This was the last one with ceramic filters:
https://www.nxp.com/docs/en/data-sheet/TEF6730A.pdf , see pages 45...48.
https://www.nxp.com/docs/en/data-sheet/TEF6730A.pdf , see pages 45...48.
Last edited:
Terrifying at first blush, but then it looks more classical after study. It seems like a classical view of tuning would apply?
Are the latest FM tuners completely free of "tuning"?
Much thanks, as always,
Chris
ps: haven't forgotten my promise of a working MvdG phono eq. Have been without mains power six weeks, and other excuses, less dramatic
Are the latest FM tuners completely free of "tuning"?
Much thanks, as always,
Chris
ps: haven't forgotten my promise of a working MvdG phono eq. Have been without mains power six weeks, and other excuses, less dramatic
Talking about IF filters, this might be relevant:
https://www.diyaudio.com/community/...mpletely-different.422870/page-2#post-7952822
https://www.diyaudio.com/community/threads/ceramic-filter-selection.261137/
https://www.diyaudio.com/community/...mpletely-different.422870/page-2#post-7952822
https://www.diyaudio.com/community/threads/ceramic-filter-selection.261137/
178BBR-3132A six-pole linear-phase LC IF filters are offered on eBay now, under item number 285930909082. I have no affiliation with the seller.
@Chris Hornbeck
Six weeks? That's quite long...
The TEF6730 was fairly classical, although DACs and digitally stored trimming values were used to let the RF filter track the LO, rather than manually adjusted trimmers and padders.
Low-IF receivers usually have a fixed LC filter at the input, then a half-complex mixer (a.k.a. I/Q mixer) and a complex-valued IF filter (a.k.a. I/Q filter) to get some image rejection. Anything that requires tuning is tuned automatically, with a PLL for the LO and an on-chip tuning loop for IF filter tuning, if the IF filter is analogue.
You can describe what I'm about to write in terms of goniometric functions, but the description in terms of complex numbers is much clearer once you get used to it.
Mathematically, in terms of double-sided spectra, when you frequency-convert a signal by multiplying it with an LO signal cos(2πfLOt), you mix it both up and down in frequency, because cos(2πfLOt) = 1/2 exp(2jπfLOt) + 1/2 exp(-2jπfLOt).
If you want to, for example, convert a (real-valued) 100 MHz signal to 300 kHz by multiplying it with a 99.7 MHz cosine, you actually have signals at -100 MHz and +100 MHz that get converted to -199.7 MHz, -0.3 MHz, +0.3 MHz and +199.7 MHz. A real-valued signal at 99.4 MHz with a 99.7 MHz LO would result in -199.1 MHz, +0.3 MHz, -0.3 MHz, +199.1 MHz. Hence, you get image reception at 99.4 MHz.
Now suppose you use an LO signal exp(-2jπ • 99.7 MHz • t) instead. A real-valued 100 MHz RF signal with spectral components at +/- 100 MHz gets converted to -199.7 MHz and +0.3 MHz, while a real-valued RF signal at 99.4 MHz only produces -199.1 MHz and -0.3 MHz. A complex IF filter can pass +0.3 MHz and suppress -0.3 MHz, so there is no image reception.
In practice, the image suppression you can get is limited by mismatch in the half-complex mixer and complex IF filter. With digital IF processing, there are algorithms that automatically correct for this mismatch.
Six weeks? That's quite long...
The TEF6730 was fairly classical, although DACs and digitally stored trimming values were used to let the RF filter track the LO, rather than manually adjusted trimmers and padders.
Low-IF receivers usually have a fixed LC filter at the input, then a half-complex mixer (a.k.a. I/Q mixer) and a complex-valued IF filter (a.k.a. I/Q filter) to get some image rejection. Anything that requires tuning is tuned automatically, with a PLL for the LO and an on-chip tuning loop for IF filter tuning, if the IF filter is analogue.
You can describe what I'm about to write in terms of goniometric functions, but the description in terms of complex numbers is much clearer once you get used to it.
Mathematically, in terms of double-sided spectra, when you frequency-convert a signal by multiplying it with an LO signal cos(2πfLOt), you mix it both up and down in frequency, because cos(2πfLOt) = 1/2 exp(2jπfLOt) + 1/2 exp(-2jπfLOt).
If you want to, for example, convert a (real-valued) 100 MHz signal to 300 kHz by multiplying it with a 99.7 MHz cosine, you actually have signals at -100 MHz and +100 MHz that get converted to -199.7 MHz, -0.3 MHz, +0.3 MHz and +199.7 MHz. A real-valued signal at 99.4 MHz with a 99.7 MHz LO would result in -199.1 MHz, +0.3 MHz, -0.3 MHz, +199.1 MHz. Hence, you get image reception at 99.4 MHz.
Now suppose you use an LO signal exp(-2jπ • 99.7 MHz • t) instead. A real-valued 100 MHz RF signal with spectral components at +/- 100 MHz gets converted to -199.7 MHz and +0.3 MHz, while a real-valued RF signal at 99.4 MHz only produces -199.1 MHz and -0.3 MHz. A complex IF filter can pass +0.3 MHz and suppress -0.3 MHz, so there is no image reception.
