The Beehive

[ppppenguin]

While building an attenuator is an instructive exercise I'd be very tempted to look for an old Hatfield unit at a reasonable price. I have a couple of 2115 which are 75 ohm. I think the 2105 is the 50 ohm version: https://picclick.co.uk/Hatfield-Attenuat...38795.html

[BusyBee]

I agree that it may be a good idea but for a number of reasons I find it better to build one. It's sometimes better the devil you know rather than one you don't as in an old piece of equipment which is an unknown quantity. I don't have the room for much anyway so am very limited. That means that, as much as I would like, a stack of equipment is not possible. Many of the things I make, I end up dismatling if I find I am not using them just to make space. Little units like the one I have just made I can usually find room for.

As to the making - With the noise generator I am now looking into I see that the ones available from the far east are fairly cheap. If I were just wanting one to use it would be more sensible to buy one but with me it's far more often the thrill of the chase rather than the end result, if you get my meaning. It keeps my mind exploring and learning new things. There are times I cut my losses and just go out an buy things but it's often when I find I have bitten off a bit too much for me to readily handle.

Tracy

[Mike Watterson]

My noise generator is a zener as noise source, R C high pass filter and a MAR6 amp, optimised for about 20 MHz to 1.2 GHz.

[BusyBee]

Thanks Mike. It seems like I am looking along similar lines, just concentrating on a different frequency band. As I am looking to lower frequencies I thought it may be prudent to cut the gain at the higher to limit the possibility of spurious oscillations etc.

[Mike Watterson]

The power is the issue. You need to limit the bandwidth. I'd need a 10W amp to get a decent level from mine, the wideband nature means it's quite low level, but I used to use it to see filter shapes with an HP141T spectrum analyser. I never use my now because my Rigol spectrum does 19 kHz to 1.5GHz and has a tracking generator (can be a signal generator if span = 0). Though there are other uses for a noise source.

A zener is quite noisy. Also good is a high value carbon composite resistor, with enough current to warm it.

Actually even a perfect resistor is an ideal noise source!

Thermal noise in an ideal resistor is approximately white, meaning that its power spectral density is nearly constant throughout the frequency spectrum (Figure 2). When limited to a finite bandwidth and viewed in the time domain (as sketched in Figure 1), thermal noise has a nearly Gaussian amplitude distribution

https://en.wikipedia.org/wiki/Johnson–Nyquist_noise
This is why low noise circuits need to be cool and have low value resistors, thus a 10 M ohm grid bias 100x noise of 100 k Ohm grid bias, and the grid resistors want to be at a cool place on the chassis / PCB, as the noise is amplified by the valve. The resistor noise is proportional to resistance x temperature x desired bandwidth.

So at room temperature, 10 MΩ provides about 3 mV of RMS noise for 20 kHz bandwidth and 100 kΩ thus 33 uV RMS noise. Note that 20 °C to 200°C isn't x10. It's 293 K to 473 K, only 1.6x worse, so the component placement and temperature is less critical than resistance value. Maybe only radio telescope LNAs need to be cooled, but a satellite dish certainly doesn't want to be reflective to IR at the equinoxes when there is more RF noise picked up from the sun as it passes behind the satellite.

An amp that does 100 mW max (= -10 dBW) without distortion can only do that with a single tone (frequency). If you have two equal signals, the level per tone is less. The white noise total power N integrated over a bandwidth B is N = BNo
https://en.wikipedia.org/wiki/Noise_spectral_density, thus it's challenging to get a significant signal level the wider band the noise source is.

[BusyBee]

Yes, I have come across issues to be aware of with the bandwidth. With the spectrum analyser I have, I have seen a video warning of input overload because of the total level rather than just that of one point in the spectrum. Apparently it's something to be aware of when buying a secondhand analyser as faulty input circuitry is not uncommon.

