8648B signal generator, is it supposed to do that (harmonics)

[hp3310a]

I have a new-to-me Agilent 8648B 9kHz-2GHz signal generator. This is my first signal generator which is also faulty (low amplitude due to broken RPP), so I'm not entirely sure: Is it supposed to produce harmonics, at all/that strong?

The specifications say:
Harmonics (output <= + 4 dBm)
 < -30 dBc
Subharmonics (output <= + 4 dBm)
< 1001 MHz......... -60 dBc
<= 3200 MHz....... -50 dBc
<= 4000 MHz....... -40 dBc

In the screen shot I see the first harmonic at < -40dBm lower. Is that good or bad (not sure what dBc is in relation to dBm)?

[ftg]

You have the fundamental off scale or right at the reference level, so the spectrum analyzer might also be distorting and producing some of those spurious responses internally.

I'd first adjust the reference level to be 10dB higher than the signal generator output before considering the generator broken.

I recall the HP 8648B being generally well liked so I don't think it should be that dirty.
But the strongest harmonic is over 30dB down from the fundamental, so the <–30 dBc (output ≤+4 dBm) spec is being met.
Even if output power is above at or above +10dBm (SA reflevel in the picture).

dBc = deciBells relative to carrier.
So the difference compared to the fundamental emission.
And <-30dBc means they promise all harmonics will be at least 30dB weaker than the fundamental output, when output power is +4dBm or less.

The spec sheet for it can be found here if others want to take a look as well.
https://testequipment.center/Product_Documents/Agilent-8648B-Specifications-9F224.pdf

Edit: Just noticed that the reflevel is at +10dBm, so the generator output is at that or above.

[jonpaul]

Learn how to setup, use and protect the SA.

Possible to BURN OUT the input mixer with high power from a gen.

Get the HP manua;s READ AND LEARN

j

[hp3310a]

You have the fundamental off scale or right at the reference level, so the spectrum analyzer might also be distorting and producing some of those spurious responses internally.

I'd first adjust the reference level to be 10dB higher than the signal generator output before considering the generator broken.

I recall the HP 8648B being generally well liked so I don't think it should be that dirty.
But the strongest harmonic is over 30dB down from the fundamental, so the <–30 dBc (output ≤+4 dBm) spec is being met.
Even if output power is above at or above +10dBm (SA reflevel in the picture).

dBc = deciBells relative to carrier.
So the difference compared to the fundamental emission.
And <-30dBc means they promise all harmonics will be at least 30dB weaker than the fundamental output, when output power is +4dBm or less.

The spec sheet for it can be found here if others want to take a look as well.
https://testequipment.center/Product_Documents/Agilent-8648B-Specifications-9F224.pdf

Edit: Just noticed that the reflevel is at +10dBm, so the generator output is at that or above.

Many thanks for the detailed response. The picture was taken while I tested the generator's internal output on the A10 board which turned out to be fine. The signal was between 1 and 4 dBm over ref level for about 5 seconds.

I just took another picture with wider span and the ref level set appropriately. The second harmonic is about 32dBc lower than the fundamental. I assumed that signal generators (in general and this one specifically) would produce much purer tones. Because I'm new to all this, I wasn't aware that this is normal.

I do wonder how one avoids ever subjecting the spectrum analyzer to signals above ref level. I mean, this is something that can happen realistically. The signal is more powerful than anticipated, and the ref level needs to be adjusted after the fact. It sounds as if the spectrum analyzer wouldn't be able the handle short periods of slightly too low ref-leveling. Is that really the case?

[hp3310a]

Actually, please disregard the previous picture. The output was set to 10dBm - due to the faulty RPP only -4dBm exit the generator.

Here is one with the generator set to 0dBm amplitude (again, due to the fault, the SA shows less), the difference between fundamental and 1st harmonic is slightly better and higher order harmonics fall off much more quickly.

[Miek]

I do wonder how one avoids ever subjecting the spectrum analyzer to signals above ref level. I mean, this is something that can happen realistically. The signal is more powerful than anticipated, and the ref level needs to be adjusted after the fact.

The way to avoid it is to adjust your setup so that it can't happen. Set the ref level higher before you ever connect the signal. If there is still some chance that the signal could be more powerful than the safe limit for the analyser, hook it up with an external attenuator in line first, take a measurement, then decide whether you're happy to remove the attenuator.

It sounds as if the spectrum analyzer wouldn't be able the handle short periods of slightly too low ref-leveling. Is that really the case?

It depends on the model, it'll usually tell you in the data sheet or user manual but there isn't always much detail. For example, the 8560/8590 series datasheets specify a safe input level of 30 dBm given an input attenuation >= 10 dB. If you set the ref level low enough that it switches to 0 dB input attenuation, then the safe limit goes down (I don't think it actually specifies but I'd assume 20 dBm).

