THD measurement

Use some input signal to cancel a copy of the output, in a summing amplifier. See AoE III Figures 5.43, 5.44, page 331, distortion curves for many op-amps. At 1MHz quite a few come in well under -80dB, -90dB actually, and at 10MHz under -80dB, so you should be able to make such a circuit. Then your analyzer can handle the rest.

I copied these curves from manufacturer's datasheets. Many were at higher signal levels than you need to use and/or more severe output loads than you need to use.

--
 Thanks, 
    - Win

That's a nice idea. I'll simulate and see how much the phase shift of the amp affects the cancellation. I guess I could tweak the phase shift a little somewhere to improve the null.

I should still start with a pretty good sine wave, I guess. It's impressive how bad a lot of RF signal generators are on THD... like

-20 dBc!

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

You only need a partial cancel to help your analyzer.

Minicircuits sells passive 50-ohm RF filters you can use to remove or attenuate the harmonics.

--
 Thanks, 
    - Win

That should be a very easy measurement to make. Do you just need to measure at the two frequencies, 1 MHz and 10 MHz?

What is the amplifier gain?

Yes.

3, single-ended input to differential output, all 50 ohms.

Easy?

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Sure. What signal generators do you have access to? What SA do you have? A HP 8568 can do the measurement easily.

Build a couple of 1 and 10 MHz notch filters to get rid of the original signals and maybe the second harmonics. maybe a LPF to reduce the higher noise and spurious.

Sounds like a fun project.

What is the amplifier bandwidth?

We used to use Winfield's method quite successfully.

Around 70 MHz.

RF generators and spectrum analyzers are optimized for close-in phase noise, and often have horrible harmonic distortion. The 8568 is spec'd for

You can easily clean up the signal generator with a $10 minicircuits LPF. For the 1 MHz, use the LPF-B0R7+. It is SMT so use one of your nice test boards you just displayed. Or buy the test board for $30. For the 10 MHz test, use the SCLF-10+.

Both of these are 50 ohm non-reflective filters.

Get yourself one of the ebay HP 8566B or 8568B spectrum analyzers and displays. While these oar very old school, they are quite a bargain. Originally approaching $100k in 1980 dollars. Ask PH.

It's not specified for attenuation at 2F, but it's only down 66 dB at

5F. Looks like maybe -50 at 2F. I might have to make my own bandpass on the generator side. Nuisance.

I could maybe roll my own true sinewave XO, optimized for low distortion, and filter that a bit. Ditto nuisance. A crystal has insane Q, so should make a good sine wave.

It is SMT so use one of your nice test

I'd need two at least, for test set spares. And we'd have to talk to them, probably 488. Yuk.

Gonzo filters and a scope FFT, or Win's cancellation and scope FFT, maybe.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

ge

t of

is

them

ew

S to

wave

ally

the

e? A

l

er

.

have a look at

Check out TTE Inc. You can have any filter you like, in a week, for a few hundred bucks.

Cheers

Phil Hobbs

Yeah, Divide and conquer! In the audio range I made/used an opamp HP, (Q=50) filter to measure the third and fifth harmonics, of an oscillator. There would be some 1st background, but the ~50dB 'scope FFT would/can measure it. (only a two pole filter.)

What happens (goes bad) if you put a fast opamp in an SV filter?

George H.

A notch filter would be easier.

--
 Thanks, 
    - Win

Bandpass on the generator side, notch after the amp.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Or, one could use a phase-locked amp to detect harmonic amplitude and phase in the source, and again to detect the amplitudes and phases in the output. It's something you'd want to program an instrument for, so you could walk away while it twiddles. Assuming 'total harmonic distortion' means no more than a dozen harmonics, two phases each, and taking ten seconds for each PLL to settle into lock then ten seconds to measure... you could get a fairly fine result by the end of lunch.

At least, you could if it were 100 kHz to 700 kHz, rather than '70 MHz'. How good is downmixing for this kind of thing?

Are you guys talking in two different directions? Win seems to be talking about making it easy on the analyzer, and John seems to be thinking about how to make a pure sine wave.

Just noticing...

--
Tim Wescott 
Control systems, embedded software and circuit design 
I'm looking for work!  See my website if you're interested 
http://www.wescottdesign.com

e

need both, like bandpass in put to get clean sine, notch on output to see how much of output is not fundamental

That's a SINAD measurement, not THD. In audio work THD is conventionally measured as SINAD, but in a bandwidth of 100 MHZ or more the difference is pretty important!

Cheers

Phil Hobbs