1meg to 50ohm Impedance Matching Amplifier for general instrumentation use

[David Hess]

Quote from: tggzzz on Yesterday at 06:47:19 PM

Look at any of their scope front panels before the 24xx series. The timebase has units of mS.

Well, let's see.  I have a 2232 and 2247A in front of me.  And I have a 2230, 465, various 7000, and a 547 in the other room.  All use lowercase "s" except for the 547 which uses all lowercase "sec" on their front panels and except for the 7000 readouts, they also all use lowercase "s" but the later later readouts had better precision.

So do you have a specific example?

I'm going to keep using mhos just to spite SI; and because it is much more mnemonic and avoids confusion with seconds no matter how marked.


Insatman, below is an example of a simple 3 transistor impedance converter used as an active probe.  For lower noise where low input capacitance of the FET is actually a problem, a JFET traditionally used for low noise audio (Linear Systems makes these now) can still achieve 100+ MHz.  I would probably add a trimmer capacitor to shunt the input so the input capacitance can be trimmed to match your oscilloscope using a normalizer or x10 passive probe.

A feedback amplifier will of course have lower distortion at low and mid frequencies.  A more complex discrete circuit could add a cascode to the FET for better performance up to at least 100 MHz at the cost of more noise from active devices and a good chance of VHF oscillation without careful layout.

I seem to recall that TI has some 8-pin DIP high speed operational amplifiers with a pinout that includes 2 output pins shortening the feedback path for better performance.  LT had a part I would have recommended but AD has absorbed their web site and apparently discontinued it.

Burr-Brown made some really nice operational transconductance amplifiers (OPA860 and OPA861) which would work well in this application but TI never followed up on them and none had FET inputs so they would require an input buffer.  They could effectively implement the x2 gain stage and cable driver shown in the Tektronix P6202A example with shunt termination instead of series termination in a single part.

[tggzzz]

Quote from: tggzzz on Yesterday at 06:47:19 PM

Look at any of their scope front panels before the 24xx series. The timebase has units of mS.

Well, let's see.  I have a 2232 and 2247A in front of me.  And I have a 2230, 465, various 7000, and a 547 in the other room.  All use lowercase "s" except for the 547 which uses all lowercase "sec" on their front panels and except for the 7000 readouts, they also all use lowercase "s" but the later later readouts had better precision.

So do you have a specific example?

Now that really annoys me: you are right and I am wrong.
 
Where on earth did I get that misapprehension - which has lodged in my brain over the decades? I must be channelling Aristotle, who maintained women had fewer teeth than men, but never opened either of his wifes' mouth to count them!

I'm going to keep using mhos just to spite SI; and because it is much more mnemonic and avoids confusion with seconds no matter how marked.

That is unambiguous, if archaic.

[David Hess]

Now that really annoys me: you are right and I am wrong.

It happens to me also.  Forget about it.
 

Where on earth did I get that misapprehension - which has lodged in my brain over the decades? I must be channelling Aristotle, who maintained women had fewer teeth than men, but never opened either of his wifes' mouth to count them!

I had examples at hand to look at first hand so did not have to rely on memory.  For all I know, the other 465 and 22xx series models use an uppercase "S".

[tggzzz]

Now that really annoys me: you are right and I am wrong.

It happens to me also.  Forget about it.
 

Where on earth did I get that misapprehension - which has lodged in my brain over the decades? I must be channelling Aristotle, who maintained women had fewer teeth than men, but never opened either of his wifes' mouth to count them!

I had examples at hand to look at first hand so did not have to rely on memory.  For all I know, the other 465 and 22xx series models use an uppercase "S".

So do I, and I didn't choose to look. Hence Aristotle, and hence it being so damn embarrassing/annoying

[Insatman]

Unfortunately, all this nomenclature fussiness has distracted from any real technical discussion of the; Impedance changing amplifier, Delta-Z Buffer or thing-a-ma-jig in question.  I do find this item quite useful.   It can be put into a spectrum analyzer - tracking gen loop to allow for characterization of scope probes or other 1Meg output impedance items.  It is also useful for extending the cable on such items in certain test situations.   

