I am an ex marine radio officer, ex film sound studio technician, and a ham. I want to catchup with the new technologies.
There is a net distinction between atmospheric noise, and inherited radio reciever noise. In theory, it’s not much of use to go for a low noise receiver circuitry, but, in practice, aiming at lower receiver noise, will give a better reception of a weak signal, almost buried in the atmospheric noise.
The atmospheric noise changes hourly, if not less, due to varying propagation conditions, and the solar radiation, known as the sunspotcycle. There is nothing we can do here, except go digital mode!
In the old days, Hf receivers used to have one, or even more, Rf amplifier stage to compensate for the noise created by the mixer and the IF stages. The latest technology is to make use of switching mixers, using the cmos 4066 bilateral switch, or , much more improved, later versions, of the FST type, which have very low mohm resistance.
Also, many recent designs are using the direct conversion method, eliminating both the RF and IF stage, and all the gain, and filtering of the receiver is done at audio level. Hence the need of low noise, but, not necessarily lowest distortion, as we are talking of radio, not pro or hifi here!
There is a net distinction between atmospheric noise, and inherited radio reciever noise. In theory, it’s not much of use to go for a low noise receiver circuitry, but, in practice, aiming at lower receiver noise, will give a better reception of a weak signal, almost buried in the atmospheric noise.
The atmospheric noise changes hourly, if not less, due to varying propagation conditions, and the solar radiation, known as the sunspotcycle. There is nothing we can do here, except go digital mode!
In the old days, Hf receivers used to have one, or even more, Rf amplifier stage to compensate for the noise created by the mixer and the IF stages. The latest technology is to make use of switching mixers, using the cmos 4066 bilateral switch, or , much more improved, later versions, of the FST type, which have very low mohm resistance.
Also, many recent designs are using the direct conversion method, eliminating both the RF and IF stage, and all the gain, and filtering of the receiver is done at audio level. Hence the need of low noise, but, not necessarily lowest distortion, as we are talking of radio, not pro or hifi here!
Aren't you worried about intermodulation between different stations then? Or is there a channel-selective passive low-pass filter between mixer and amplifier?
Do you know the on resistance of the switches and the aerial/antenna impedance for frequencies around the RF frequencies you tune to? What kind of aerial/antenna do you use?
I'm still wondering if the source impedance range can't be narrowed down a bit. The typical ultra-low voltage noise, ultra-high current noise op-amps (LT1028, AD797) should work fine (F < 3 dB) for anything between 50 ohm and 1 kohm, but not at 10 kohm.
Do you know the on resistance of the switches and the aerial/antenna impedance for frequencies around the RF frequencies you tune to? What kind of aerial/antenna do you use?
I'm still wondering if the source impedance range can't be narrowed down a bit. The typical ultra-low voltage noise, ultra-high current noise op-amps (LT1028, AD797) should work fine (F < 3 dB) for anything between 50 ohm and 1 kohm, but not at 10 kohm.
It still isn't clear to me whether the thread starter prefers op-amps over discrete circuits.
The BF862 is very good, but it is also out of production. The ON Semiconductor 2SK932 is cheap, still in production and quite similar to the BF862, but its white noise specs are worse - which is strange, because FETs with similar transconductance are supposed to have similar white noise.
An InterFET IF3601 would be even better and it is available at Mouser, but it is quite expensive.
The BF862 is very good, but it is also out of production. The ON Semiconductor 2SK932 is cheap, still in production and quite similar to the BF862, but its white noise specs are worse - which is strange, because FETs with similar transconductance are supposed to have similar white noise.
An InterFET IF3601 would be even better and it is available at Mouser, but it is quite expensive.
Dimitri measured them and they have excess GR noise making them not suitable for audio. Dimitri updated his FET noise article in AX (I can't post it because it's copyrighted) but the CPH3910 (same AM radio app as BF862) looks very good.
+1 on Ulrich Rohde his stuff is great.
Last edited:
Scott, can you link us to it please?
Serracin, I'm sure you don't need reminding....
Sun up frequency up,
Sun down frequency down.
Semper Fi brother
Cheers,
Last edited:
Thanks again, folks for your kind interest and valuable info on the subject.
