TheRadioBoard

A couple of questions on this point, all of which may be summarized as "how difficult is it to achieve this sensitivity in practice?"

First, consider SM0VPO's "quick receiver":

http://www.sm0vpo.com:800/rx/quickrx.htm

With sufficient audio gain, is this receiver capable of -130 dBm sensitivity?

Second, consider a homebrew diode ring mixer with matched 1N4148 or 1N914 diodes, symmetrical physical construction, and toroidal, trifilar-wound baluns. Again: is -130 dBm reasonable to expect? Are there specific pitfalls that will compromise this sensitivity?

Finally, regens have a reputation for being the most sensitive of the "simple" receivers. But it we consider SM0VPO's "quick receiver", that also seems pretty simple, assuming we have a VFO and AF amp available. Is there anything that in practice makes it difficult for simple DC receivers to match regen sensitivity?

Correct.

Below that frequency, the capacitor voltage (Q output) will increase, and above that frequency the capacitor voltage will decrease. My guess is that, in a pinch you could cover the entire 3 to 30 Mhz range with a single fixed capacitor, since the 10:1 frequency range would translate to a 10:1 voltage range. Assuming that you're using TTL level (5 volt) signals for the LO, your I and Q outputs would only vary between about .5 to 5 volts (depending on your input transformer turns ratio), which is going to be easy to amplify and clip.

If "simple" means just a detector and maybe an AF amp, then a "regen DETECTOR" is the most sensitive with one exception! (Super-regen).




Other than the above fact?
NO Not really, but it depends what you call "simple".

Gaijin,

To quote from Harry...

The final receiver is as stable as the RF Sig-gen and as sensitive as the AF Amplifier. A typical stereo amplifier will give out its full audio if there is one millivolt at the PHONO input. You can hear signals that are less than 1% of this (10uV).

A simple DC rx using diode switches has sensitivity limited by the losses incurred by the input network, the on-resistance of the diodes, output network losses and input noise figure of the audio amplifier. Harry bases his figures on the use of the phono input sensitivity of a typical stereo amplifier. That type of amplifier is not designed for extreme sensitivity since the signal source will be relatively noisy to begin with. Despite that he claims (and I concur) that <10uV signals can be heard. 10uV is equal to -87dB in a 50 Ohm system.

A typical DBM will have a loss factor of 7dB. Add to this perhaps 3dB for input matching losses. That leaves output network losses to consider. Harry's circuit is not optimized in that regard. To extract maximum performance the diodes should be followed by a diplexer to terminate them in a load that is essentially resistive from DC to daylight. That can be approximated with as little as one resistor, two capacitors and one rf choke. You can see several potential designs on VE7BPO's QRP POPS website. Lets say that you can reduce output coupling losses to 3dB. Next will be the input noise figure of the audio stages. Using a common base configuration a noise figure of 1 to 2dB can be achieved with nothing more exotic than a 2N3904. So we have roughly 15dB of losses that must be added to the -130DB at this point if we want sensitivity of a bit less than 0.1uV. That works out to 145dB of gain to bring that signal up to 0dB (actually 0dBm is more correct) to 1mW or 223.6mV. That's a powerful lot of gain except.... we don't need that much!! My headphones have an SPL of 95dB at 1mW. That's ear-piercing loud. I can throw away 40dB and still have reasonably (and comfortable) headphone volume. Now we are down to only needing 105dB gain... or so it would seem. The catch there is we only need that much gain if we are fortunate enough to live where the noise floor is at -130dB or lower. For most of us that is not the case. At low through mid-HF ranges the realistic noise floor tends to be around -90 to -100dBm (7.1uV to 2.2uV). Realistically speaking, we can throw away another 40dB of gain requirement leaving us needing only 65dB gain to hear 'band noise'. That's easy to do with with only a few stages.

