Message	Date	From	Subject
7634	2012-07-24 11:55:05	Ashhar Farhan	measuring IMD
correct me if i am wrong... when measuring IMD, we feed two signals, let's say 20 KHz apart from stable, low noise sources like crystal oscillators through a hybrid coupler into the circuit under test. at the other end of the DUT, we attach a step attenuator and a spectrum analyzer. i would guess that the f1-2f2 will show up at a further 20 khz away. now, my question is, if let's say, we use the W7ZOI specan, will it resolve each of these 20 KHz spaced spikes well enough to measure the IMD? considering that the resolution bandwidth is just 15 KHz then, won't the fundamental signals show up inside the skirt too? on the other hand, if we consider using a direct conversion receiver in place of the specan feeding a PC-based audio spectrum analyzer, then the plain vanilla audio cards will give us a 16-bit resolution (-32,000 to +32,000 reading of the voltage), which can give a best case dynamic range of 90 db, but will probably settle at 80db (given the noise figure of the PC's ADC). which is the simplest way to make these measurements? i can trade-off making these measurements at just one of few frequencies, using crystal filters, crystal sources, the 8703 power meter with the RF attenuator, etc. I am trying to understand and study IMD performances of individual blocks like mixers, amps, etc. - farhan
7635	2012-07-24 12:32:32	William Carver	Re: measuring IMD
As long as the down conversion to the audio spectrum analyzer is stronger than the DUT the audio spectrum analysis should be OK. One should be careful that signals, or harmonics, beyond half of the sampling rate do not contaminate the real data. The usual approach to identifying the IMD behavior is to vary the input signals by 1 dB, see if the third order IMD changes by 3 dB, for an example. While this works when the IMD is produced by one analog stage one can extrapolate at that 1:3 ratio to produce an "intercept" number. I have read that ADC can produce signals at the IMD frequencies but those will not follow the 1:3 ratio of the analog processes we are used to. So one has to be very careful to understand what you "expect" when you change input levels, and what you get. But your email indicates you have a good handle on that. But I can confess than correct interpretation of data is sometimes difficult. I have occasionally made some real stupid blunders! W7AAZ On Wed, 2012-07-25 at 00:25 +0530, Ashhar Farhan wrote: > > correct me if i am wrong... > when measuring IMD, we feed two signals, let's say 20 KHz apart from > stable, low noise sources like crystal oscillators through a hybrid > coupler into the circuit under test. at the other end of the DUT, we > attach a step attenuator and a spectrum analyzer. i would guess that > the f1-2f2 will show up at a further 20 khz away. > now, my question is, if let's say, we use the W7ZOI specan, will it > resolve each of these 20 KHz spaced spikes well enough to measure the > IMD? considering that the resolution bandwidth is just 15 KHz then, > won't the fundamental signals show up inside the skirt too? > on the other hand, if we consider using a direct conversion receiver > in place of the specan feeding a PC-based audio spectrum analyzer, > then the plain vanilla audio cards will give us a 16-bit resolution > (-32,000 to +32,000 reading of the voltage), which can give a best > case dynamic range of 90 db, but will probably settle at 80db (given > the noise figure of the PC's ADC). > which is the simplest way to make these measurements? i can trade-off > making these measurements at just one of few frequencies, using > crystal filters, crystal sources, the 8703 power meter with the RF > attenuator, etc. > I am trying to understand and study IMD performances of individual > blocks like mixers, amps, etc. > > - farhan > > > >

As long as the down conversion to the audio spectrum analyzer is
stronger than the DUT the audio spectrum analysis should be OK. One
should be careful that signals, or harmonics, beyond half of the
sampling rate do not contaminate the real data.

The usual approach to identifying the IMD behavior is to vary the input
signals by 1 dB, see if the third order IMD changes by 3 dB, for an
example. While this works when the IMD is produced by one analog stage
one can extrapolate at that 1:3 ratio to produce an "intercept" number.

I have read that ADC can produce signals at the IMD frequencies but
those will not follow the 1:3 ratio of the analog processes we are used
to. So one has to be very careful to understand what you "expect" when
you change input levels, and what you get. But your email indicates you
have a good handle on that.

But I can confess than correct interpretation of data is sometimes
difficult. I have occasionally made some real stupid blunders!

W7AAZ

On Wed, 2012-07-25 at 00:25 +0530, Ashhar Farhan wrote:
>
> correct me if i am wrong...
> when measuring IMD, we feed two signals, let's say 20 KHz apart from
> stable, low noise sources like crystal oscillators through a hybrid
> coupler into the circuit under test. at the other end of the DUT, we
> attach a step attenuator and a spectrum analyzer. i would guess that
> the f1-2f2 will show up at a further 20 khz away.
> now, my question is, if let's say, we use the W7ZOI specan, will it
> resolve each of these 20 KHz spaced spikes well enough to measure the
> IMD? considering that the resolution bandwidth is just 15 KHz then,
> won't the fundamental signals show up inside the skirt too?
> on the other hand, if we consider using a direct conversion receiver
> in place of the specan feeding a PC-based audio spectrum analyzer,
> then the plain vanilla audio cards will give us a 16-bit resolution
> (-32,000 to +32,000 reading of the voltage), which can give a best
> case dynamic range of 90 db, but will probably settle at 80db (given
> the noise figure of the PC's ADC).
> which is the simplest way to make these measurements? i can trade-off
> making these measurements at just one of few frequencies, using
> crystal filters, crystal sources, the 8703 power meter with the RF
> attenuator, etc.
> I am trying to understand and study IMD performances of individual
> blocks like mixers, amps, etc.
>
> - farhan
>
>
>
>