om:
emeasured as SINAD, but in a bandwidth of 100 MHZ or more the difference is pretty important!
or THD+N , but anywho, doesn't it still boil down to getting a clean input and getting rid of the fundamental so you can actually see the harmonics
t!
** Claiming that audio use SINAD is plain silly, the many differences make it so.Traditional audio THD testing involves removing the fundamental and measuri ng what is left over with an average responding meter AND in a bandwidth of 80 to 100kHz. This means supply frequency components and random noise well beyond the audible range get included.
More recently, spectrum analysis is used and individual harmonics under 22k Hz summed to get a number. This eliminates hum, most noise and harmonics b eyond 22kHz.
Some folk still like to quote IM distortion figures using a twin tone test with frequencies of say 11&12 kHz or 18&19 kHz - looking for the 1kHz diffe rence tone. Ought to be a tough test of high frequency linearity, especiall y with a bandwidth limited device like a CD player or class D amplifier.
But it isn't reliable - as my friend and colleague Rod Elliot found when he ACTUALLY tested the method.
The presence or near absence of a difference tone depends strictly on the n ature of the non-linearity.
Darn.
.... Phil
Win's "bridged differentiator" notch filter - fig 6.40 in AoE3 - is a good place to start. You can notch out the fundamental fairly emphatically, and amplify whatever's left.
-- Bill Sloman, Sydney
Hmmm.. maybe subtract amplified output from input to get residual artifacts, which may or may not be distortion (like line frequency bazz-fazz). That way,input waveform is theoretically not relevant.
If it's the customer's request, then just tell them "we don't have the equipment on hand to test this, so we need a purchase or rent of $xxx on the PO to support item YY on your spec"...
They'll either drop it, or give you the extra bucks (don't forget your standard markup on the gear).
I mean, it would be foolish of them to think you know everything, and have all the gear in the world, to be able to design for, and test to, any arbitrary specification. (Unless you misled them into thinking this wouldn't be an added cost, in which case that's bait and switch...)
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:
I have never tried this, but it may work:
Use a LC tank, tuned for the frequency you need. Kick that tank into oscillation by for very short time duration driving it to one of the rails by a FET
Then let the tank ring and do your measurements during the ringing. The LC is clean and THD should be very very low. The limited Q will result in an envelope that is falling slowly
You need a good low loss capacitor and inductor. So maybe a aircore might come to mind
Cheers
Klaus
Better be a low impedance (small L, huge C) to keep the transistor Coss variation in the ppms.
But then you probably won't be able to "drive it to the rails". Well, for small signal purposes, you neither need nor want rails, so that's okay I guess.
I mean, John has plenty of PHEMTs and stuff handy, but even so, you're talking >nF.
What about distortion varying as signal strength, though?
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
THD (total harmonic distortion) is only about harmonics and not any other spurious junk that comes along for the ride. What I think they want is THD+N (total harmonic distortion plus noise). Also, I'm rather surprised that they didn't also ask for IMD (intermodulation distortion). Keeping harmonics down is much easier than maintaining linearity over the operating range.
I used various RF notch filters to improve SA (spectrum analyzer) dynamic range. Once the fundamental signal (or two signals as in IMD measurement) has been reduced by the notch filter, the attenuation between the signal source and the spectrum analyzer can be reduced so that you can see below the previous noise floor.
When you do see a signal worth measuring, be sure to change the input attenuation a few dB and see what the display shows. If the display follows the change in attenutation, your system is linear and not generating any additional harmonics or mixing spurs. If the change on the display is different, or the displayed signal totally disappears, the system is not linear and you're generating that signal either in the attenuators, connectors, or SA.
Also a word of warning. Do NOT tune the notch filter while the input attenuation is reduced. Tuning off frequency can blow the SA RF input fuse, trip the protection circuitry, or as I've done a few times, burn up the input attenuator or SA front end.
Is that sine or sin()? So much sin, so little time. Note that there is sin all over the blackboard.
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
There is a separate, borderline impossible, noise spec. I expect a lot of 2nd harmonic and maybe a bit of 3rd.