Comparing apples and oranges... let's take a common NE602/LM386 DC rx and compare it with an 0V1 regen. The NE602 has conversion gain of about 12dB with relatively low noise figure and the LM386 at gain of 200 is about 20dB... total 32dB gain and somewhat noisy mostly due to the LM386. Now let's suppose a J310 FET detector driving an LM386 at gain of 200. The regen detector is capable of 20dB (maybe a bit more) gain but will exhibit a much higher noise figure than an NE602. The LM386 is still at its somewhat noisy 20dB. The 0V1 would seem to have an 8dB advantage in sensitivity over the DC rx... but does it? My contention is that the noise factor of the regen detector will negate that advantage. Adding more gain between the NE602 and the LM386 will improve the weak signal performance providing it also lowers the noise floor in the audio chain. Adding the same gain to the 0V1 will not achieve the same effect since the noise floor has already been set by the regen detector. It will only serve to make both signal and noise equally stronger.

I can attest from experience that at my location any signal I can *hear* with Icom 7200 I can also hear with my modified Polyakov DC rx. Both of them feeding my computer for WSPR/QRSS work exhibit the same SNR. I cannot say the same for my regens. Signals that are close to the noise floor on the Icom and the DC rx are buried in the noise on the regen. I would estimate, at best, my regens need a signal of at least 2-3uV to achieve audibility and at least 10uV for readable CW, 20uV for SSB and 30uV or greater for AM. That sounds like a lot but by comparison to the sensitivity specs for commercially made military communications grade regens (the RAK/RAL7s) it is about the same!!

What the regen has going for it that is lacking in the DC rx is its selectivity. At the threshold of oscillation I see selectivity of about +/-2KHz centered on the frequency of oscillation. A DC rx would need the addition of a 2KHz lowpass filter added after the detector. That adds to the complexity and cost. Both detection methods suffer from the same DC offset problems but the DC rx is more amenable to use as a converter to an audio frequency IF. The same, of course, could be said of a regen+BFO configuration.

Enough obfuscation for now. Time for me to go to bed <G>

73,

'Bear' NH7SR

Thanks for the very informative post, qrpbear. I'm sure I'll be referring to it again to digest it fully.

About your experience that regens need 2-3 uV for audibility, here are a few links to regens that claim very high sensitivity:

A Charles Kitchin design with cascode RF amp and 0.3 uV sensitivity in oscillating mode:
http://www.arrl.org/files/file/Technolo ... hb1768.pdf

A PA2OHH regenerative reflex design with common-emitter RF preamp and 0.05 uV sensitivity (!) for CW:
http://www.qsl.net/pa2ohh/02reg.htm

Perhaps the pre-detector RF amps in these designs are important to improve weak signal response, though PA2OHH reported problems with overloading.

It seems most phasing receiver designs I've seen use 90 degrees shift in the LO and 90 degrees shift in the AF chain. Some use 90 degrees in the RF signal path and 90 degrees in AF. An interesting variant, TinySDR, apparently uses 45 degrees in RF, 90 degrees in LO, Polyakov mixers, and 90 degrees in AF (via software). (EDIT: I'm not sure if I interpreted the TinySDR phase shift scheme correctly - corrections would be welcome.)


One unusual combination that I haven't seen implemented is 90 degrees in the RF signal path and 90 degrees in the LO, with no phase shifting in the AF chain. This combination is described here but with no implementation: http://www.qsl.net/g3cwi/dc.html

Taking the example of TinySDR, if we replace the RF RC phase shift network (which, I think, may provide a 45 degree shift) with a transformer and amplitude-balancing pot as described earlier in this thread, then we could get wideband 90 degrees phase shift in the RF path, while needing only one adjustment as we tune up and down the band. Since the LO also produces 90-degree phase-shifted I/Q outputs, we mix (using normal mixers, not frequency-doubling Polyakov mixers) the 0 degree RF signal with the I and the 90 degree with the Q, add the AF outputs, and it seems we're ready to go for LSB reception. Swap the I/Q for USB. No AF phase shifting needed!

Seems like this could simplify the overall design. Thoughts?

EDIT: this won't work (and explains why I couldn't find any implementations of the idea). The reason is that we're essentially multiplying sin*sin and cos*cos in this scheme, both of which yield cosines. Therefore the AF signals generated in this way cannot be summed to eliminate one sideband.