Impedance converters or buffers are very handy.  Another place they can be used is to drive a 50 ohm frequency counter with a standard high impedance passive prove.  I have seen them used for low frequency network analysis.

With an adapter that includes enough shunt capacitance, an active probe can be used as an impedance converter.

When I designed this buffer, I set up the input to look like a typical scope input ie 1M/18pf specifically so scope probes would work correctly.  That would also apply to your active probe I would guess in most cases.   Trim adjustment of the probe would probably be advisable for best performance.

[Insatman]

I built another version of this type of Impedance Matching Amplifier based on a different op-amp.  This version, while more complicated, uses a lower cost chip ($5 vs $15) and has a few other advantages and some disadvantages:

Advantages:
1) It's gain is 0dB rather than -6dB nominal.   
2) It has a slighter higher -3dB point and the roll off is slower, resulting in usability over a greater range.   
3) The 2nd Harmonic at 100MHz is down about -50dB rather than -40dB for the previous circuit.
4) Op-Amp is cheaper
5) Possible performance improvement through circuit tweaks.
6) Increased gain possible.

Disadvantages:
1) Circuit is more complicated and requires +/-5 volt regulators.
2) Previous circuit would tolerate higher input voltages as it used 9V supplies rather than 5V.
3) Harmonic data shows some spurious signals between the 2nd and 3rd Harmonic, largest of which is -50dB down from fundamental.

This version of the circuit board had an error on the IC pinout so some bodges were required.  Also a pot was bodged in to adjust the gain to as close to 1:1 as possible over the frequency range of interest.
I believe that further improvement can be obtained from this op-amp, after another run of PCBs to correct errors and add a few elements for flexibility we will try again.

Yet another version of this circuit using the VCA821 suggested amplifier is in progress but I've yet to receive the PCBs.   

[David Hess]

When I designed this buffer, I set up the input to look like a typical scope input ie 1M/18pf specifically so scope probes would work correctly.  That would also apply to your active probe I would guess in most cases.   Trim adjustment of the probe would probably be advisable for best performance.

High frequency active probes are designed to have very low input capacitance for obvious reasons so they require added input shunt capacitance to work as an impedance converter for a standard passive probe.

I have not seen this so much in modern oscilloscopes but older oscilloscopes usually have input capacitance adjustments for each channel which allows all channels to have a matched input capacitance so that a compensated probe can be moved between channels without affecting the compensation.  This also allows different oscilloscopes, or at least those with the same nominal input capacitance, to be adjusted to have the same input capacitance.  This input capacitance adjustment on an impedance converter would allow probes to be moved around without extra compensation adjustments.

One thing I have thought about making is a differential impedance converter to use a pair of x10 or x100 probes as a differential probe.  Old oscilloscope differential amplifiers rely on the probes for DC and AC common mode adjustments but I would want to move these adjustments to the amplifier instead of buying specialized probes; I think LeCroy is the only supplier for these today.

[Insatman]

I built another version of this type of Impedance Matching Amplifier based on a different op-amp.  This version, while more complicated, uses a lower cost chip ($5 vs $15) and has a few other advantages and some disadvantages:

Advantages:
1) It's gain is 0dB rather than -6dB nominal.   
2) It has a slighter higher -3dB point and the roll off is slower, resulting in usability over a greater range.   
3) The 2nd Harmonic at 100MHz is down about -50dB rather than -40dB for the previous circuit.
4) Op-Amp is cheaper
5) Possible performance improvement through circuit tweaks.
6) Increased gain possible.