I did wonder about a resistor noise source. I did think too, although only briefly, about a valve based noise source too. I do have a 'noise diode' although still intend it for it's original purpose when I get around to it. Using one as a noise source would not likely be as good as it may seem as I would need to under run it to get good life. It did get me thinking about other diodes though. I wonder about a rod pentode in triode mode with g1 used as a level control? Still I am digressing a bit.

For interest, I have found references to a company who do give a bit of information, much of what you give, in their articles and videos. Apparently their diodes are very expensive though, even if you could get them in small quantities: https://noisecom.com/products

Yesterday I did get the noise diode I have, which is a purpose produced zener, onto a board with it's bias circuitry. I also produced another with RF amplifier. Both manhattan style so should be good in the frequency ranges I am likely to use. I hope to get round to some testing later in the week

I am still wondering about filtering though; at this point whether to approach this seperately as my needs will likely be varying but fairly narrow bandwidth specific.

[Mike Watterson]

Neons are also quite noisy, but likely not very wide band.

A lowest capacitance zener (typically lowest power) has the better HF response.

Some filtering (most?) should be between the noise source and the amplifier to minimise distortion and maximise output.

[Mike Watterson]

Also a peak reading or RMS AC meter will give a much higher reading than the level seen on a spectrum analyser with wide band noise.

There are two main kinds of spectrum analyser. The analogue kind is really a superhet radio with the local oscillator swept by the X timebase of the scope. A purely digital kind uses an ADC and DSP (FFT typically). Some are hybrids of the two approaches. At least with a digital display rather than CRT you don't have a problem with the slow sweeps needed for accurate views of filters. That's why most serious CRT based spectrum analysers used very expensive "storage" CRTs. My HP141T was nice with 4 plugins for HF to 10 GHz, but the input on those is easily damaged and the "wrong" settings can give misleading results. Sadly the IF or part of PSU or both died. The Rigol digital model had an initially harder to understand interface, but I've mostly mastered it now. I can go above the 1.5 GHz using various LNBs (MMDS, C Band, X Band, Ku Band and Ka Band), if I really it, but I don't now, though I also had a HP141T in work and needed up to 10.5 GHz

[BusyBee]

Neons are also quite noisy, but likely not very wide band.

A lowest capacitance zener (typically lowest power) has the better HF response.

Some filtering (most?) should be between the noise source and the amplifier to minimise distortion and maximise output.

I did come across the use of gas filled tubes. There were ones produced which had this as one of their uses. I have some neons but, apart from what you say, the operating voltage would be high.

With my initial designs I have a two stage amplifier following several similar designs I came across online. I was thinking of the filtering between the two stages. Where the signal is still low but not too low as to be possibly lost, to a great extent, in the filter.

For my first tests of the noise source though I will likely seperate things as I do have one of those 30dB wideband amp modules that come cheaply from China. With the noise source connected into it I would expect some kind of signal which can be measured out. It avoids dealing with two unknown quantities at once.

Also a peak reading or RMS AC meter will give a much higher reading than the level seen on a spectrum analyser with wide band noise.

I will see what I get.

Tracy

[BusyBee]

A bit of an update. I have been busy with various things and have got some results.

Shown here are the little RF amplifier board using the MAX2611 I mentioned earlier. I actually constructed two of them as it is likely both would be needed. During the initial experiments I have been conducting using my TinySA Ultra I have been testing a number of oscillators I have as regards to purity of signal. Using one of these I tested both amplifiers. They appear to be working as I would expect although a lash up of connections does not help with stray pickup. The circuits shown have temporary connections for testing and are not in their final setup.

I also show the noise generator board. It is supplied via a zener setup regulating it's feed voltage to 30V, rather than perhaps having an indeterminate supply providing input. On test the board appears to output wideband noise but I haven't tested this extensively.

As implied earlier, on looking at signals I have noticed the results obtained by using a lash-up of connections. The indication is that below 1MHz or so things are not too bad but above can be a bit of a nightmare. If there is one thing the TinySA Ultra is telling me, it is just how poor my testing setup is for anything above audio. It's beginning to pay for itself already