Designers and critics , much more technical than yours truely, have long established that the mixer, known as the Tayloe mixer, adapted as the H mode switching mixer, using very, very low “on” resistance have excellent intermodulation properties, high sensitivity and very low noise . Classic example is the FST3253. They also have rediscovered the old, direct conversion receiver, where, audio, is directly detected, straight after the mixer, eliminating the IF, RF stage. So, all the receiver processing , are at audio level, hence the need of high performance audio circuitry. It’s much less noisy, for an audio filter, than an RF crystal filter, as used in IF amplifier, where, peaking, and restricting certain RF frequencies, tends to peak some annoying inside noise.
IC’s are a convenient way to make a high performance circuit “ in a nut shell” . But, still have to convince myself, that IC’s have lower noise than discrete! I come from the 741 era!!! The aim here is noise, rather than fidelity!
Designers and critics , much more technical than yours truely, have long established that the mixer, known as the Tayloe mixer, adapted as the H mode switching mixer, using very, very low “on” resistance have excellent intermodulation properties, high sensitivity and very low noise . Classic example is the FST3253. They also have rediscovered the old, direct conversion receiver, where, audio, is directly detected, straight after the mixer, eliminating the IF, RF stage. So, all the receiver processing , are at audio level, hence the need of high performance audio circuitry. It’s much less noisy, for an audio filter, than an RF crystal filter, as used in IF amplifier, where, peaking, and restricting certain RF frequencies, tends to peak some annoying inside noise.
IC’s are a convenient way to make a high performance circuit “ in a nut shell” . But, still have to convince myself, that IC’s have lower noise than discrete! I come from the 741 era!!! The aim here is noise, rather than fidelity!
They haven't. If you have no problem with discrete circuits, the clear recommendation is to make a first gain stage with a BF862, CPH3910 or IF3601 JFET. The unclarity about the source impedance that drives the stage doesn't matter then, because these FETs combine low voltage noise with low current noise (as long as you don't forward bias the gate-source junction and don't make the drain-source voltage higher than needed).
The fact that you worry about mixer intermodulation but not about audio amplifier intermodulation means that you either have a low-pass filter between mixer and audio amplifier, or that you haven't given it enough thought.
The fact that you worry about mixer intermodulation but not about audio amplifier intermodulation means that you either have a low-pass filter between mixer and audio amplifier, or that you haven't given it enough thought.
Yes there are several composite amps like gerhard's that would give you the low noise and everything else.
Not to hijack this thread, I have moved the JFET noise discussion to here :
More FET noise measurements (for EUVL)
Patrick
If you give it even more thought, you'll notice that mixers make their living on
non-linearity, while amplifiers don't.
BTW I have made a redesign of the 20 x ADA4898 preamp. It now has an
alternative FET op amp input for browsing with a scope probe, with much worse
noise, of course. The lo noise input now has wet slug tantals, gain and corner
frequencies are switched with ADG1404 etc, controllable in SPI style via
opto couplers.
Early this morning I have sent the Gerber files to PCBWAY in China; 10 boards
cost $5 (together!), DHL gets $25 for the transport. Weird, isn't it?
Board size is 10*9 cm, I wonder if they can produce anything that works
at these prices.
Cheers, Gerhard
Attachments
Just as an FYI to anyone interested Tom Lee has published some excellent stuff on mixer noise (actually brilliant stuff). I don't know if any of it exists outside of a paywall.
In general, they don't.
At high impedance, it is a wash. Above a few dozen Kohm you can't beat a 19 cent TL072.
At low impedance you need a BIG device. IC makers are all about getting the most functionality in the least Silicon. The LM394 was one of the first "ICs" (100 transistors in one can) which broke that model. There are several others optimized for ~~100r hiss. The Tayloe paper suggests a very fine one.
You can also slap in a fat FET. Except the Tayloe needs four inputs. And as discussed, the bias resistors add hiss. _I_ would first-crack just adopt Tayloe's design verbatim.
On a broader look-out: in audio when facing hundred-ohm sources, a transformer is often "a good way" to transform to a better noise fit. Except in studio work we regret the band-limit of transformers; also the cost to fit 64 transformers in a studio console. However here your bandwidth demand is small (and easily met at low Z) and you only need two. A couple 150:600 telephone (modem) transformers may get good noise figure with 59-cent NE5532 chips. There are also now for-purpose microphone preamp chips.
