|
I want to add this post to Martin's MLA-30 thread, but it is locked. Hence, my starting this new topic. I know the MLA-30+ has been done to death where evaluation is concerned, but I can't resist jumping into the fray! So, here's the Thanks for reading!?
73, Chin-Leong Lim, 9W2LC.
|
Hi Leong,
?
Good work, but the LZ1AQ IOP3 looks far too low.
?
It is quoted as being
?
1 dB output compression point(9) +19dBm (5.6 V p-p), equal to +125 dB(uV/m) at input
Second harmonic OIP2(7) +82dBm to +105dBm
Third harmonic OIP3(8) +41dBm to +42dBm
?
Looking back at my results, the best I could manage to measure was:-
?
OIP2 +74dBm
OIP3+42dBm
?
My normal sanity check is to check that the IOP3 is typically 10 to 15dB higher than the P1dB value.
?
What's your test setup, and how do validate it ?
?
Regards,
?
Martin
|
This is what a LZ1AQ amp is with 2SC5551 transistors.
?
?The Improved LZ1AQ at 13.8VDC @120mA has an IMD of (OIP2) 1 MHz +98.4 dBm, 7 MHz +93.8 dBm. (OIP3) 2 MHz +42.2 dBm, 5 MHz +40.35 dBm
?
Everett N4CY
In a message dated 10/6/2024 12:52:06 PM Central Daylight Time, martin_ehrenfried@... writes:
?
Hi Leong,
?
Good work, but the LZ1AQ IOP3 looks far too low.
?
It is quoted as being
?
1 dB output compression point(9) +19dBm (5.6 V p-p), equal to +125 dB(uV/m) at input
Second harmonic OIP2(7) +82dBm to +105dBm
Third harmonic OIP3(8) +41dBm to +42dBm
?
Looking back at my results, the best I could manage to measure was:-
?
OIP2 +74dBm
OIP3+42dBm
?
My normal sanity check is to check that the IOP3 is typically 10 to 15dB higher than the P1dB value.
?
What's your test setup, and how do validate it ?
?
Regards,
?
Martin
|
The is the original built as per circuit on line using 2N2222 transistors
?
LZ1AQ Loop amplifier, with 2N2222 Transistors. The Orginial LZ1AQ at 13.8VDC @130mA has an IMD of (OIP2) 1 MHz +77.1 dBm, 7 MHz +85.2 dBm. (OIP3) 2 MHz +38.5 dBm, 5 MHz +37.5.35 dBm
In a message dated 10/6/2024 1:53:20 PM Central Daylight Time, everettsharp@... writes:
?
This is what a LZ1AQ amp is with 2SC5551 transistors.
?
?The Improved LZ1AQ at 13.8VDC @120mA has an IMD of (OIP2) 1 MHz +98.4 dBm, 7 MHz +93.8 dBm. (OIP3) 2 MHz +42.2 dBm, 5 MHz +40.35 dBm
?
Everett N4CY
In a message dated 10/6/2024 12:52:06 PM Central Daylight Time, martin_ehrenfried@... writes:
?
Hi Leong,
?
Good work, but the LZ1AQ IOP3 looks far too low.
?
It is quoted as being
?
1 dB output compression point(9) +19dBm (5.6 V p-p), equal to +125 dB(uV/m) at input
Second harmonic OIP2(7) +82dBm to +105dBm
Third harmonic OIP3(8) +41dBm to +42dBm
?
Looking back at my results, the best I could manage to measure was:-
?
OIP2 +74dBm
OIP3+42dBm
?
My normal sanity check is to check that the IOP3 is typically 10 to 15dB higher than the P1dB value.
?
What's your test setup, and how do validate it ?
?
Regards,
?
Martin
|
On Sun, Oct 6, 2024 at 10:52 AM, Martin - Southwest UK wrote:
Good work, but the LZ1AQ IOP3 looks far too low.
Hi Martin,
Sharp eyes! ?
You r correct, the LZ1AQ OIP3 looks too small. I think the problem lies in the large variation with freq - the OIP3 varies from 16 dBm at 5 MHz to 30 dBm at 20 MHz. In contrast, the competition - MLA-30+, M0AYF & PA0FRI have flatter OIP3 variation.
?
My normal sanity check is to check that the IOP3 is typically 10 to 15dB higher than the P1dB value.