There is an analog multiplexer in the signal path. I'm guessing that it will have more distortion than the -80 dB limit. That's hard to estimate.
Bite your tongue! You never know who might tbe reading this. Don't give them ideas!
Yeah. If I notch it deep enough, I can maybe use an oscilloscope FFT to measure the harmonics.
Scopes are more rugged than SAs!
I could imagine sin-ing with that professor.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
I like to measure the 1dB gain compression point at various frequencies. That's usually a little above the level where the amp begins to produce harmonics.
You have it easy. For a genuine nightmare, look into PIM (passive intermodulation) which measures the distortion products produced by the non-linearities in cables, connectors, housings, packages, racks, cavities, circulators, isolators, power combiners, and anything trying to rust. To pass, RF connectors needs to be silver plated and properly torqued: Near the bottom of the page are some typical specs under "PIM limits". For LTE cellular service, about -107dbm max level and -150dBc (dB below carrier power) PIM level is about right. For test signal generators, it's common to use two 40 watt mobile transmitters into a power combiner and use cavity notch filters to remove the two carriers. Your -80dB spec seem trivial by comparison.
Would burning my lip with overheated cheese suffice as penance? I just did that with a sandwich and my microwave oven.
Dunno. I've never tried that.
Scope have high impedance inputs. It takes far more volts input to a scope to produce enough heat to fry the input attenuator and input amplifier. A spectrum analyzer has a 50 ohm input attenuator which heats up nicely with minimal voltage and signal levels. If the scope had a 50 ohm input, it would be equally susceptible to overpower screwups as the SA.
You might be a bit late. The advertisement was from 1972, when she was 28 years old. Today, she would be 73 years old. I couldn't find much about her with Google except that she was probably working at NASA Langley when the picture was taken:
-- Jeff Liebermann jeffl@cruzio.com 150 Felker St #D http://www.LearnByDestroying.com Santa Cruz CA 95060 http://802.11junk.com Skype: JeffLiebermann AE6KS 831-336-2558
Air core inductors tend to be lossier than cored inductors, but cored inductors generate harmonics all on their own.
LT Spice allows you to plug in the John Chan model of hysterisis, and even with a heavily gapped ferrite core, I got harmonics about 105 dB below the funadmental (admittedly at 17kHz, or 100,000 radians per second).
John would be looking at nickel-zinc ferrites at 1MHz and 10MHz, but he's nervous about winding inductors at the best of times.
-- Bill Sloman, Sydney
Low pass filter on the gen, high pass filter on the SA. No big thing. M
The source will need to be low pass filtered. Yes, a bandpass will also wor k.
The output leg for measuring THD will likely need a high pass filter, as th is will relax the dynamic range requirements of the measuring spectrum anal yzer. Starting with the 2nd, then 3rd, and so on, add the power of the harm onics up until they are small enough such that they don't make a significan t difference to the sum.
Depending on your setup, you may actually need a diplexer on the amplifier output. The high pass port will go to the spectrum analyzer (SA) and low pa ss port to a proper termination. In fact, both ports need to be properly te rminated. IOW, the SA must also be at "system impedance." Amps need to be p roperly terminated, or the measurement doesn't mean anything.
It is easy to design diplexers with your Nuhertz filter synthesis tool.
I don't think a bandpass will work - you need to measure 5-7 harmonics at at least. Multiple band passes might work
-- Using Opera's mail client: http://www.opera.com/mail/
On a sunny day (Thu, 01 Jun 2017 17:30:42 +0300) it happened snipped-for-privacy@downunder.com wrote in :
Of course, the classic way, it was described by others here already, I used to do that every so often on the land audio lines for the Europe new feeds, they send a tone, you tune to the harmonics one by one and measure those, do the adding, and there is your distortion number. Cannot remember the type / name of the test equipment, was LF only anyways.
But again, good clean test sinewave, spectrum analyzer.. should work.
The problem will be coupling this to the amplifier. Its input Z (guessing 50R) will give a very low loaded Q.
Pere
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.