|
Components and Circuits For discussions about component types, alternatives and availability, circuit configurations and modifications etc. Discussions here should be of a general nature and not about specific sets. |
|
Thread Tools |
19th May 2022, 4:57 pm | #21 | |
Pentode
Join Date: Nov 2021
Location: Toronto, Canada
Posts: 131
|
Re: Noise Source
Quote:
Thanks. Doing it, with a BAT42 Right now, trying to measure NF of amplifiers |
|
19th May 2022, 5:25 pm | #22 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,903
|
Re: Noise Source
You can use a commercial microwave grade coaxial attenuator if you can pick one up at a rally. This will give you good impedance and good flatness to quite high frequencies. You still need to do a decent job on the diode and decoupler, but this is a short cut for the rest. Surface mount with low strays is a further help.
David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
19th May 2022, 5:27 pm | #23 |
Dekatron
Join Date: Nov 2010
Location: Oxford, UK.
Posts: 4,999
|
Re: Noise Source
John Betts was at Southampton University, so I had the pleasure of being lectured to by the guy. He had a clear way of describing everything we have been discussing.
He had a racing green TR5. Which he drove with gusto - too much it turned out, and ended up discussing communications beyond the pearly gates. Sometime in the late 70's. By 1988 the University was offering a memorial prize for communications excellence, and still does. Adam Malpass John Betts. Never knew he was burdened with Malpass in his name. He kept that very quiet! Craig
__________________
Doomed for a certain term to walk the night |
19th May 2022, 7:01 pm | #24 |
Heptode
Join Date: Jan 2011
Location: Croydon, London, UK.
Posts: 773
|
Re: Noise Source
I don't think a BAT42 will work in John's circuit. It has a reverse breakdown voltage of 30V minimum, compared with a maximum continuous reverse voltage of 4V for the BAT17.
Paula |
19th May 2022, 11:44 pm | #25 |
Pentode
Join Date: Nov 2021
Location: Toronto, Canada
Posts: 131
|
Re: Noise Source
|
20th May 2022, 6:58 am | #26 |
Pentode
Join Date: Nov 2021
Location: Toronto, Canada
Posts: 131
|
Re: Noise Source
Used a BFR96 Transistor in the circuit (Base + Collector to ground), Vebo was 7.2V, no noticeable noise output
1 out of 2 1n4733 5.1V Zeners "Worked" but the noise peaked at about 5.1V. Increasing the voltage above 5.1 reduced the noise output What else makes the BAT17 work? What else is there to look for in a diode that will work Meanwhile, trying this npn noise circuit here : http://www.w1ghz.org/noise/noise99.pdf @G0HZU, please is the ENR definition in post 9 correct? Thanks |
20th May 2022, 8:59 am | #27 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,903
|
Re: Noise Source
There are two major flaws in the circuit in post No 9.
"Resulting in a strong noise source suitable for measurements" Nope! It should read "Resulting in a noise source far too strong to be suitable for the measurement of normal low-noise equipment and components" The second flaw lies in the two amplifiers (whose gain will wander with temperature) and also in the imprecise control of diode current. This is not going to be a stable enough source to be worth calibrating to the resolution needed to make noise figure measurements, even if you attenuate the output down to the necessary level. As you say, the amount of noise from a zener changes very rapidly with voltage. This is the diode telling you that driving it from a voltage source isn't a good idea. Constant current source drive has been mentioned a few times. You've overlooked an important clue. What frequency range do you want your noise source to work over? David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
20th May 2022, 9:49 am | #28 |
Dekatron
Join Date: Nov 2010
Location: Oxford, UK.
Posts: 4,999
|
Re: Noise Source
The Pye 1961 patent I attached mentions the difference between Zener and Avalanche, says Avalanche is what you need, that you can use a transistor instead of an Avalanche diode, and constant current is the preferred mode. They say that a breakdown voltage of 6-7V is preferred.
What's that - 61 years ago? Craig
__________________
Doomed for a certain term to walk the night |
20th May 2022, 12:53 pm | #29 |
Pentode
Join Date: Nov 2021
Location: Toronto, Canada
Posts: 131
|
Re: Noise Source
I see : constant current
Thankssssss @David : 1 MHz to 0.15 GHz ( ) @Craig , thanks! overlooked |
20th May 2022, 1:08 pm | #30 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,903
|
Re: Noise Source
150MHz is within home construction capabilities. Base-emitter junction of an RF transistor. Surface mount will help.
1MHz is a problem. The national standards organisations only go down to 10MHz so any calibration below that isn't going to be traceable to provable standards. There is another way, but 150MHz might be a limitation. Agilent/Keysight have a US patent on an FPGA driving a DAC and filter to make a calibrate-able noise source compatible with Y factor measurements to lower frequencies. It looks like an arbitrary waveform generator driving a lot of attenuation, but there are some quirks needed to make the cal provable and you'd need to get at the arb's software. Can't remember the patent number and it's got my name on it... David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
20th May 2022, 4:57 pm | #31 |
Dekatron
Join Date: Nov 2010
Location: Oxford, UK.
Posts: 4,999
|
Re: Noise Source
Any thoughts on long cycle length pseudo random noise generation David? Easy to do at audio frequencies of course.
Craig
__________________
Doomed for a certain term to walk the night |
20th May 2022, 6:30 pm | #32 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,903
|
Re: Noise Source
That's what the patent is on, and on using other digitally generated signals to allow the DAC etc to be characterised so that noise calibration can be mathematically derived.
