R&S SML01 1.1Ghz Generator

[Martin72]

EDIT: Title changed.

Hi,

I would like to buy a 1Ghz generator, but currently only to determine bandwidths of oscilloscopes, nothing else.
So I don't want to buy a new one, that would be too expensive for this one application.
So I looked for used generators and I found one:

https://www.aimtti.com/product-category/legacy-products/aim-tgr1040

I could get exactly this model used for under 400€, because it is supposed to be technically flawless, but the display has a small quirk and the transformer hums.
Neither is a problem for me, from the side everything is clear.
What I want to know is if this generator is good enough for what I want to do.
I do not trust the whole even cheaper China stuff over the way - Or is that a misjudgement ?
What do you think, buy this generator or is it cheaper ?

[nctnico]

2dB level accuracy is nothing to write home about. You can find an R&S SMIQ which has much better performance for about the same amount of money with some patience.

[Martin72]

1050€ minimum currently and huge dimensions..It should be space saving I forgot to say.

2dB level accuracy

Hm, OK...For determining the -3dB frequency this is worse.

[TurboTom]

The price for a discuntinued RF signal generator of these specs is borderline acceptable, but I doubt that the money is well spent for the application that you have got in mind. Amplitude accuracy over the span is specified to be +-2dB. In order to properly characterize oscilloscopes even half of that accuracy appears to be too much.

A plain adjustable sine wave oscillator in combination with an RF level meter may be a more accurate approach, though it may be more troublesome to put to use.

Okay, @nctnico beat me on that one . But I may add that some Spectrum analyzer tracking generators may not be much worse suitable for that application. So if you have got an SSA3000 (Plus) or DSA800TG series device, these may provide a suitable test signal as well.

[tautech]

Hi,

I would like to buy a 1Ghz generator, but currently only to determine bandwidths of oscilloscopes, nothing else.
So I don't want to buy a new one, that would be too expensive for this one application.

Profile:

Siglent SDS2504X HD, Siglent SDG2122X, Neutrik A1, Hioki 3532-50 LCR Meter, Brymen BM869s,Siglent STB-3, Siglent DF2001A, UT 210-E, Hioki 3450 Milliohm Meter, Andonstar ADSM302, several DC-Supplies and other things.

No analyzer to set a zero span and use its TG ? 

[nctnico]

1050€ minimum currently and huge dimensions..It should be space saving I forgot to say.

Lowballing has served me pretty good every now and then.

[Martin72]

@tautech:

A R&S at work...

SA is on my list for my home equipment.

[nctnico]

Hi,

I would like to buy a 1Ghz generator, but currently only to determine bandwidths of oscilloscopes, nothing else.
So I don't want to buy a new one, that would be too expensive for this one application.

Profile:

Siglent SDS2504X HD, Siglent SDG2122X, Neutrik A1, Hioki 3532-50 LCR Meter, Brymen BM869s,Siglent STB-3, Siglent DF2001A, UT 210-E, Hioki 3450 Milliohm Meter, Andonstar ADSM302, several DC-Supplies and other things.

No analyzer to set a zero span and use its TG ? 

Won't do much good as tracking generators typically aren't levelled.

Something like this could be interesting:
https://aaronia.com/en/shop/signal-generator/battery-powered-signal-generator

[tautech]

Hi,

I would like to buy a 1Ghz generator, but currently only to determine bandwidths of oscilloscopes, nothing else.
So I don't want to buy a new one, that would be too expensive for this one application.

Profile:

Siglent SDS2504X HD, Siglent SDG2122X, Neutrik A1, Hioki 3532-50 LCR Meter, Brymen BM869s,Siglent STB-3, Siglent DF2001A, UT 210-E, Hioki 3450 Milliohm Meter, Andonstar ADSM302, several DC-Supplies and other things.

No analyzer to set a zero span and use its TG ? 

Won't do much good as tracking generators typically aren't levelled.

Of course, that's why analyzers have a Normalize function.

However the analyzer typically has far better accuracy than TG where if we are to worry about TG flatness for zero span tasks a marker dropped on the frequency of interest will display the amount of level correction one needs to make.
There are ways around many inconveniences. 

SA is on my list for my home equipment.

Consider the above. 

[nctnico]

And now use the TG output to measure the amplitude flatness of an oscilloscope which is the intend of the OP. How is normalisation going to help there? Not to mention the harmonics a TG output may have that will show up on an oscilloscope.

[tautech]

And now use the TG output to measure the amplitude flatness of an oscilloscope which is the intend of the OP. How is normalisation going to help there? Not to mention the harmonics a TG output may have that will show up on an oscilloscope.

It's late there so maybe you need some rest to think clearly about this.

You can't use Normalization on TG for use other than with the analyzer but we can measure TG unflatness with the analyzer and account for these errors/inaccuracies for other use needs.
TG can be used as a fixed (zero span) or sweeping frequency for other than analyzer needs and at levels within TG amplitude capabilities.

[nctnico]

That is rather cumbersome and still doesn't deal with harmonics. For all you know the TG output looks like a square wave. That doesn't matter so much for a spectrum analyser itself as this is basically sweeping a narrow filter over the input signal to determine amplitude versus frequency, but if you want to look at the bandwidth on an oscilloscope you will want to have a signal that looks like a single tone.

Getting some form of a real RF generator is the easiest and most straightforward path.