[G0HZU]

If you want to measure harmonic levels using a spectrum analyser, then the usual rule of thumb is to always make sure the harmonic distortion produced by the analyser itself is at least 15-20dB lower than the amount of distortion you want to measure or prove.
In other words, if the sig gen (genuinely) produces a second harmonic at -40dBc then the analyser ideally needs to keep its own internal harmonic distortion 20dB lower (at -60dBc) if you want to keep the measurement uncertainty (due to the analysers distortion) below about +/- 0.9dB.

The analyser should then correctly report the sig gen distortion as -40dBm +/- 0.9dB (ignoring other contributors to uncertainty for now).


The datasheet can give you some idea as to how much distortion the analyser produces for second harmonic distortion in the form of the 2HI spec. For the 859x series of analysers this is usually in the region of +25dBm for the mixer with 0dB attenuation. However, this is usually a conservative estimate and the 2HI might typically be +40dBm for a 100 MHz test signal. But I'm just guessing here as I've not used one of these old analysers for quite a while.

If the mixer 2HI really is +40dBm at 100 MHz and you want -60dBc internal 2nd harmonic distortion then the mixer input level should be -20dBm or less. So if the sig gen produces +10dBm you would need 30dB internal attenuation selected in the analyser menu. If you opt for the worst case 2HI of 25dBm as given in the datasheet the mixer would need to be driven at -35dBm and this would require >45 dB attenuation ahead of the mixer for an input level of +10dBm from the signal generator.

[Sorama]

Actually, please disregard the previous picture. The output was set to 10dBm - due to the faulty RPP only -4dBm exit the generator.

Here is one with the generator set to 0dBm amplitude (again, due to the fault, the SA shows less), the difference between fundamental and 1st harmonic is slightly better and higher order harmonics fall off much more quickly.

Fyi:the fundamental IS the first harmonic.

[hp3310a]

I do wonder how one avoids ever subjecting the spectrum analyzer to signals above ref level. I mean, this is something that can happen realistically. The signal is more powerful than anticipated, and the ref level needs to be adjusted after the fact.

The way to avoid it is to adjust your setup so that it can't happen. Set the ref level higher before you ever connect the signal. If there is still some chance that the signal could be more powerful than the safe limit for the analyser, hook it up with an external attenuator in line first, take a measurement, then decide whether you're happy to remove the attenuator.

OK, that is good to know, thanks!

It sounds as if the spectrum analyzer wouldn't be able the handle short periods of slightly too low ref-leveling. Is that really the case?

It depends on the model, it'll usually tell you in the data sheet or user manual but there isn't always much detail. For example, the 8560/8590 series datasheets specify a safe input level of 30 dBm given an input attenuation >= 10 dB. If you set the ref level low enough that it switches to 0 dB input attenuation, then the safe limit goes down (I don't think it actually specifies but I'd assume 20 dBm).

That would put me on the safe side (accidentially) as the attenuation was set to 20dBm (by the SA when I set the ref level to 10dBm). Still, by accident.

If you want to measure harmonic levels using a spectrum analyser, then the usual rule of thumb is to always make sure the harmonic distortion produced by the analyser itself is at least 15-20dB lower than the amount of distortion you want to measure or prove.
In other words, if the sig gen (genuinely) produces a second harmonic at -40dBc then the analyser ideally needs to keep its own internal harmonic distortion 20dB lower (at -60dBc) if you want to keep the measurement uncertainty (due to the analysers distortion) below about +/- 0.9dB.

The analyser should then correctly report the sig gen distortion as -40dBm +/- 0.9dB (ignoring other contributors to uncertainty for now).

So if I understand you correctly: Assuming the signal is within ref level (like in the last picture) it can happen that all the higher harmonics I see could be a product of the spectrum analyzer and not be actually in the signal? That blows my mind a little.

The datasheet can give you some idea as to how much distortion the analyser produces for second harmonic distortion in the form of the 2HI spec. For the 859x series of analysers this is usually in the region of +25dBm for the mixer with 0dB attenuation. However, this is usually a conservative estimate and the 2HI might typically be +40dBm for a 100 MHz test signal. But I'm just guessing here as I've not used one of these old analysers for quite a while.

If the mixer 2HI really is +40dBm at 100 MHz and you want -60dBc internal 2nd harmonic distortion then the mixer input level should be -20dBm or less. So if the sig gen produces +10dBm you would need 30dB internal attenuation selected in the analyser menu. If you opt for the worst case 2HI of 25dBm as given in the datasheet the mixer would need to be driven at -35dBm and this would require >45 dB attenuation ahead of the mixer for an input level of +10dBm from the signal generator.

Thanks for the explanation, really appreciate the detail. I will have to think this over a bit.

[G0HZU]

The last time I tried to explain (on Eevblog) how distortion terms behave (between stages) it caused quite an argument as someone (who seemed to think they were an authority on this stuff) disagreed quite strongly. The main thing to bear in mind is that the harmonic (or IMD) signal from the sig gen and the internally generated harmonic in the analyser will have the same frequency but the phase will be different.