Gaijin,

I have no doubt that the Kitchen design measured 0.3uV sensitivity. I would attribute that to the cascode rf stage which probably has a gain of 10-12dB over a simple grounded-gate preamp. That is enough to bring 3uV sensitivity down to 0.3uV. I seriously doubt that the PA2OHH rx exhibits 0.05uV sensitivity. That may be a typo or the result of generator leakage around the attenuators during measurement. Something on the order of 0.5uV is more likely.

In either case the improvement of sensitivity is a trade-off against reduced dynamic range. PA2OHH mentioned overload and you will notice that Kitchen does not use a cascode rf stage ahead of the detector in his more recent designs.

In the 3-30MHz spectrum sensitivities of less than 2uV really don't provide any benefit below 20MHz. There is a natural null in the atmospheric/galactic noise between 3-4MHz falling to about 1.8uV. The natural noise then rises back up to a peak around 10MHz before falling off to the 2uV region at 20MHz and less approaching 30MHz. Natural noise at the 10MHz peak can be as high as 20uV depending on time of day and geographical location. I have the 9uV figure reported by several sources as being the common atmospheric/galactic noise level for 7MHz at night. These figures are for rural areas. 'Quiet' suburban areas will generally be about 6dB worse and urban/business areas up to 20dB worse.

The upshot is that I don't worry overly much about sensitivity specs since the band noise will usually be much higher. Where I do pursue it is for WSPR/QRSS reception. WSPR can dig 28dB below the noise in a 2KHz passband to 100% decode a transmission. QRSS can go even deeper into the noise... possibly as low as -36dB in the same passband. Far and away below the threshold of audibility. A very good CW op can copy a signal up to 12dB below the noise in 2KHz bandwidth provided there is no QRM. The average CW op, in my experience, won't even try to copy your signal unless it's at least 6dB above the noise.

Anyhoo... an estimated 2-3uV sensitivity for my regen on 7MHz is good enough for me. I hear plenty of band noise at my location. On 3.5MHz the band noise is much lower and I could use a bit more audio gain following the detector... especially since my antenna is a cloudwarmer in that frequency range.

73,

'Bear' NH7SR

I quote "The upshot is that I don't worry overly much about sensitivity specs since the band noise will usually be much higher. "

That basically says it all right there!

One could use two regenerative stages to get the sensitivity down to say 0.01uV, but it would NOT do us any good, in most cases!

I was considering using homebrew diode ring mixers, but I noticed more recent designs use CMOS logic as you say. How much LO drive would be required by, say, a 74HC4066 used as a mixer? PA2OHH uses this mixer in his single-signal DC rx here: http://www.qsl.net/pa2ohh/98txrx.htm

I'm wondering if CMOS mixers might require less LO drive than the 7 dBm that diode ring mixers require.

Gaijin,

I suggest you take a look at the work Tasa has done with CMOS dc receivers.

http://yu1lm.qrpradio.com/

Extensively and professionally documented.

73,

'Bear' NH7SR

qrpbear, would it be possible to ask you to perform an experiment? Next time you hear a weak signal on the DC rx, that you can't hear on the regen, could you try running the regen just below threshold and use a separate and nearby beat oscillator (e.g. your DC rx LO) to generate the beat note in the regen? This increases apparent signal clarity when I try it, but I'm curious if it can actually make a previously inaudible signal audible. I'd like to try this myself, but don't have any DC receivers right now to compare with my regens.

Would a Schmitt trigger be appropriate for doing this kind of clipping?

The primary reason for using a Schmitt trigger is to deal with hysteresis, which is not a problem in your case. And a Schmitt trigger by its very nature, won't necessarily work with a low level input signal. I would use a high speed comparator instead.

Thanks Bob for the advice. Looks like my local suppliers stock 30-80 ns comparators for modest prices; 2.5 ns comparators start to get a little pricey.

On the other hand I have a bunch of 1N4148 diodes lying around. Looking at their spec sheet, it seems they can switch in a very respectable 4 ns (i.e. good up to 250 MHz). Would there be any foreseeable problems using diodes to do the clipping of the RF I/Q sine waves, using this sort of a circuit?
[/img]
Source: http://www.allaboutcircuits.com/vol_3/chpt_3/6.html

You can certainly use diodes. They're nice and simple. It's just a question of whether they're compatible with the signal levels that you're dealing with.