The |OIP3 - P1dB| in the table looks deceptively small because it is the worst case value, taken at different frequencies, i.e. P1dB is taken at 20 MHz, while OIP3 is at 5 MHz. If the OIP3-P1dB is taken at the same frequency, say 20 MHz, the delta is a more believable 15 dB ( see this graph). When I find the time, I will create another graph showing |OIP3-P1dB| vs. freq - hopefully, it will clarify things. ?
What's your test setup, and how do validate it ?
FY6900 function gen, Minicircuits ZSC-2-4 hybrid combiner, & TinySA. I admit I haven't validated the setup.
Hi Everett, thanks for sharing your measurement results.
?
73, Chin-Leong Lim, 9W2LC.
?
|
Hi Martin,
73, Chin-Leong Lim, 9W2LC
?
|
Hi Leong,
?
OK thanks for that.
?
From experience, I don't think you will be able to achieve the required performance of your test setup without building fundamental frequency notch filters to place on you analyser input.
?
Ideally you need to be able to measure signal levels below -100dBm, in the presence of test signals at 0dBm or more. Even with a high end SA, this is asking a lot.
?
Steve and Everett wrote this useful practical guide for Siglent.
?
https://www.siglenteu.com/application-note/inter-modulation-distortion-imd-testing/
?
I think the variation you are seeing is due to the external and internal IP generation aiding and abetting each other as the phase and amplitude relationships vary with frequency.
?
In the old days of analogue TV transmitters, we used to deliberately generate IP products at low level, and inject them into the RF transmission chain, to null out IP's that were being produced in the high power amplifier stages. We called this feed-forward correction.
?
Regards,
?
Martin
?
|
On Sun, Oct 6, 2024 at 07:32 PM, <biastee@...> wrote:
?
?
What's your test setup, and how do validate it ?
FY6900 function gen, Minicircuits ZSC-2-4 hybrid combiner, & TinySA. I admit I haven't validated the setup.
?
Hi Chin-Leon Lim Measurements of high IP3 values are challenging. It is very likely that the measured IM products will be created in Tiny SA itself, not in the DUT. Your limitation is the internal IP of the Tiny SA
?
regards, Fred
|
Hi Martin & Fred,
?
Thanks for the article links. I will dig into them later.
?
At this moment, I don't have notch filters for the 5-30 MHz range that I want to measure. To quickly verify whether the IP is internal to the TinySA, I will try to increase its attenuation to see if the displayed IMD reduces.?
What I find hard to understand is why the measured OIP3 is 30 dBm at 20 MHz ( see graph), but progressively drops as the frequency pair is lowered, and finally reaching 16 dBm at 5 MHz. My expectation is internally generated IMD should be quite constant over frequency. On the other hand, if the external & internal IMD are combining constructively and destructively at different frequencies, as Martin has speculated, I would expect the OIP3 trace to be wavy. Or did I miss something? Anyway, more investigation is required to understand the setup's eccentricities.
At the project onset, I thought it would take 1 month max to build a working active loop, but now it has already taken a year of my life ?. What a deep rabbit hole!
?
73, Chin-Leong Lim, 9W2LC
|
Hi Chin-Leong Lim ( the first time that i got right)
?
Besides the TinySA, whose own IP3 is 17 dBm at best (thats your absolute limitation), the Signal generators Output Stages and even the Splitter with its Ferrite nonlinearity are also a source of Intermodulation which limit your measurement dynamics. Again, IP measurement of high IPs is demanding, dont't fall in your own traps.
?
regards, Fred
|
I would agree with Fred.
?
Despite what many articles and videos on the internet may suggest, including many professional ones, it is not easy to make practical and repeatable IMD measurements, especially at the IMD levels we are interested in.
?
I've struggled with this, and have my own built band pass and notch filters and crystal oscillator based test sources. But even things like dirty connectors, especially things like BNC test leads, can cause results to vary considerably.
?
The first step is to check that your IMD test rig is capable of producing results that are at least 10dB better than the device you are testing. Even this isn't really good enough, but it's something to aim for. Then try measuring known devices, such as the Mini-Circuits PGA-103+ or GALI-74+.
?
I've always found these perform pretty closely to their published specification, and I tend to use one as a sanity check, before I try measuring something else.
?
Regards,
?
Martin
?
?
On Mon, Oct 7, 2024 at 04:19 PM, Fred M wrote:
the Signal generators Output Stages and even the Splitter with its Ferrite nonlinearity are also a source of Intermodulation which limit your measurement dynamics. Again, IP measurement of high IPs is demanding, dont't fall in your own traps.
|
?
