Calculating SMPTE IMD %

[compresearch]

Hello, i have been trying to find the correct calculation for an IMD % figure using the SMPTE method.

I found the following statement here http://www.rane.com/note145.html

"the modulation components of the upper signal appear as sidebands spaced at multiples of the lower frequency tone. The amplitudes of the sidebands are RMS summed and expressed as a percentage of the upper frequency level"

I presume from this the sidebands created at the upper frequency are used in the calculation? What is not clear however is what sidebands should be included? Let's call the sidebands either side of the frequency tone the upper and lower sidebands. So should the calculation include only non-harmonic upper side bands? Use only the lower sidebands? Or a mixture of both?

 

[JohnRoberts]

SMPTE IMD is an archaic spec that dates back to old movie sound systems when 7kHz was HF and 60Hz was LF.

The distortion components should be spikes +/- 60Hz from the 7kHz, so  6940 and 7060. RMS summed means Square root of the sum of those two terms squared, and then that represented as a fraction of the 7kHz amplitude.

I stopped using SMPTE in the 70s because it didn't tell me anything meaningful about circuits (too easy). I modified my SMPTE IM analyzer (heathkit) to use 19kHz and 20 kHz tones 1:1 and measured the IM product down at 1kHz.

This HF two-tone IMD was much more revealing than THD for phono preamp design because the RIAA feedback rolled off harmonic distortion products -6dB/oct while boosting 1kHz IMD products +20 dB relative to 20 khz level. 

If you want to make a product spec sheet look good use SMPTE but IMO it's meaningless (too easy) for modern audio paths.

JR

 

[compresearch]

The distortion components should be spikes +/- 60Hz from the 7kHz, so  6940 and 7060. RMS summed means Square root of the sum of those two terms squared, and then that represented as a fraction of the 7kHz amplitude.

Thanks for the reply. Just to be clear,  are including  both the upper and the lower sidebands that cluster around the 7kHz tone and not just the upper or the lower?  How many of the components are used in the calculation? Just the 6940 and the 7060, as in your example, or the 6880 and the 7120 as well?

 

[JohnRoberts]

Include them all... higher and lower...

Using the square root of sum of squares addition will make the smaller spikes less significant in the total..


JR 

 

[compresearch]

Include them all... higher and lower...

Using the square root of sum of squares addition will make the smaller spikes less significant in the total..


JR

Thanks for the reply. Funny enough thats what i have been doing but then i read a passage from Temme Audio measurments that got me thinking it should only be the upper components. It says 

"A good rule of thumb is to measure more than N times above the fixed tone (f2) for IM distortion"

http://www.bksv.co.uk/doc/BO0385.pdf on page 11.

 

[PRR]

The classic 60+7KC IM meter has a 1,500CPS high-pass, rectifier, 600CPS low-pass, and then measures everything which is left. So a very wide band around 7KHz, multiple high-order products on both sides.

60+7K (or 40 or 90, maybe 2150) IM was illuminating in the late 1930s. Many amps would pass single tones tolerably well but garbled 2-tone, especially above half-power. Attention to SMPTE numbers gave us the amps of the 1950s, which didn't outright suck on complex orchestral passages (especially if you never hear an orchestra live). An amp can give low SMPTE IM numbers and still haze-up complex music with thousands of high-order low-level (too low to read on an IM meter) IM products.

I'd think at this date, any amp which reads above hiss on a 1940 IM test *wants* to distort. Guitar amps, flavor-boxes, etc. I don't think is any good correlation between an IM number and a "sound", except between amps of identical topology and margin.

 

[JohnRoberts]

Back in the day when I had both THD and SMPTE IMD on my bench it noticed that most decent audio paths showed a simple ratio between IMD and THD, both are caused by non-linearity. It got so I could just about predict IMD from the THD @ 20kHz.  The typical circuit using NF with dominant pole stability compensation (and flat frequency response high enough to capture and measure several harmonics) would show increasing THD with frequency, as predicted by the falling open loop gain and shrinking loop gain margin. 7 kHz  just isn't heavy lifting for a 20-20kHz path.

The reason I embraced two-tone IMD (19-20kHz) is because it stressed the audio path with a slew rate roughly 2x 20kHz, but with in-band distortion products that didn't get rolled off by LP band limiting filters, making the measurements look better than they are (don't get me started on RIAA preamps and THD specs.)

At the end of the day it all comes down to linearity and loop gain margin....  SMPTE IMD just isn't hard enough.

JR

 

[compresearch]

The classic 60+7KC IM meter has a 1,500CPS high-pass, rectifier, 600CPS low-pass, and then measures everything which is left. So a very wide band around 7KHz, multiple high-order products on both sides.

Great, seems John was right and i was doing it correctly to start with.

Thanks for the reply

 

[compresearch]

Back in the day when I had both THD and SMPTE IMD on my bench it noticed that most decent audio paths showed a simple ratio between IMD and THD, both are caused by non-linearity. It got so I could just about predict IMD from the THD @ 20kHz.  The typical circuit using NF with dominant pole stability compensation (and flat frequency response high enough to capture and measure several harmonics) would show increasing THD with frequency, as predicted by the falling open loop gain and shrinking loop gain margin. 7 kHz  just isn't heavy lifting for a 20-20kHz path.

I have read, i think in Glenn Ballous book, that IMD can be predicted to be 3 or 4 times more than THD. Does that sound about right?

The reason i asked this in the first place is i was measuring THD and IMD for an 1176 compresor (set at 20:1 with fastest attack and release) but noticed the IMD seemed to be perhaps 5 or 6 times greater. I was scratching my head thinking why that could be the case and my first thought was i had used too many IMD artefacts in the % calculation.

Any ideas why that could be? 

 

[PRR]

> IMD can be predicted to be 3 or 4 times more than THD

For "many" older amplifier topologes.

With "cleverer" amplifiers the ratio will not be the same.

Especially at different level. For most "simple" systems, 2nd harmonic rises roughly as amplitude, 3rd as square(?) of amplitude, etc. IMD will follow similar but different trends. The ratio *can't* be constant.

> 1176 compresor

Simple old-style push-pull amp but with a side-chain hammering on it. Without racking my brain, I would agree it may have IMD artifacts above what we'd expect from THD number.

And at 20dB limiting the poor input stage is "starved". No clean-amp designer would work it at such low current; but low current is essential for reduced gain.

And returning to my point: if you wanted a "no IMD" compressor you'd start from a THAT VCA and a sophisticated sidechain. Committing to a 2-bottle design means you "want" a sound. Whether IMD is 4X or 6X THD is probably unimportant; though you can fool with biasing or re-design the level structure to see if it changes.

 

[compresearch]

Thanks for that PRR, your reply raises a lot of good points!

Cheers