Disadvantages:
1) Circuit is more complicated and requires +/-5 volt regulators.
2) Previous circuit would tolerate higher input voltages as it used 9V supplies rather than 5V.
3) Harmonic data shows some spurious signals between the 2nd and 3rd Harmonic, largest of which is -50dB down from fundamental.

This version of the circuit board had an error on the IC pinout so some bodges were required.  Also a pot was bodged in to adjust the gain to as close to 1:1 as possible over the frequency range of interest.
I believe that further improvement can be obtained from this op-amp, after another run of PCBs to correct errors and add a few elements for flexibility we will try again.

Yet another version of this circuit using the VCA821 suggested amplifier is in progress but I've yet to receive the PCBs.

I finally built and tested the VCA821 IC based impedance matching amplifier as suggested by another member of this blog (thanks MasterT).   This IC is surface mount only and the version I bought costs about $9 from Digikey.  The BW is the best so far at over >300Mhz.   This op-amp is different from most and the datasheet is confusing to say the least.  I spent most of a day getting it to work this well and further tweaking is probably possible.  If particular importance is R5,R12 and C6 to tweak high end frequency response.

I will order a revised PCB eventually and see if I can improve things next version.   Like my previous posted version this one has variable gain and is tested at 1:1 into 50 ohms. 

Advantages:
1) It's gain is variable and tested at 0dB not -6dB as in my first version
2) It has a significantly higher -3dB point at ~330MHz  and the roll off is reasonable.   
3) Op-Amp is cheaper than 1st version.
4) Probable performance improvement through circuit tweaks.

Disadvantages:
1) Circuit is more complicated and requires +/-5 volt regulators.
2) Original circuit would tolerate higher input voltages as it used 9V supplies rather than 5V.
3) Harmonic data at 10Mhz 2nd/3rd is down >60dB but at 100MHz 2nd/3rd is down 35dB/53dB.

All three version of the PCB share identical BNC connector spacing (2") so one housing design works for all of them.  All can use 9V batteries as on board voltage-regulators are incorporated.

[Alex Eisenhut]

Quote from: tggzzz on Yesterday at 06:47:19 PM

Look at any of their scope front panels before the 24xx series. The timebase has units of mS.

Well, let's see.  I have a 2232 and 2247A in front of me.  And I have a 2230, 465, various 7000, and a 547 in the other room.  All use lowercase "s" except for the 547 which uses all lowercase "sec" on their front panels and except for the 7000 readouts, they also all use lowercase "s" but the later later readouts had better precision.

Meh? Not mine! Indeed, not in any picture I could find, they all have SEC in upper case. And I leafed through Stan Griffith's book and looked at tons of old Tek stuff.

[tggzzz]

Quote from: tggzzz on Yesterday at 06:47:19 PM

Look at any of their scope front panels before the 24xx series. The timebase has units of mS.

Well, let's see.  I have a 2232 and 2247A in front of me.  And I have a 2230, 465, various 7000, and a 547 in the other room.  All use lowercase "s" except for the 547 which uses all lowercase "sec" on their front panels and except for the 7000 readouts, they also all use lowercase "s" but the later later readouts had better precision.

Meh? Not mine! Indeed, not in any picture I could find, they all have SEC in upper case. And I leafed through Stan Griffith's book and looked at tons of old Tek stuff.

I refer you to my post that immediately followed David Hess' post:
https://www.eevblog.com/forum/projects/1meg-to-50ohm-trans-impedance-amplifier-for-general-instrumentation-use/msg1467894/#msg1467894
(plus the ensuing conversation between David an me )

[IDEngineer]

I also put slashes through my zeros and dashes through my 7s and Zs.

As do I.

I also write my dates in DDMMMYYYY format, thus: 24May2018. This completely eliminates the possibility of confusion between the day and month, since depending upon the writer and reader the two formats MM/DD/YYYY and DD/MM/YYYY are not definitive for days <= 12. The four digit year might seem unnecessary, except to those of us who write software and/or lived through the whole Y2K thing.

