Noise Source

[dmowziz]

Hii...

Trying to measure the Noise Figure of an amplifier with homemade noise source but calculation not working out well

Using a spectrum Analyzer

My first question, what's ENR?
Is it the difference in noise level on Analyzer when noise source is off and on?


Thank you very much

[G6Tanuki]

"Equivalent Noise Resistance" is the noise-level generated by your amplifier equivalent to the noise generated by a 'perfect resistor' at the same ambient temperature.

https://en.wikipedia.org/wiki/Equiva...ise_resistance

[And remember that the calculations are thermally referenced to Kelvin, not Centigrade].

In times-past, whrn I was measuring noise-factor of receiver front-ends and tuning them for best behaviour we used a 'switched noise-source' which commutated at 1KHz between a thermionic-diode noise-source running with its heater being the emission-limiting factor, the amplifier, and a resistor in a tank of liquid Nitrogen. The switching resulted in a 1KHz component which was a reflection of the noise generated in the amplifier compared to that from the resistor and the diode noise-source.

Using this sort of gear you soon learn that an amplifier's maximum gain is rarely its lowest-noise-contribution to the overall signal-chain, which leads you to being able to sacrifice a few dB of all-outn gain if it gets you an 0.5dB noise-figure improvement.

"Tuning for maximum S-meter reading" means nothing if half of that meter-reading is noise from the amplifier!

[Radio Wrangler]

HP/Agilent/Keysight AN57-1 and AN57-2

-1 explains things generally. -2 goes into sources of error and how to minimise them.

In the 2000 re-write of them Bruce Erickson at HP wrote -1 and Ian White (GM3SEK) wrote -2. I was the reviewer.

David

[G0HZU_JMR]

R&S also provide an application note and this offers a step by step guide to uncorrected and corrected measurements.

https://scdn.rohde-schwarz.com/ur/pw...oiseFigure.pdf


What is handy here is that there is a lengthy worked example in the app note and the equations can be entered in excel and compared to the worked example to make sure you get everything in the spreadsheet correct.

[dmowziz]

Thanksss for the PDF

I will read and try again soon

But from page 6, it seems the ENR value has to be known prior it seems

[G0HZU_JMR]

Can you describe the component technology used for the homebrew noise source? I've used quite a few noise sources and also built and tested a few. Also, it would help to know the make/model of the spectrum analyser. Does it have a preamp option?
For many years I used a classic HP8566B spectrum analyser with a homebrew preamp to measure noise figure. I used a homemade noise source and it was quite a good one with very low source VSWR in both hot/cold states. My noise source had an ENR of about 13dB and this was quite stable. Despite this, the results I achieved were never that great in terms of repeatability especially when I tried using a second noise source that had a lower ENR.

My advice would be to start with a noise source that has an ENR up towards 15dB and try and measure amplifiers with a noise figure of at least 4dB. Otherwise the uncertainty/repeatability of the results you might achieve could become a bit disheartening.

[Radio Wrangler]

The ENR value represents the calibration factor of your noise source.

Whatever you're using as a receiver/detector has to be able to compare levels with nooise on and noise off, expressing the difference accurately in deciBels.

You need to know the ENR of your noise source at the frequency you are working. This is the amount of extra noise it generates, in dB above thermal noise at 290 Kelvin temperature.

Finally you need to know the temperature (in Kelvins) of the noise source (and its integral termination)

If any of these factors is wrong or missing, there will be an error in your result. The HP/Agilent/Keysight (now what name wioll they have next? ) webpage has an error allowance calculator. It won't cheer you up. Even with state of the art care over your equipment and calibration, Noise Figure is likely to be one of the least accurate measurements you'll ever make.

For measuring low noise devices with an NF of 1dB or below, a 6K error in temperature will give around the same error as a freshly calibrated noise source just back from a national standards lab.

There is one elephant in the room... why do people fight so hard to get noise figures below 1dB. Surely you'd expect it to make less than 1dB difference?

The answer lies in the fine print. People will glibly say that noise figure is the degradation in the S/N ratio od a signal coming out of the DUT compared to the S/N ratio going into it.

Just a minute, I wasn't putting any noise in added to the signal when the DUT is doing its job. There is no defined S/N at teh input.

Well, ther is, via the back door. It's ASSUMED (oh dear!) that any signal from a resistive source impedance will carry thermal noise from that impedance. But this depends on temperature... and here is the problem. They ASSUMED the temperature would be a nice comfortable 290K. OK in the lab.

