Out of the various op and small power amps , similar to NE5532, TDA 2822 etc , want the best as of background noise, something like a practical, not theoretical at or less than 1 uvolts noise level. It’s needed for a performance HF radio reciever. They often use the noisy LM386, want something better.
Again, what's the impedance driving the amplifier?
If it's very low, you can make a gain stage with an LT1028 or AD797 followed by a headphone amplifier, when it's not so low, a modern JFET op-amp is more suitable. In both cases you have to keep the resistance of the feedback network low because of its thermal noise.
In both cases a discrete gain stage with a BF862 or similar high transconductance JFET should work well, or with a bunch of them in parallel if needed. If the impedance is low, you could also make a discrete amplifier with a good bipolar transistor (like the 2SC2547, if you can still get them somewhere).
If it's very low, you can make a gain stage with an LT1028 or AD797 followed by a headphone amplifier, when it's not so low, a modern JFET op-amp is more suitable. In both cases you have to keep the resistance of the feedback network low because of its thermal noise.
In both cases a discrete gain stage with a BF862 or similar high transconductance JFET should work well, or with a bunch of them in parallel if needed. If the impedance is low, you could also make a discrete amplifier with a good bipolar transistor (like the 2SC2547, if you can still get them somewhere).
If the impedance is low, this thread could be of interest:
ultra-low noise transistors.
By the way, are you sure it is the amplifier setting the noise floor, not the atmospheric noise or the radio front-end?
ultra-low noise transistors.
By the way, are you sure it is the amplifier setting the noise floor, not the atmospheric noise or the radio front-end?
That's exactly where the dominant source of noise is. Reducing noise down stream will have little effect.
Sounds like you want to use ICs only.
Then maybe try 4x AD8429 in parallel, set to 100x gain with 0.1% resistors.
And then sum the output to reduce noise to 0.5nV/sqrtHz.
The input impedance per AD8429 is 1.5G//3pF.
So it should be > 350Mohm even with 4x in parallel.
So no issue with 10k source impedance.
And I am not sure whether you can do better with discrete JFETs (maybe only with cascoding, etc.)
Cheers,
Patrick
Then maybe try 4x AD8429 in parallel, set to 100x gain with 0.1% resistors.
And then sum the output to reduce noise to 0.5nV/sqrtHz.
The input impedance per AD8429 is 1.5G//3pF.
So it should be > 350Mohm even with 4x in parallel.
So no issue with 10k source impedance.
And I am not sure whether you can do better with discrete JFETs (maybe only with cascoding, etc.)
Cheers,
Patrick
The AD8429 has an input noise current of 1.5 pA/sqrt(Hz). Put four in parallel and you have 3 pA/sqrt(Hz) (it increases with the square root of the number because the noise currents will be uncorrelated). 3 pA/sqrt(Hz) across 10 kohm is equivalent to 30 nV/sqrt(Hz), so your four AD8429's could be outperformed by a single TL071 or by almost any discrete JFET if the source impedance is indeed 10 kohm.
Last edited:
If you want to stick to ICs: an OPA827 has an input noise voltage of 4 nV/sqrt(Hz) at 1 kHz and an input noise current of 2.2 fA/sqrt(Hz) at 1 kHz. To get below the thermal noise of 50 ohm, you could put (at least) 20 op-amp stages in parallel. The noise current is then still small enough for 10 kohm source impedance. The only problem is that it becomes quite expensive: at Farnell, 20 pieces OPA827 in SO-8 package would cost you 166 euro.
Use lm4562 with battery supply. When need driving higher power ad lme49600 like the wire design.
Just for fun
Just for fun
OP, what sort of supply voltages do you have available? Will you need any gain from the stage? Is power consumption critical?
In terms of maximum output, it's generally nice to have about 2 Vrms available. It may be hard to accommodate both very sensitive IEMs (>=130 dB / 1 Vrms) and very insensitive phones like some 600 ohm jobs or orthodynamics (which can be down at 90 dB / 1 Vrms) without some sort of gain selection.
A step up from an LM386 is, let's say, not exactly hard. This thing is not super low noise and its minimum gain is 26 dB. For the purposes of a radio, I'm pretty sure a unity gain follower using an NJM4556A with some decent close-by rail bypassing and 22 ohm in series will be fine. That's about 2 µV or so right there, which should be absolutely inaudible in just about anything, and the IC remains useful down to relatively low voltages (like 9 V and below). Of course if you need any more gain, you'd have to provide it beforehand and follow it up with the volume pot. (Chances are a '4558 would get the job done fine, just go easy on output loading with this oldie. It likes 10k up at higher levels. You can always use a 4580 or 5532 if you want, those have a lot more oomph.)
If you need some gain at same or even lower-noise, you could employ an NJM2068 with some sort of buffer stage. (JDS Labs' latest amp does just that, actually.)
Hanging around the audio guys is not a bad idea if you're into RF, you can always pick up a trick or two to make your supply rails quieter (though not necessarily more stable). It can only ever be good for phase noise...
In terms of maximum output, it's generally nice to have about 2 Vrms available. It may be hard to accommodate both very sensitive IEMs (>=130 dB / 1 Vrms) and very insensitive phones like some 600 ohm jobs or orthodynamics (which can be down at 90 dB / 1 Vrms) without some sort of gain selection.
A step up from an LM386 is, let's say, not exactly hard. This thing is not super low noise and its minimum gain is 26 dB. For the purposes of a radio, I'm pretty sure a unity gain follower using an NJM4556A with some decent close-by rail bypassing and 22 ohm in series will be fine. That's about 2 µV or so right there, which should be absolutely inaudible in just about anything, and the IC remains useful down to relatively low voltages (like 9 V and below). Of course if you need any more gain, you'd have to provide it beforehand and follow it up with the volume pot. (Chances are a '4558 would get the job done fine, just go easy on output loading with this oldie. It likes 10k up at higher levels. You can always use a 4580 or 5532 if you want, those have a lot more oomph.)
If you need some gain at same or even lower-noise, you could employ an NJM2068 with some sort of buffer stage. (JDS Labs' latest amp does just that, actually.)
Hanging around the audio guys is not a bad idea if you're into RF, you can always pick up a trick or two to make your supply rails quieter (though not necessarily more stable). It can only ever be good for phase noise...
Last edited:
Uhm, the antenna hiss is microVolt, but the whole point of the IF stages (or whatever technology) is usually to bring signal up to Volt levels before detection.
Ah, there is direct conversion to Audio in which the initial audio level may be less than antenna level. Depending what the RF is, this would often lead to a LOW impedance at the audio port. But none of this is in my old-old "Radio For Boys" books. I have an impression that an audio transformer is the next best step after the multiplier.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.