Maybe, but only for their default process. Any option drives the
price high. Immersion gold = 3 times, purple solder mask = 11 times,
plugged vias +$150.
And they seem to be fast. It is just the next day, and they have
declared delivery to DHL!
I have not even ordered the missing parts from Digi-Key.
cheers,
Gerhard
He has written that book on low noise oscillators together with A. Hajimiri.
Google finds at least 1 powerpoint/pdf and a thousand citations.
Cheers, Gerhard
My friend mentioned this before it's part of his research I think and that's must be keeping him busy and this makes me want to know more of it.
That's all good and well, but how are you going to build a general purpose receiver just with direct conversion? I/Q mixing, as for an SDR frontend? (If so, where is the processing supposed to take place?) Or do you want this strictly for CW? How do you intend to handle AGC or at least manual gain control? (While I have dabbled in RF before, I am not familiar with how these direct conversion rigs tend to be constructed.)
BTW, if IF noise is dominating your noise floor, then either that's not a particularly good receiver or you mistakenly connected a bad excuse for an antenna. Receiver designers have generally heard of noise figure, Friis' formula and atmospheric noise levels.
Bringing the signal up to a level where it can be fed to a headphone amp of moderate gain is the job of the stuff between the mixer and the volume pot. You wanted a headphone amp, you got one. It just seems you're still missing all of that stuff in between, and I see some very large question marks there. You don't even seem to know where you want to go, let alone being able to specify what you need to get there. Engineers always want specs - input level, source impedance, output level, output impedance, frequency response, all of that good stuff.
Is an input voltage noise density of 0.9 nV/sqrt(Hz) low enough? That's the performance level of the lowest-noise integrated opamps available, it's equivalent to about a 50 ohm resistor. If you need even less, then yes, you will need to go discrete (or at least hybrid with a discrete input stage) with a bunch of parallel low-noise BJTs or JFETs. That's what people do in high-performance MC phonopres (or MC prepres). You can get down to 0.5 nV/sqrt(Hz) or so without any extreme measures. Suffice it to say that the power supplies and/or PSRR must be up to par!
Making use of very low input noise in practice can actually be quite hard if you need no more than moderate gain (to maintain high dynamic range), as feedback network impedance inevitably becomes quite low. At high levels, the output stage can break quite a sweat!
Let's say you wanted to make a 14 dB amp fully making use of an AD797 and used resistor values of 22 and 91 ohms, respectively - you probably wouldn't get 6 Vrms out of the poor thing, as it would start clipping first, not to mention distortion performance. (Mind you, that's asking for 138 dB of dynamic range, 20 kHz BW. You'd also exceed the dissipation limits of 1/4 W resistors along the way.) And that's a high-performance part. Trying the same exercise with an NJM2068 is going to result in similar issues, just with about 10 dB less dynamic range per side to begin with and more distortion issues. (I guess you could still parallel both halves since it's a dual, that should work OK again.) That's when discrete buffer stages come in handy, as otherwise used for headphone amps.
No way around it - high analog dynamic range requires corresponding power levels. Hence why portable audio players traditionally rely on DACs of relatively modest dynamic range (and minimal power draw) plus a PGA to shift this dynamic range around as needed for the headphones being used. Us humans tend to be quite happy with about 90 dB in the right spot after all.
Last edited:
And they seem to be fast. It is just the next day, and they have
declared delivery to DHL!
I have not even ordered the missing parts from Digi-Key.
In the mean time, the boards of the new version of the 220pV/rtHz
amplifier have arrived. Less than a day production, 3 days to
Leipzig, Germany and a week+ at the customs office, since 10 boards
must be big business and the import is important for the federal statistics. :-(
I have finally found a home for some 35year+ optocouplers, original HP.
I have also already found a missing wire, happened by ctl-c ctl-v in the
schematics, but harmless.
Looking at the semiconductors, that should be an AD app note.
There are some excess boards, if someone seriously feels like
soldering under the microscope.
cheers, Gerhard
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.