In practice, the image suppression you can get is limited by mismatch in the half-complex mixer and complex IF filter. With digital IF processing, there are algorithms that automatically correct for this mismatch.
Please forgive my ignorance (as you have done so graciously in the past, enboldening me yet again) but, is this something only possible by serious digital calculations, and not possible in real-world-RLC components? That seems likely true because it was never done, but also seems unlikely (to be true) because ordinary L's and C's have complex values, so.... If the subtleties are beyond me, it won't be the first or (hopefully) the last time.
As always, much thanks,
Chris
Unfortunately these low IF, DSP based radios rarely seem to work as well as a properly done 10.7 MHz design. Cheaper to make and more versatile, which is why they have taken over.
A great pity that antenna diversity for car radios never happened. It helps at VHF FM and even more at DAB frequencies.
A great pity that antenna diversity for car radios never happened. It helps at VHF FM and even more at DAB frequencies.
Hi Chris,
The filter powered jig is simply a splitter and impedance match (50R to 330R I think), the filter is plugged in after you do a correction on the network or spectrum analyzer, then the load to impedance match / buffer out to the network or spectrum analyzer. Using the network analyzer I get amplitude and phase information. This is from distant memory when I designed it.
I use an HP 3585A spectrum analyzer, or HP 4195A Network Analyzer for this. For quick stuff, the 3585A gets the nod. Compensating for the jig takes less time with the 3585A.
Anyway, it is about as simple as I could make it while keeping performance high enough for the job.
The filter powered jig is simply a splitter and impedance match (50R to 330R I think), the filter is plugged in after you do a correction on the network or spectrum analyzer, then the load to impedance match / buffer out to the network or spectrum analyzer. Using the network analyzer I get amplitude and phase information. This is from distant memory when I designed it.
I use an HP 3585A spectrum analyzer, or HP 4195A Network Analyzer for this. For quick stuff, the 3585A gets the nod. Compensating for the jig takes less time with the 3585A.
Anyway, it is about as simple as I could make it while keeping performance high enough for the job.
Not the case
- my '06 VW Golf has antenna diversity ( roof mounted stub antenna and rear window antenna). It isn't present in the base model though.
- my old '03 Saab 9-5 also had it.
The distance between the two car mounted antennas wasn't enough to make a large enough difference for one. Now most people listen to MP3 or stream from their phones (like me), although I do listen to FM in the car as well.
The two Revox tuners I have are set up for two antenna inputs. Never used more than one.
The two Revox tuners I have are set up for two antenna inputs. Never used more than one.
With the exception of chips meant specifically for cheap aftermarket radios, all car radio chips I worked on in the past two decades had to support diversity reception. Not just switching diversity, but also phase diversity, that is, taking a weighted sum of the signals from the two aerials while correcting for phase differences, turning them into a kind of self-adjusting phased array. No idea how the radio software people managed to make that work, but they did.
Last edited:
What's a bit confusing is that we often use complex numbers to analyse circuits that have only real-valued inputs and outputs. The real-valued signals are then (more often than not implicitly) written as sums of two complex-conjugate signals, just to make the circuit analysis easier. In the case of a low-IF receiver with complex IF, the circuit really has complex-valued in- and outputs.
How would one go about recording a complex signal on, for example, a compact cassette? That's simple: define that the left channel represents the real part and the right channel represents the imaginary part, and any old stereo cassette recorder can record and play back complex signals.
A half-complex mixer has a real RF input, a complex LO input and a complex IF output. When you work it out, it turns out that a half-complex mixer is nothing but two ordinary mixers getting the same RF signal and getting LO signals that are 90 degrees out of phase. The output signal of one mixer is then regarded to be the real part of the complex IF signal and the output signal of the other mixer regarded to be the imaginary part.
A complex filter is a bit more complicated. You could just make two identical analogue state variable filters, call the input and output signals of one of them the real parts of the complex signals and call the input and output signals of the other the imaginary parts of the complex signals, but you then always get a response that is the same for positive and for negative frequencies (or same magnitude, opposite phase response to be precise). It turns out that you can solve that by adding cross-couplings between the two filters.
So all in all, complex filters can be fully analogue and can be digital as well. The amount of image rejection you can get with an analogue implementation is limited by analogue imperfections such as mismatch between the real and imaginary paths, usually to somewhere between 30 dB and 50 dB. With a partly digital implementation that corrects for the analogue imperfections (so-called I/Q correction), that can increase to 80 dB...100 dB.
Someone who really, deeply understands something can convince even a dummy like me that he could also understand it with a little effort. You're a wonderful teacher.
The basics seem related enough to SSB (single-sideband) modulation and de-mod that I might be able to approach from that angle.
Again, much thanks for your time and effort,
Chris
The basics seem related enough to SSB (single-sideband) modulation and de-mod that I might be able to approach from that angle.
Again, much thanks for your time and effort,
Chris
That's correct, it is very much related to the phasing or Hartley method of SSB modulation and demodulation (which apparently dates back to 1924 - there is nothing new under the sun).