But if your antenna sees a lot of sky in its view, the sky is cooler than the Earth, and in the gaps without many stars, it goes down to 2.8K This is much colder than 290K.

So for something like a satellite receiver, you want your receiver noise to be as little above 2.8K as you can get. Measured in the lab with gear at and calibrated for 290K, then the noise from a good DUT will only be a tiny amount more than the thermal noise from the source at 290K, Hence you want to see a very small Noise Figure under these circumstances, well less than 1dB if you can get it.

On the other hand, the HF bands have a lot of natural noise, well above 290K, and noise figures of 10dB and more are of little concern.

David

[Radio Wrangler]

The variation of VSWR with noise on/off can be a dominant error. Your DUT will respond to the changing impedance with changing gain, and this will distort the slight difference you measure in noise on/off with the DUT compared the the difference with the noise source bypassing the DUT. So you get quite large errors on low noise devices.

The solution is to use a noise source in the 5 dB region for very low noise DUTs, and to use the more normal 15dB noise source for less demanding DUTs. Each will give you better accuracy in its home ground.

The 5dB noise sources are essentially 15dB ones with more attenuation on their output to better swamp the impedance changes of the noise diode as it switches on and off. You can add an attenuator to a 15dB noise source, but you have to calibrate them together as a pair.

Note that a calibrated noise source has different calibration values and uncertainties at each calibration frequency.

You normally get calibration points at 10MHz, 100, and then integer numbers of GHz from major cal labs like the UK National Physical Laboratory or the US National Institute of Science and Technology. This is because they have to use different test rigs and noise sources for each of those frequencies. They will sell you calibrations at other frequency points, but as they only have rigs for the frequencies I listed, anything else is done by interpolation and can be a bit less certain.

OK, so you aren't going to get your noise source calibrated by the big boys, but as all other calibration routes traces back to them, you are still tied to their cardinal frequencies, and interpolation for anything else. Also each generation your source is removed from the top level thermal standards, some more uncertainty creeps in.

Back in 2000, about to bring out a new family of noise sources, the issue of calibration came up. Talking to NIST, who we'd always previously used as the source of our calibration standards elicited a bit of an Oooo-err response. For political reasons, their budgets had been cut and some maintenance hadn't been done out in Boulder, Colorado and a roof had sprung a leak, The water damaged their noise source standards. They'd done repairs and replaced things, but it was going to take years of getting proof of stable performance, involving swapping comparison noise sources with other standards labs around the world. Consequently the uncertainty figures they could offer on their calibrations were not as good as they used to be. However, in the UK, with all our rainfall, the NPL reference noise sources were in fine shape and NPL could offer much tighter uncertainties. We shopped around and got quoted uncertainties from other national labs. One offered uncertainties so low they weren't considered believable. I won't say where.

A company called Noise/Com used to sell noise diodes to amateurs and would also help calibrate people's finished noise sources. They also sold finished sources to professional standards. They bought one of our noise source calibration rigs. Nice to know that the transfer standard noise source in it, and the comparison receiver, were HP ones, so the noise source source cals on their new products ought to be directly comparable to the HP ones.





David

[dmowziz]

Thanks David and Jeremy

@Professor Jeremy,
Using thru hole components (Dead bug on copper clad). BFR96 Transistor

The circuit is the first image followed by a 50 Ohms 15 dB pad. Thinking of using one amplifier and not 2.
Siglent SVA1015X (It has a preamplifier, it was turned off before)

https://vintage-radio.net/forum/showthread.php?t=97043
Making a pcb with smd to make this but right now my circuit is the first image with thru hole.

The 2nd image is output of noise source with it off (the "glitches" are from the sma cable : Please how to correct?)

The 3rd image is output of the noise source with it on

So is the ENR (110.40 - 56.77) dB ?

Thankss... @Jeremy Still going through the R&S Pdf. Thanks

[Craig Sawyers]

The idea of using a reverse biased diode as a noise source was originally patented by Pye in 1961. One of the authors was my old friend and mentor Gordon Edge (RIP).

Attached

Craig

[dmowziz]

Nicee, being an inventor. Hope you learned well



Post #9
"The 2nd image is output of noise source with it off (the "glitches" are from the sma cable : Please how to correct?)"
This occurs when the attenuator is turned off

[Radio Wrangler]

Virtually any diode type? No. The diode needs to go into avalanche breakdown with the available supply voltage. This means an avalanche diode with a specified breakdown voltage.