So this can cause the analyser to show a harmonic level that is higher than it should be or lower or it might make no difference as the phase angle might be additive (in phase) or subtractive (out of phase) or anything inbetween. It depends on the relative levels and the phase difference. The person I argued with didn't agree but I have Agilent (and R&S) and many decades' experience on my side of the argument

There's an old app note from Agilent for one of my old spectrum analysers here :

https://www.testunlimited.com/pdf/an/5980-3079en.pdf

On page 21 this stuff is described. The graph in fig 2-17 shows that it's best to try and keep the analyser's internal distortion level about 20dB lower than the expected DUT distortion if the uncertainty of the displayed distortion on the analyser is to remain within about +/- 0.9dB.

[hp3310a]

Thanks for the link, I'll see if I can make any sense of it.

I've taken another screen shot. The difference to the previous one is that I removed the faulty RPP and connected the output directly to the step attenuator assembly - not as a permanent solution if I can help it.

If I didn't know that I know nothing I'd say that it is much better now (-50dB @ 0.0dBm output power vs -32dB @ -10dBm). A little suspicious is the 3rd harmonic being very slightly over the 2nd as well as the 5th over 4th.

[chilternview]

Here's my HP 8648B outputting 100MHz at -60dB into a FPC1500.

No harmonics above the noise level which is 45dB less at -105dB.

Edit: added -40dB signal which gave a 2nd harmonic at about -95dB.

[hp3310a]

Here's my HP 8648B outputting 100MHz at -60dB into a FPC1500.

No harmonics above the noise level which is 45dB less at -105dB.

Edit: added -40dB signal which gave a 2nd harmonic at about -95dB.

How does it look at 0.0dBm? I don't see harmonics at -40dBm or lower (that is above the -90dBm noise floor).

[G0HZU]

Normally, most of the classic RF sig gens from the big manufacturers will try and achieve harmonic distortion levels better than -30dBc. Below about +7dBm they might typically achieve -40dBc to -55dBc.

However, it also depends on the frequency plan the sig gen uses. Many modern sig gens use a BFO/HET/Downconverter system to produce signals in the lower part of the frequency range. When used within this downconverter range, the sig gen will typically produce much lower harmonic distortion.
Above the BFO/HET/Downconvert range the sig gen will either use a divider system or it might even operate within the main (switched?) VCO frequency range or use a multiplier. This will usually produce higher harmonic distortion compared to the BFO/HET/Downconvert range.

The frequency range of the downconverter range varies from model to model.

The Marconi 2018/9 only operates in BFO/HET/Downconvert mode below about 2 MHz.
The Marconi 2024 only operates in BFO/HET/Downconvert mode below about 10 MHz.
The Marconi 2022 only operates in BFO/HET/Downconvert mode below about 62 MHz.
The Agilent ESG sig gens typically only operate in BFO/HET/Downconvert mode below about 250 MHz.
The IFR 341x sig gens typically only operate in BFO/HET/Downconvert mode below about 375 MHz.
The HP/Agilent 8648 sig gens typically only operate in BFO/HET/Downconvert mode below about 249 MHz.

So in the case of the 8648 sig gens I'd expect to see quite low harmonic distortion when used below 249 MHz and below about 0 dBm. A good sig gen model in BFO/HET/Downconvert mode will show distortion better than -50dBc at 0dBm output. Above about 250 MHz the harmonic distortion may be a bit higher. I haven't used an 8648 for quite a while so I can't remember. I've got lots of sig gen models here but not an 8648.

Generally speaking, it's a bad sign if the harmonic distortion exceeds -30dBc as this will notably impact the accuracy of the ALC detector and this will result in less 'flat' levelling vs frequency. So some sig gen models will show degraded levelling at higher output power levels (> +7dBm) because the harmonic distortion will begin to creep up. The diode detector used for the ALC system is typically a single diode and this will be very susceptible to 2nd harmonic distortion. Distortion at -30dBc can begin to show obvious bumps and dips in the levelling accuracy vs frequency. In theory this can be >+/- 0.25dB but often it degrades the levelling by 0.1 to 0.15dB when the harmonics creep up to -30dBc.

[chilternview]

How does it look at 0.0dBm? I don't see harmonics at -40dBm or lower (that is above the -90dBm noise floor).

Still over 50dB down with a -20dB input, but have to turn the SA preamp off else the IF gets overloaded.

Edit: With a 500MHz -20dB input, 2nd harmonic is -61dB so a bit worse as G0HZU says. The 8648 is not the worlds best sig gen but it's not too bad.

[hp3310a]

How does it look at 0.0dBm? I don't see harmonics at -40dBm or lower (that is above the -90dBm noise floor).

Still over 50dB down with a -20dB input, but have to turn the SA preamp off else the IF gets overloaded.

Edit: With a 500MHz -20dB input, 2nd harmonic is -61dB so a bit worse as G0HZU says. The 8648 is not the worlds best sig gen but it's not too bad.

Just noticed your RESBW is 1MHz, try 10kHz. When I use 1MHz there are no harmonics visible at all. The noise floor is on my SA is much higher though at 80dBm.