In my opinion, the goal of written communication is to communicate. Details that reduce ambiguity, such as slash-throughs and dash-throughs and choice of formatting, are a good idea.

I do have a story about this, though. I adopted the slash-zero in junior high school (grades 7-8). One day a student was asked by the teacher to grade a test, and I got something like 50% wrong (?!?). Turned out the student thought my slash-zeroes were where I'd changed my answers, so she marked every single one incorrect. I pointed out that every single zero on the page had a slash through it, none of the ones had a circle, and asked her the odds that I'd "correct" literally every single zero on the page to a one. She didn't relent. Grrrrr. 

[David Hess]

I also write my dates in DDMMMYYYY format, thus: 24May2018. This completely eliminates the possibility of confusion between the day and month, since depending upon the writer and reader the two formats MM/DD/YYYY and DD/MM/YYYY are not definitive for days <= 12. The four digit year might seem unnecessary, except to those of us who write software and/or lived through the whole Y2K thing.

I write out my dates in some form of numeric YYYYMMDD to match what I type into computers so they may be easily sorted.  This is essentially ISO 8601 and I refuse to conform to the US civilian standard.  For convenience I may include the text month.

I also set all of my clocks to 24 hour time if possible.

And I swap the control and caps-lock keys on my keyboards so that the control key is next to the A key where God intended it.  This has occasionally resulted in hilarity when someone tries to use my computer.

[IDEngineer]

I write out my dates in some form of numeric YYYYMMDD to match what I type into computers so they may be easily sorted.

Good point, I too do that with file and directory names (e.g. "2018-05-23_mnemonic_info.ext") so they will sort by date. I tried that for a while in written communications but it made things worse, not better. Try dating a signature with that format at a bank or the DMV and settle in for a long afternoon....

And I swap the control and caps-lock keys on my keyboards so that the control key is next to the A key where God intended it.  This has occasionally resulted in hilarity when someone tries to use my computer.

I used to design keyboards in a previous life. Strictly speaking, there's no way to follow historical standards with today's keyboards. Caps Lock should be SHIFT Lock, for example (though frankly Caps Lock is more useful). And the -Lock key is what belongs to the left of the A key.



There was no Ctrl key on traditional keyboards, so it was just tacked on someplace. Later the same was done with the Alt key. Note the singular... early 80's keyboards only had a single Ctrl and single Alt key. Later designs started having two, like shift keys (which they functionally resemble). I bear some responsibility for that.

I just purchased a new ThinkPad X270. They reversed the Ctrl and Fn keys (Fn is to the outside on the extreme left  ). Get this: They know it's so non-standard that there is an option in the BIOS to flip those two keys! Seriously, if what you're doing is so contrary to societal norms that you ALREADY know you need a bailout option, why not just do it properly in the first place? Or at least make the standard the default, so the keys are silkscreened properly?

[ebastler]

I just purchased a new ThinkPad X270. They reversed the Ctrl and Fn keys (Fn is to the outside on the extreme left  ).

Your first Thinkpad, I reckon? They have been doing this since 1992:
https://forums.lenovo.com/t5/ThinkPad-T400-T500-and-newer-T/Why-is-the-FN-key-on-the-outside-left-instead-of-CTRL/ta-p/359099

(Which still doesn't make it a good idea, in my book...)

[IDEngineer]

Your first Thinkpad, I reckon? They have been doing this since 1992.... (Which still doesn't make it a good idea, in my book...)

Amazing. I wonder who thought that was a good idea in the first place.

Yes, my first ThinkPad. I have zero brand loyalty, I just start with a list of required specs. I build my own home machines but you sort of have to buy laptops.  So I start with specs and winnow the field.