One secret: Zener diodes only exist up to a few volts, every 'zener' above this is actually an avalanche voltage regulator diode. It matters not to circuit designers, but the internal mechanism is different.

The avalanche diode needs resistance in the DC path, or a current source, to control the DC current and keep it stable over temperature.

Voltage regulator diodes ("zeners") are usually designed to handle a few hundred mW, so they are big and suffer high capacitance which limits the frequency range of their noise.

A trick is to use the base-emitter junction of a GHz type transistor. These break down in the 5-8v region most often and have the low capacitance needed for VHF-UHF use. Don't connect the collector. Keep the avalanche current low or you will suffer progressive degradation.

That said, a diode in avalanche should make quite enough noise to not need any amplification at all, and still allow for some attenuation to dilute the variation in impedance and still give you about 15dB ENR.

You really really want to avoid having RF amplification in the noise path. It varies in gain with temperature and there goes the accuracy of your calibration.

The N4000 series noise sources are a diode with attenuator and DC block. Part of the attenuator serves as the DC bias feed in a way more broadband than any choke-T can do.
The current source for the diode is digitally programmable from the instrument using the source. The cal table is in EEPOM, and there is a digital thermometer IC to measure the 'cold' temperature. As a bonus the thermometer has ADC inputs which check diagnostic voltages.

Stripped of the fripperies it's a diode, resistive attenuator with DC feed and block powered from a constant current source.

David

David

[kalee20]

Originally Posted by G6Tanuki

"Equivalent Noise Resistance" is the noise-level generated by your amplifier equivalent to the noise generated by a 'perfect resistor' at the same ambient temperature.

I'm no guru on noise, but... that doesn't look right to me!

Resistors produce noise which fades to zero as temperature drops. But amplifiers have shot noise (eg partition noise in pentodes, etc) which is independent of temperature - and other excess noise sources.

That means that an amplifier can never be represented by an equivalent resistor at the same temperature - the 'equivalent resistor' value would have to be specified at a stated temperature.

[Craig Sawyers]

David is entirely correct. Equivalent noise resistance and noise temperature are familiar concept to anyone involved with sensitive receiver design. My somewhat ancient reference on this is Signal Processing, Modulation and Noise by JA Betts.

Craig

[kalee20]

I know ENR is a familiar concept, but if David (G6Tanuki) is 'entirely correct' then our amplifier's noise would go to zero as ambient temperature went to zero, whatever the 'equivalent resistor' happened to be.

And while the Johnson noise contribution does, shot noise doesn't. So, there's something to be reconciled.

[G0HZU_JMR]

ENR is Excess Noise Ratio.

It's easy to get in a muddle with the definition of ENR but most publications now seem to agree that it is defined in terms of noise temperatures and it is the ratio of:

(noise source hot temperature - noise source cold temperature)/(reference temperature)

where the reference temperature is usually 290K.

ENR (dB) =10*log((Thot - Tcold)/290K)

[dmowziz]

Originally Posted by G0HZU_JMR

ENR is Excess Noise Ratio.

It's easy to get in a muddle with the definition of ENR but most publications now seem to agree that it is defined in terms of noise temperatures and it is the ratio of:

(noise source hot temperature - noise source cold temperature)/(reference temperature)

where the reference temperature is usually 290K.

ENR (dB) =10*log((Thot - Tcold)/290K)

Please post #9 is not correct then?

[G0HZU_JMR]

Sadly, no it's not correct. Note that the glitches are probably caused by signal pickup from VHF FM broadcast stations.

[G0HZU_JMR]

That noise source has an ENR that is way too high for receiver noise figure testing.

Try making the noise source recommended by John in post #5 in the thread below. This uses a BAT-17 Schottky diode. It will need a decent attenuator after it to improve the source match and to reduce the ENR down to something suitable for receiver noise figure testing.

https://www.vintage-radio.net/forum/...d.php?t=163568

[Radio Wrangler]

Still got my copy of Betts from my student days.

If you look at noise performance and measurement in terms of equivalent temperature, a number of things suddenly become clear and obvious.

Noise figure because of a built-in reference temperature which is not always appropriate, can be obscure and counter intuitive in some cases. Being expressed in dB, noise figure looks straight-forward, which it isn't.

David