Incidentally, I've just been comparing IMD measurements using a Rigol DSA815 and a TinySA Ultra.
?
I've been using my own test source and notch filters, and providing the notched test tones are kept to a level of -60dBm or lower, the TinySA Ultra actually produced better results.
?
Every time I make comparisons against my other Spectrum Analysers, including HP 8500 series, I gain more respect for the TinySA Ultra.
?
Regards,
?
Martin
?
On Mon, Oct 7, 2024 at 04:19 PM, Fred M wrote:
Besides the TinySA, whose own IP3 is 17 dBm at best
|
On Sat, Oct 12, 2024 at 07:28 AM, Martin - Southwest UK wrote:
Every time I make comparisons against my other Spectrum Analysers, including HP 8500 series, I gain more respect for the TinySA Ultra.
Definitely, but without? notch filters for the original 2-tone signals the tiny-SA will become overloaded. It's own IP3 is about +17 dBm.
?
regards
Fred
|
Absolutely, it's no good unless you can notch out the fundamental tones.
?
My own notches provide about 40dB rejection, which is just about good enough.
?
Regards,
?
Martin
?
On Sat, Oct 12, 2024 at 03:21 PM, Fred M wrote:
Definitely, but without? notch filters for the original 2-tone signals the tiny-SA will become overloaded. It's own IP3 is about +17 dBm.
|
There is no way to run a proper IMD without a band stop filter that rejects the 2 tones and a band pass filter for each of the tones.
?
Everett N4CY
In a message dated 10/12/2024 11:55:40 AM Central Daylight Time, martin_ehrenfried@... writes:
?
Absolutely, it's no good unless you can notch out the fundamental tones.
?
My own notches provide about 40dB rejection, which is just about good enough.
?
Regards,
?
Martin
?
On Sat, Oct 12, 2024 at 03:21 PM, Fred M wrote:
Definitely, but without? notch filters for the original 2-tone signals the tiny-SA will become overloaded. It's own IP3 is about +17 dBm.
|
On Sat, Oct 12, 2024 at 12:11 PM, Everett N4CY wrote:
There is no way to run a proper IMD without a band stop filter that rejects the 2 tones and a band pass filter for each of the tones.
A decade back, I made use of an old tube-type Stoddart NM-22A Field Strength EMI/RFI receiver for this purpose, as, a tuned RF front end serves the purpose of fundamental filtering or notching. Beware using broad-band SDR receivers for this application w/o taking similar precautions (of course) as one would for a broad band Spec An.
?
Anything as broad in response as a spectrum analyzer is going to require assurances (such as outboard LP/HP/BP filtering) to assure the Spec An is not overloaded, or its dynamic range is not exceeded.
?
I was running tests on an active antenna mounted on a car. The 'antenna' (wire) was about 30 inches and driving the gate of an JFET. I could vary the bias point and observe 3rd order distortion products as created by several local AM radio stations in the area? ... different circuit configurations were tried, including buffer stages following the JFET input amplifier/Source follower stage. Sorry, no detailed notes were taken of this quick proof-of-concept testing.
?
de AA5CT Jim
?
?
|
Just wanted to give my 2c in regard to IMD/IP3 measurements as I did and continue doing lots of them at my work.
?
30-40dBm IP3/OIP3 measured levels are not that high. One just need a good "Big Box" Spectrum Analyzer (HP/Agilent/Keysight) with good LowPIM attenuator in front of it. No Bandstop/notch filters are really needed! Signal Combiner can produce some IMDs but usually they are pretty low for those levels of desired measured IP3. Bandpass filters for F1 and F2 are also not needed along with attenuators between those filters and Combiner. Actually if filters are used I'd recommend increasing the attenuators to 10dB each. Instead of Bandpass filters normally we would use an Isolator (circulators with termination on top). But.... that's mostly for VHF+ frequencies. Not sure if proper isolators do even exist for LF/MW/HF frequencies. So, filters may be the best F1/F2 solution before Combiner.
