2 + RF^2).
Here, the desired mixing action is the 2*LO*RF term, and LO^2, RF^2 are
both non-linear terms. These latter terms rectify the LO and RF signals,
respectively and cause (undesired) AM demodulation, intermodulation
products and DC offsets.
> Just as an example to make talking about this easier, can I propose aYes, this is a consequence of switching between the two linear cases:
> four diode, high level balanced modulator, with input RF of 10MHz, local
> oscillator square wave driven of 14MHz and a wanted difference IF of
> 4MHz:
>
> If you mean that the wanted 4MHz output has a peak to peak rf voltage
> that is a simple ratio of the input 10MHz peak to peak rf voltage unless
> it exceeds some input overload level, then there is a form of linearity
> between the wanted IF out and the RF input.
output = input and output = -input.
> That doesn't mean there is anything analogue or linear involved in theIndeed; in this case, the mixer also switches between two states: output
> local oscillator operation.
>
> Does what I have said above make sense to you? I can grasp the basic
> analogue operation of a fet mixer with square wave drive - it is the
> same exactly as a fast digital sampling scope. It gives frequency
> translation but maintains the linearity of the peak waveform voltage and
> so retains any amplitude modulation.
= input and output = 0. Theoretically, there is very little difference
between this mixer and a double balanced diode mixer. The LO has become
uni-polar (1, 0, 1 ,0 ..etc) instead of bi-polar (1, -1, 1, -1).
Consider now that unipolar = 0.5 + 0.5 * bipolar.
Then, we can write your uni-polar mixer as:
output = (0.5 + 0.5 * bipolar) * input.
which is
output = 0.5 * input + 0.5 * bipolar * input.
Note that the term '0.5*bipolar*input' is the double balanced mixer
(DBM) multiplied by one half. So the unipolar mixer output contains the
same signals as the DBM output at half the energy. Additionally, the
unipolar mixer output contains _all_ the input signals.
> Local oscillator harmonic content makes sense to me although yourIt is very similar in that they both use a multi-phase LO, but the
> information on the harmonic levels is new to me and appreciated. Your
> description of parallel mixers with complex different LO waveforms is
> also a new to me and a fascinating idea. At first glance it looks like
> an extension or development of the Tayloe circuit, or is it another
> thing altogether?
Tayloe mixer does not apply weighting and therefore suffers from the
same harmonic downmixing phenomena as a normal switching mixer. However,
the tayloe mixer has less conversion loss compared to a normal diode mixer.
> Sorry to sound pedantic, but you said: "A balanced switching mixer canWhat I mean is that there is no theoretical difference between a
> be viewed as a perfect multiplier which is driven by a bi-polar square
> wave." If it is a perfect multiplier, then it is an analogue device by
> nature - no? Or is it a switching device? I can't see how it can be
> anything but one or the other. Don't ask me to visualise photons...
> *laughs* Or do I have a conceptual problem here?
switching mixer and a perfect multiplying mixer, if the latter is driven
by a bi-polar square wave LO. They both switch between two states:
output = input and output = -input.
There are, of course, practical differences such as device noise
contributions, the circuit topologies, the effects of impedance
mismatches at the ports. But in essence, they implement the same
"mechanism", if you will.
> I've read descriptions about the Gilbert cell being an analgoueYes, I believe there are noise advantages when driving the Gilbert Cell
> multiplier, but lots of articles also talk about the upper four
> transistors operating in saturation as switches. It looks to me as
> though the Gilbert cell can be operated as a analogue multiplier and
> also as a switching mixer depending on local oscillator drive level.
in a switching mode. Perhaps Wes, Chris or Allison can comment on this?
Using the switching mode will give you harmonic downmixing as a
side-effect. I measured the harmonic downmixing effect of a NE602 (with
a 50% duty-cycle square wave LO) a while back; it follows the
20*log10(1/n) relation up to about 200 MHz pretty well.
> Is that what is muddying the waters of language? (More like quicksandWhat is muddying the waters is probably that the mixer is a three port
> of language. *laughs*) That the Gilbert cell can be a true analogue
> multiplier and a switching mixer, primary difference being that the
> local oscillator is either low enough for totally linear operation, or
> else so large that the upper transistors are in saturation?
device. If one of the two input ports is constant, it behaves as a
linear device. If the two input ports are tied together, it is a
non-linear squarer. If the two input ports are independently varied is
is also a non-linear operation (but can be sometimes considered a
time-varying linear system). Now, I hope I haven't confused you any
furhter...