High on my list this time around was native Win7 so Win7 drivers were downloadable from the manufacturer's support site for all of the onboard peripherals. I just can't stand the post-Win7 user interfaces. Why Microsoft insists on forcing their users to constantly relearn how to use a tool is beyond me. Computers aren't a hobby for me, they are a TOOL and I resent having to relearn where they've hidden all the important stuff every couple of years. Remember when Office already had 95% of its market, yet they completely redesigned the UI?  Here's an idea: Let's put the steering wheel in the back seat and have the driver sit sideways. No, really, it's better because they can look forward AND backward with just a turn of the head! Trust us, we know better than you.

It's notable that, despite there being provably better keyboard layouts (example: Dvorak), we continue to raise new generations of typists on QWERTY whose reason for existence died along with typewriters that had typebars instead of the Selectric's spherical element. If the marketing research says to leave the keyboard layout alone, why doesn't that apply to the GUI?

The X270 was one of the last laptops to have a replaceable battery, RJ-45 Ethernet port, and a good array of USB and other comm ports. I'm often in a hotel room late at night, revving code based on customer meetings that day in preparation for more customer meetings the next day, and I usually have a backpack development environment with me (portable scope, power supply, ICE/debugger, various sniffers and other USB-interfaced tools, etc.). The hyper-thin laptops are emotionally appealing but lack Ethernet ports or enough USB's to do that job. The X270 is still an ultraportable, amazingly compact by the standards of just a few years ago, which is thin and light enough (~3 pounds) for me. More importantly, it has ports for everything and I get 20+ hours of battery life before I have to swap in another battery or plug in somewhere. Not even Lenovo's X1 Carbon can do that (if it could, I'd own that instead because it really is a cool design).

Interesting about ThinkPads swapping those keys so long ago. That was back when they were owned by IBM, and then Motorola, which means they would have had a more US-centric viewpoint (instead of China today). What could have made them buck an industry standard like that?

EDIT: I just read the comments in that link you provided. Here's one that sums it up: "My business partner standardized our laptops on Lenovo. Let me tell you, I remedied that in less than a year. Why, and what would make me switch back to these otherwise great machines? This button arrangement. Love the explanation, and I wish the others had all followed suit, but they didn't. We've swapped the buttons in BIOS but it causes unfamiliar users tons of pain as they keep raising issues thinking the button is broken because they're pressing the wrong one now."

[ebastler]

Yes, my first ThinkPad. I have zero brand loyalty, I just start with a list of required specs. I build my own home machines but you sort of have to buy laptops.  So I start with specs and winnow the field.

That's how I have selected notebooks, ever since parametric search engines were made available by a few vendors and price comparison portals. But despite the fact that various generations of Thinkpads had attractive specs (both back in IBM days and later Lenovos), I never bought one for personal use since the keyboard layout did put me off.

EDIT: I just read the comments in that link you provided. Here's one that sums it up: "My business partner standardized our laptops on Lenovo. Let me tell you, I remedied that in less than a year. Why, and what would make me switch back to these otherwise great machines? This button arrangement. Love the explanation, and I wish the others had all followed suit, but they didn't. We've swapped the buttons in BIOS but it causes unfamiliar users tons of pain as they keep raising issues thinking the button is broken because they're pressing the wrong one now."

My employer just went through an extended phase of buying Lenovo machines as a standard. (Just switched back to DELL after > 5 years, whew...) I have activated the Fn/Ctrl BIOS swap on my corporate notebook, and it works for me. But it sure confuses others who have to type something on my machine, e.g. during a meeting, and I struggle and curse when I have to use theirs. Why couldn't Lenovo at least make the two keycaps the same size, so you can physically swap them?

Regarding quotes on that web page, my favorite one is from the main author of that Q&A: "Every other notebook personal computer manufacturer that I know of  has the Fn and Ctrl key positions swapped." 

[Mechatrommer]

Impedance converters or buffers are very handy.  Another place they can be used is to drive a 50 ohm frequency counter with a standard high impedance passive prove.  I have seen them used for low frequency network analysis.

why dont just build an active probe (Hi-Z in 50 ohm out) with selectable 1x 1/10x input? less one Hi-Z probe to take care about. cost should be less than $100 alltogether.