I spent some years working on instruments for long PRBS test datastreams for evaluating error creation in higher order PCM telecoms systems. The bit I was interested in was in jitter and jitter transfer functions. From that area I moved into the RF/Microwave division. I got around! As part of the Master's there was a lot of material on the mathematical analyses of PRBS properties, lots of modulo-2 arithmetic based algebra, as per Solomon Golomb's book. David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
22nd May 2022, 2:08 pm | #33 | |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Noise Source
Quote:
Ideally, the noise measuring receiver itself should have a noise figure of less than 5dB. This is what many commercial noise figure receivers achieve. I think your Siglent spectrum analyser has a noise figure of about 13dB with the preamp on. This isn't low enough if you want to be able to make good noise figure measurements. The other issue with your Siglent analyser is that its internal preamp could be prone to overload damage if you start measuring amplifiers that can deliver >15dBm output power if they go unstable. My advice would be to make an external wideband preamp for the Siglent analyser. A good compromise would be to have a preamp with about 25dB gain and 3.5dB noise figure, then follow this with a protection limiter, then select 10dB internal attenuation in the Siglent analyser and then turn on the Siglent internal preamp. The limiter and the 10dB attenuator should protect the Siglent analyser from overload damage. I think the overall system noise figure of the above setup will be about 5dB. Note also that the noise floor of your analyser will degrade at lower frequencies due to noise on the internal local oscillator. So you may find the analyser is still too noisy down at 1MHz. I'm not sure how low the noise floor is on your analyser at 1MHz. Usually the typical performance is much better than the datasheet in this respect.
__________________
Regards, Jeremy G0HZU |
|
22nd May 2022, 2:33 pm | #34 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Noise Source
See below for some plots of my old homebrew external preamp box. This was what I used for many years with my Advantest TR4172 and HP 8566B spectrum analysers for noise figure measurements. I'm lucky to own a newer analyser now that has an internal preamp and it can also mimic a noise figure analyser as it has the relevant hardware and software options fitted. So this preamp rarely gets used.
It's old and a bit battle scarred but this preamp was designed for about 3dB noise figure and flat 32dB gain from 1MHz to about 1GHz. It has been repaired and tweaked several times over the years and it does work very well in my opinion. A key requirement for a noise figure preamp is that it should have low input VSWR and high reverse isolation. This preamp has ultra low VSWR as you can see in the plots below. The noise figure is about 2.7dB across 10MHz to 1GHz. The reverse isolation isn't shown below but it is about 55-60dB. It isn't necessary to have something this good but you should be able to make an analyser preamp using a couple of low cost MMIC amps from Minicircuits. In your case I think you only need about 25dB gain in the preamp. I needed more because the HP8566B doesn't have a preamp. The Advantest TR4172 does have a preamp but its noise figure was too high. I also wanted very low input VSWR for the preamp as this minimises mismatch uncertainty effects. Note that the analyser source power is set really low at about -35dBm to prevent any compression effects. So the traces are a bit noisy.
__________________
Regards, Jeremy G0HZU |
22nd May 2022, 3:17 pm | #35 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,903
|
Re: Noise Source
The Agilent NFA has TWO different preamps, It uses Silicon below 500MHz, and switches to GaAs above 500MHz.
Low noise GaAs devices may offer the lowest NF at UHF, but they tend to give rising noise at lower frequencies. Plain old Silicon can win at lower frequencies. Jeremy's preamp is rather good. The VSWR will help. 26dB gain is about right in front of a spectrum analyser. David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
22nd May 2022, 9:56 pm | #36 | ||
Pentode
Join Date: Nov 2021
Location: Toronto, Canada
Posts: 131
|
Re: Noise Source
Thanks David, Craig and Jeremy.
I initially thought of having a simple way to estimate NF of Norton amplifier and other amplifiers. But seems, what is worth learning should be worth learning well. I made the Noise source here (Fig. 5) Not current source http://www.w1ghz.org/noise/noise99.pdf On the Spectrum Anlayser, the noise rises by 9dB (including 15 dB pad). Thru hole, dead bug. Noise peaks at 7.5Vdc. Quote:
Yes I noticed this for an amplifier yesterday and stopped. Please how did you get 13 dB ? From the 1st image, 10 MHz - 1 GHz, 18 dB ? Is this what it means? (NF of Pre amp) not sure Quote:
https://www.farnell.com/datasheets/2342538.pdf Also, don't know what that limiter is Meanwhile, I got a cheap laser printer last week to do Pcb for SMD soldering. Starting with this Thankssss Any suggestion/advice, highly welcome |
||
23rd May 2022, 4:48 pm | #37 |
Dekatron
Join Date: Sep 2010
Location: Cheltenham, Gloucestershire, UK.
Posts: 3,077
|
Re: Noise Source
I guessed the noise figure based on the datasheet for the SSA3021X plus model as it quotes a displayed average noise level (DANL) of -161dBm/Hz.
-161- (-174) = 13dB Also your noise floor plot showed a DANL of about -110.4dBm in a 100kHz bandwidth (= 50dBHz) so this would be -110.4dBm - 50 = -160.4dBm/Hz although I haven't included any corrections in this for log amp or detector type. It's best to make a two stage amplifier rather than try and use that NXP part. The NXP part also has poor input VSWR so it isn't really suitable for use as a noise figure analyser preamp. The limiter could be something as simple as some back to back diodes. Avago application note 1050 shows a few diode limiter designs using low cost PIN diodes. The alternative is to use back to back 1N4148 silicon diodes although the 1N4148 diodes will only provide effective limiting up to about 500MHz.
__________________
Regards, Jeremy G0HZU |