73,
Niels.3651 2009-10-24 08:18:49 Chris Trask Re: Mixers
> > I've read descriptions about the Gilbert cell being an analgoue
> > multiplier, but lots of articles also talk about the upper four
> > transistors operating in saturation as switches. It looks to me
> > as though the Gilbert cell can be operated as a analogue
> > multiplier and also as a switching mixer depending on local
> > oscillator drive level.
>
> Yes, I believe there are noise advantages when driving the Gilbert
> Cell in a switching mode. Perhaps Wes, Chris or Allison can comment
> on this?
>
My goal since then has been to further linearize that last circuit, but I have also been interested in linearizing the LO path so as to further improve the overall NF by turning it into a linearized high frequency multiplier. Not an easy task, but I do have an idea or two to try, just not the time to breadboard them.
Chris
,----------------------. High Performance Mixers and
/ What's all this \ Amplifiers for RF Communications
/ extinct stuff, anyhow? /
\ _______,--------------' Chris Trask / N7ZWY
_3654 2009-10-25 10:30:54 tim Re: Mixers
The two mixers I've looked at the most, diode and FET, both have a transfer
characteristic that is closely approximated by an exponential.
An exponential has three main areas. The low-slope area leading up to
the "knee", the "knee" itself, and the high-slope area after the "knee".
Am I to take it that when you talk about about linearity in a mixer made up of
these components you are talking about either the low-slope area or the
high-slope area of the transfer curve?
Neither area is "quite" linear but the deviation from linear in these two
areas is small while around the knee the deviation from linear is large.
Is this what you folks are talking about with a "linear" non-linear mixer?
That the more time the mixer spends in a "linear" non-linear area the better
the performance?
tim ab0wr3655 2009-10-25 18:03:40 Russell Shaw Re: Mixers
> Niels,All areas simultaneously, because the large LO drive sweeps over all those
>
> The two mixers I've looked at the most, diode and FET, both have a transfer
> characteristic that is closely approximated by an exponential.
>
> An exponential has three main areas. The low-slope area leading up to
> the "knee", the "knee" itself, and the high-slope area after the "knee".
>
> Am I to take it that when you talk about about linearity in a mixer made up of
> these components you are talking about either the low-slope area or the
> high-slope area of the transfer curve?
areas switching the transistor between off and on.
> Neither area is "quite" linear but the deviation from linear in these twoTurn the volume pot on your TV up and down 5 times/sec. That is acting as
> areas is small while around the knee the deviation from linear is large.
>
> Is this what you folks are talking about with a "linear" non-linear mixer?
>
> That the more time the mixer spends in a "linear" non-linear area the better
> the performance?
a linear time-varing resistor and you now have a 5Hz modulated volume.
No matter how much audio signal is applied to the input of the pot, the
*amplitude* or envelope/spectrum of the modulated output varies linearly (or
proportionally) with the amplitude of the input. That is an ideal linear
mixer.
If the audio speaker output becomes distorted as the input to the pot (mixer)
is increased so high that the pot starts arcing over, then the system is
becoming nonlinear.
The pot makes an ideal mixer because the LO (turning the knob) is a different
physical action to the RF (audio input), and so don't interact.
With a transistor, LO and RF are both the same physical action (voltage/current)
so they interact. At high LO and low RF, the RF affects the LO action very
little, so the IF output amplitudes (in time and frequency spectra) vary
linearly with the RF input amplitudes. The transistor is acting as a time-
varying (trans)conductance with its action controlled by the LO and very
little by the RF. When the RF becomes higher, the time-varying
(trans)conductance becomes different from just the LO alone, and so
the RF/LO interaction results in extra frequency products at the IF
output because the output signal amplitudes no longer exactly follow
the input signal amplitudes.
--
regards,
Russell Shaw, B.Eng, M.Eng(Research)3668 2009-10-26 16:27:39 Glen Leinweber Mixers
mixer linearity. Superposition is the critical test.
Mixers are a great example for testing our understanding
of linearity - their output waveforms are nothing like their
input waveforms. At first glance, you might conclude
that they fail the linearity test - using the principle that
output waveshape should be a scaled replica of input
waveshape.
We are only interested in a small fraction of the
myriad mixing products that the output offers. Our mixer
is followed by band-selective filters that reject at least
half the mixing products that we regard as spurious.