[David Hess]

Impedance converters or buffers are very handy.  Another place they can be used is to drive a 50 ohm frequency counter with a standard high impedance passive prove.  I have seen them used for low frequency network analysis.

why dont just build an active probe (Hi-Z in 50 ohm out) with selectable 1x 1/10x input? less one Hi-Z probe to take care about. cost should be less than $100 alltogether.

I suppose I could but x10 passive probes are cheap and rugged and all of my frequency counters have 1 megohm inputs and some of my frequency counters lack 50 ohm inputs.  Internally 1 megohm input frequency counters have an impedance buffer driving the low impedance circuits where amplification and comparison is made.

When working above 200 MHz, an active probe and 50 ohm input become a better option and very few frequency counters have high impedance inputs much above 100 MHz.  My fastest high impedance input is 225 MHz while 50 ohm inputs work into the microwave range of course.  Somewhere above 100 MHz you would want the lower input capacitance of an active probe (or low-z probe or 50 ohm connection) anyway so this all makes sense.

[IDEngineer]

Why couldn't Lenovo at least make the two keycaps the same size, so you can physically swap them?

That would just be further admission (beyond the BIOS option) that they made a mistake. And continue to refuse to correct it.

Regarding quotes on that web page, my favorite one is from the main author of that Q&A: "Every other notebook personal computer manufacturer that I know of  has the Fn and Ctrl key positions swapped."

Comic genius! 

[Insatman]

Impedance converters or buffers are very handy.  Another place they can be used is to drive a 50 ohm frequency counter with a standard high impedance passive prove.  I have seen them used for low frequency network analysis.

why dont just build an active probe (Hi-Z in 50 ohm out) with selectable 1x 1/10x input? less one Hi-Z probe to take care about. cost should be less than $100 alltogether.

Because often it's not simply circuit probing I'm using the buffer amp with.  For instance, maybe I want to look at the frequency response of a standard 1M probe on the spectrum analyzer.  Put the buffer in the tracking gen loop, normalize, then put a 50ohm feed-thru term on the tracking gen cable and the probe tip w/ adaptor connected to it.  The output of the probe then goes to the buffer input....now you have the frequency response and attenuation of your 1M probe displayed on your SA.   Lots of applications for this kind of buffer.

[HVUser]

I just found this thread while searching for a product to do this. Would you consider selling these? If not, maybe posting masks/Gerber files?

[TimFox]

I just found this thread.  I have a Marconi 2374 high-Z probe (specified bandwidth 200 MHz) for my Marconi 2380/2382 400 MHz spectrum analyzer (50 ohm input).
I tested the THD of the probe at 500 kHz (half the maximum frequency of my Wavetek 98 DDSvgenerator, which has low THD at its 50 ohm output.
Without probe:  50 ohm output to 50 ohm input.  THD = 0.12% for the naked generator (-63 dB 2nd, -61 dB 3rd, and -70 dB 4th) at 0 dBm/50 ohms = 0.22 V rms.
With probe:  50 ohm output to 50 ohm terminator to probe to 50 ohm input.  THD = 1.5% (-37 dB 2nd, -46 dB 3rd, and <-72 dB 4th) at same generator level.
I quote these values as what can be accomplished with a simple circuit with discrete devices;  obviously, the probe distortion is much worse than the generator's THD. 
The probe comes with a set of input attenuators, which I haven't tested.