Now, my main work IP3 bench is capable of measuring IP3/OIP3 levels of up to 90-100dBm with ~2W of F1 and F2 each... That's pretty much measuring IP3 of passive components where everything matter including connector cleanliness and proper torque (sometimes overtorquing helps!). At those levels you cannot use lump components for filters and combiners and cavity stuff is the only thing that worked for me. For example I'm using celltower cavity diplexers/duplexers for F1 and F2 mixing and a big double cavity rejection filter in front of the Spectrum Analyzer (tuned to reject CW at F1 and F2 with rejection of close to 100dB). Big negative of such setups is inflexibility. You need to swap bench components to change the frequency or have multiply setups for each frequency of interest. And of course LF-HF and cavity filters probably cannot coexist due to the size :) But any of that is not needed when you measure amplifiers.
In regard to TinySA Ultra. It's a great thing but would be my #1 suspect for creating the intermods on it's own. So, with TinySA one indeed may need rejection filters in front of it... If anything I'd probably use an SDR with good front end for the task like that instead of TinySA.
?
Thanks,
Simon
|
Hi Simon,
?
Mabe it's possible with later generation analysers that have a greater dynamic range, but most stuff available to hobby users isn't going to cut it.
?
I have a HP8563E, but it hardly gets turned on, as apart from the frequency coverage to 26GHz, it's not that much use for everyday stuff.
?
My tone sources are two crystal oscillators followed PA's and then by attenuators, individual absorptive band pass filters, a high power combiner and yet another attenuator, all built into a standalone box. I start off with 5w of RF and end up with 100mW at the output. The mini-Circuits combiners that are normally specified start to generate their own IP's at around +10dBm, and standard band pass filters don't present a 50 ohm match outside their passband, which can affect the generation of IMD products.
?
Even with this, and a decent set of notch filters, I have difficulty making repeatable IMD measurements, mainly because of leakage and minor phase shifts that affect the generation of IMD. It's hard enough at HF frequencies, and having in the past measured passive IMD, produced in cellular antennas, I know how much more difficult that can be.
?
You read the application notes from folks like Aligent / Keysight, and they make the test setup it looks easy, but at least for me, in a home workshop, it wasn't that straight forward.
?
However, I'm prepared to accept that your mileage may vary...
?
Regards,
?
Martin
?
On Mon, Oct 14, 2024 at 06:29 PM, rfsam wrote:
One just need a good "Big Box" Spectrum Analyzer (HP/Agilent/Keysight) with good LowPIM attenuator in front of it.
|
No matter what your IMD setup is, you need to run the measurement bypassing the DUT first, to see the capability of the setup.
My IMD bench is:
?- Si5351a generating F1=3.58 and F2=4.914MHz,? 4mA drive strength (important)
- 5-resonator xtal narrow bandpass filters, Cohn topology. I can slightly tweak the input frequencies to hit the maximum, it moves with temperature.
- hybrid power combiner
- step attenuator
- dummy aerial if the DUT is a loop preamp (in my case it is a 1:1:1 transformer and two 25/0.5 Ohm dividers, no reactive elements, flat response, 50dB CMRR up to 30MHz)
- here is the preamp under test
- 5-inductor bandstop filter, 2.5 .. 5.5MHz. F1/F2 Attenuation is about 35..40dB. All products of interest are outside of the stop band, but I measure their attenuation anyway to correct for it during the measurement. It is about 1 .. 1.5 dB.
- 20 or 40dB fixed attenuator to bring the signals within the range of the receiver
- Airspy HF+ Discovery
- HDSDR software
?
The floor (IMD detected without the DUT) is
+110dBm OIP2 at 2.24MHz (F2-F1)
+109dBm OIP2 at 8.5MHz (F2+F1)
+47dBm OIP3 at 2.24MHz (2*F1 - F2)
+39dBm OIP3 at 6.25MHz (2*F2 - F1) - this is the one which limits my measurements, I don't see it changing with LZ1AQ or other good preamp in the path. Probably limited by the Si5351a outputs cross-modulating.
?
73, Mike AF7KR
|
On Mon, Oct 14, 2024 at 03:10 PM, vbifyz wrote:
My IMD bench is:
?- Si5351a generating F1=3.58 and F2=4.914MHz,? 4mA drive strength (important)
- 5-resonator xtal narrow bandpass filters, Cohn topology.
?
Would it be not less effort to use two well buffered X-tal Oscillators instead of an Si5351 followed by two x-tal filters? I do it this way with a RF-2-tone generator proposed by DC4KU and are able to measure with a Siglent 3302 an IP3 up to 36 dBm without bandstop filters. To achieve a higher measuring dynamic range bandstop filters are mandatory.
?
DC4KU RF-two-tone generator (pdf in german, please use a translator)
?
regards
Fred
|
Everett N4CY