What we have left is a "good" signal that -still- has different
waveshape than the mixer's input signal. But we note
that output envelope rises & falls when input envelope
rises & falls. In this respect, the mixer is operating linearly.
For rigor, test for superposition - two signals are applied
to the mixer's input. At the output, we look to see if the
amplitude of one affects the amplitude of the other. We look
for intermodulation distortion - the mark of failed superposition.
What does intermodulation distortion sound like? One type
of IMD is caused by clipping. Two diodes across
your headphones will give you a taste, when audio level
is high enough to cause them to conduct.
I'd describe it as "muddy-ing" the sound. It is quite apparent
that IMD is unwanted in a "good sounding" receiver.
Harmonic distortion is similarly unwanted. All audiophiles
agree that they're to be avoided, although arguing continues
about which root cause sounds more objectionable.3670 2009-10-26 18:57:25 dixonglennb Mixers 3673 2009-10-27 20:53:13 David Packard Re: Mixers
understand it! Would you please use the same language to explain, to us
folks trying to understand how radios work, exactly how a mixer works?
I've been through books and websites trying to really figure it out (I
think Bill Meara is on a similar quest), but it doesn't quite make sense
to me.
Thanks and 73, Dave K1YHR.3674 2009-10-28 03:25:18 Russell Shaw Re: Mixers
> Russell, your explanation of mixer nonlinearity was so clear that IHi,
> understand it! Would you please use the same language to explain, to us
> folks trying to understand how radios work, exactly how a mixer works?
> I've been through books and websites trying to really figure it out (I
> think Bill Meara is on a similar quest), but it doesn't quite make sense
> to me.
It is hard to explain much detail without sketches. I had the lab packed
at RMIT when explaining my sketches of these kinds of things when i was
running tutorial years ago.
Imagine a 10Vpk 10MHz sinewave applied to a series diode, with a 1k
resistor connected from the other side of the diode to ground.
The output will be half-sine pulses.
Now add a 1V 10.010MHz sinewave in series.
Looking at the sum on a CRO, you'll see a ~10MHz carrier with a
sinewave AM envelope with 1Vpk amplitude and 10kHz frequency.
The large carrier is going between 9Vpk and 11Vpk 10000 times/sec.
Note that the average voltage over a 100us (10kHz) period is close
to zero.
A lowpass filter gives the *average* of its input voltage, and so
would show zero with this signal.
On the 1k resistor, you'll see the same waveform with the negative
half chopped off. Obviously now, the average of the waveform over
a large number of carrier cycles is around 10Vpk.2/pi = 6.36V,
and is alternating between 5.73V and 7.00V at a 10kHz rate.
If the LPF has a cutoff frequency higher than 10kHz, it will
display an output that is the average of the input, but fast
enough to follow the 10kHz rate. Therefore, connecting it to
the 1k resistor will show at its output a 0.635Vpk 10kHz sinewave
riding on 6.36V DC.
See the "Detector" waveform diagram at:
http://www.vintage-radio.com/repair-restore-information/transistor_if-rf-stages.html
That is mixing similar to any diode or transistor where the
RF and LO appear summed as seen by the device.
In a diode-ring mixer, the sinewave action appears more as
diode currents rather than voltages.
In a single-transistor mixer, because of an emitter bypass capacitor,
the transistor usually conducts for a shorter period like class-C,
and so there is no resistor load to think of; only the intrinsic
transistor impedances. These BJT circuits usually have high gain
but are overloaded easier. They are commonly self-oscillating
autodyne converters in broadcast receivers.
http://www.vintage-radio.com/repair-restore-information/transistor_if-rf-stages.html
At VHF/UHF, emitter inductance lengthens the conduction angle to almost class-A,
but not quite. The mixing action still happens, but with significant inductive
emitter impedance, which can have positive effects on large-signal performance.
My favourite mixer circuits are one-transistor bipolars with plenty of tuned
circuits. High gain and low current for portable battery operated ham-band
receivers.3675 2009-10-28 06:28:21 Russell Shaw Re: Mixers
> David Packard wrote:...
> On the 1k resistor, you'll see the same waveform with the negativeCorrection, the wave is halfwave rectified, so the averages
> half chopped off. Obviously now, the average of the waveform over
> a large number of carrier cycles is around 10Vpk.2/pi = 6.36V,
> and is alternating between 5.73V and 7.00V at a 10kHz rate.
should be halved (1/pi instead of 2/pi).