[Insatman]

I just found this thread.  I have a Marconi 2374 high-Z probe (specified bandwidth 200 MHz) for my Marconi 2380/2382 400 MHz spectrum analyzer (50 ohm input).
I tested the THD of the probe at 500 kHz (half the maximum frequency of my Wavetek 98 DDSvgenerator, which has low THD at its 50 ohm output.
Without probe:  50 ohm output to 50 ohm input.  THD = 0.12% for the naked generator (-63 dB 2nd, -61 dB 3rd, and -70 dB 4th) at 0 dBm/50 ohms = 0.22 V rms.
With probe:  50 ohm output to 50 ohm terminator to probe to 50 ohm input.  THD = 1.5% (-37 dB 2nd, -46 dB 3rd, and <-72 dB 4th) at same generator level.
I quote these values as what can be accomplished with a simple circuit with discrete devices;  obviously, the probe distortion is much worse than the generator's THD. 
The probe comes with a set of input attenuators, which I haven't tested.

I think you are right that a circuit with discrete devices could be designed/built for this purpose.  I chose to use op-amps simply because I am more familiar with them than discrete transistors.  It might be a nice learning project to build such a circuit someday.   

[Insatman]

1MEG TO 50ohm Impedance Amplifier Revisited.  2/2022 dwm

Recently after nearly 4 years I decided to take a fresh look at this project of mine an see if I can better characterize the performance and make some improvements.

The revised board/circuit is based on the VCA821 design described earlier.  Mechanically the new board has identical dimensions and can replace the old version except the connectors are reversed.  The prior version had the input on the right and output on the left.  This was a mistake on my part as I prefer the input on the left and output on the right, so that is corrected in this layout.

Overall performance is similar; however, I have made much more extensive measurements this time around.   

Power consumption for the board is approx. 35 to 40mA for the +/- 5 regulators.  +/-9V was supplied from a bench supply to mimic the 9V batteries normally used to power the circuit.

In general I found that THD increases quickly at higher frequencies when the input is above 0dBm.   Operation at between -10dBm (70mV RMS 50ohm) and 0dBm (.707mV RMS 50ohm) is recommended.

Frequency Response was measured at +0.85dB, -3dB from 1MHz to 280MHz at 0dBm input.  The peak value of +0.85dB occurs around 100MHz.  Instrument used was Siglent 3021X with tracking gen
Frequency Response was measured at +1.44dB, -3dB from 1Mhz to 340MHz at -10dBm input.  The peak value of +1.44dB again occurs around 100MHz.  Instrument used was Siglent 3021X with tracking gen.  Low Frequency Response was measured using ARTA, a laptop and an external USB sound “card” device.   I estimate the response to be < +/- 0.3dB.

THD was measured at 1KHz, 10KHz, 2.5MHz, 45MHz, 100MHz and 190MHz.  the 45, 100, and 190MHz measurements were performed at two power levels.   The reason for the odd frequencies is that I lack a clean signal source for frequencies above 10’s of KHz.  Both my Chinese signal generator (FY6800) and vintage HP8654A both have significant 2nd and 3rd harmonics.  So the 2.5 thru 190MHz clean tones were generated my placing a lowpass filter on the generator output.  For example the 2.5MHz signal used a 1.9MHz low pass filter.  This results in an attenuated but clean fundamental tone.   Some examples of the drive tones are shown below.

I do not believe that THD measurements below 0.3% can be counted on a accurate, so that is the lowest value I will state.
1KHz THD <0.3% at 0dBm (ARTA)
10KHz THD <0.3% at 0dBm (ARTA)
2.5MHz THD <0.3% at 0.5dBm (3021X)
45MHz THD 2% at 0dBm, 45MHz THD 1.0% at -10dBm (3021X)
100MHz THD 1.3% at 0dBm, 100MHz THD 0.8% at -10dBm (3021X)
190MHz THD 2.6% at 0dBm, 190MHz THD 0.9% at -9dBm (3021X)

Time domain SCOPE overlays are also shown below.  Scope used as TEK784D (1GHz), Signal gen was FY6800.  Signal was split using a power divider with one leg going to the DUT and the other to scope channel 3.  The DUT is displayed on CH4.  Both channels are 50ohm terminated.

[Insatman]

Additional post to add more JPG data