[TurboTom]

@nctnico - that's actually not true. Admitted, especially at low frequencies (sub-10MHz range), there are very strong harmonics present in most TG output signals, but same is true for many RF signal generators. This is the result of the output amplifier not really being designed for these low frequencies. Since the signal is synthesized by mixing two pretty decent sine wave signals, the difference of those shouldn't (and doesn't) contain that much harmonics, hence I'ld consider a TG signal (harmonics-wise) suitable for bandwidth testing of an oscilloscope, at least much better so than one of these ADF4351 based "el-cheapo" sources. But if it's comfortable to use a TG for such an application, is written on another page...

A hint for the OP: Once in a while, an R&S SM300 can be found inexpensively (though usually with a broken PA which is quite fixable). These generators perform much better than the initially mentioned AimTTi instrument and are very well suited for other test jobs as well. Yet, you've got to be patient and put some work into the device...

[tautech]

I'll just leave these here for discussion.

[rsjsouza]

For cost reasons at the lab we usually don't calibrate or rely on a signal generator amplitude accuracy - instead, we have calibrated SAs to verify and eventually normalize the output of any signal generator. Although a bit more cumbersome as it involves an additional verification step, it tends to work well. And this is true even for the SMIQs we have around.

Good luck in your decision!

[nctnico]

For cost reasons at the lab we usually don't calibrate or rely on a signal generator amplitude accuracy - instead, we have calibrated SAs to verify and eventually normalize the output of any signal generator. Although a bit more cumbersome as it involves an additional verification step, it tends to work well. And this is true even for the SMIQs we have around.

That is also a good way. You can load calibration tables into the SMIQ for this purpose (and automate the process if you want) but my guess is that this is more to compensate for dampening from cables rather than improving output accuracy. The SMIQ is specified for 0.5dB output level accuracy up to 2 GHz which is pretty good.

[Martin72]

Hi,

OK, the Aim tti is "dead", then I´ve looked for some else cheap and spacesaving generators, but no way...

First this cought my attention because of the claimed 0.5dB precision:

https://eleshop.de/rf-explorer-signal-generator.html

But in the specs it will be relative :

https://static.eleshop.nl/mage/media/downloads/RFExplorer-6G_Signal_Generator_datasheet.pdf

This one here(german language):
https://scdn.rohde-schwarz.com/ur/pws/dl_downloads/dl_common_library/dl_news_from_rs/165/165_sml_de.pdf

Sounds promising, it´s not too big and you can get it used for under 1000 bucks.

[G0HZU]

To do this properly, you have to minimise all contributors to uncertainty. The classic way to do this would be to use a self levelling system as drawn below. The level accuracy of the sig gen is improved to be similar to the accuracy of the power meter and as long as the splitter is connected directly at the input of the scope, there are no cable losses to worry about.

The other major contributor to uncertainty is due to mismatch from the sig gen mating to the device under test. With the system below, the source VSWR is defined 'only' by the quality of the 50 ohm resistor in the splitter that feeds the device under test. So as long as a precision splitter is used, you can get ultra low source VSWR at 1GHz for very little cost. The splitter can be homebrew. This reduces mismatch uncertainty to negligible levels.

The sig gen either has to have the option for external ALC or it must have external AM that is DC coupled and this allows an external leveller amp to be used.

I'm not suggesting you go out and buy a 83752A sweeper or an expensive power meter or HP 11667 splitter, but this system represents the ideal way to do this test.

You can make the splitter for a very low cost using a pair of 50 ohm resistors (Hint: use two 100R SMD resistors in parallel for each 50R resistor), and any mid range 1GHz sig gen that supports EXT AM with DC coupling can be used. The power meter could be replaced with a decent diode detector as long as the sig gen has low harmonic content. This would save even more cost. It's still going to be way better than anything else suggested so far.

[Bud]

Have you considered using a pulse gen and derive the DUT scope bandwidth from rise time?

[Martin72]

Hi,

Yes....

https://www.eevblog.com/forum/testgear/math-problems-on-sds2k-(trying-to-display-bandwith)/msg4972270/#msg4972270

[G0HZU]

If it helps, I can list a few things that can add to the overall uncertainty if you simply buy a decent 1GHz sig gen that has had a recent calibration and use it to test the 1GHz scope.

The first one is level uncertainty. This might be as good as +/-0.1dB (typical) up to 1GHz but it could be +/-0.2dB. It could be as bad as +/- 1.5dB for an older mid range sig gen that has an upper limit of 1GHz. This is because the manufacturer will opt for a more basic (cheaper) ALC system that will work well up to about 700MHz but then degrade a bit towards 1GHz.

Therefore, it's best to buy a sig gen that has optional versions that can go to several GHz as they will probably all share the same ALC (levelling) system that will be good to several GHz. The Agilent ESG series has very good flatness for example and these seem to start at about £600 on ebay.

Then there is mismatch uncertainty between the source and the load. The source VSWR of your sig gen at 1GHz at (say) +6dBm might be 1.3:1. The input VSWR of your 1GHz scope might be 1.5:1 at 1GHz. This contributes another +/- 0.23dB uncertainty to the overall system.

Then there is the loss in the connection cable and this could easily be 0.5dB at 1GHz for a decent RF cable with low VSWR.

This stuff soon adds up and a lot depends on what you consider to be acceptable.

[nctnico]

Have you considered using a pulse gen and derive the DUT scope bandwidth from rise time?

That won't work because it assumes a certain roll-off factor that isn't fixed on oscilloscopes. The only way to determine the bandwidth of an oscilloscope is to use a levelled RF generator.

[rsjsouza]

Then there is the loss in the connection cable and this could easily be 0.5dB at 1GHz for a decent RF cable with low VSWR.

I haven't tested the bandwidth of an osciloscope, but on the VNA the choice of adapters and connectors can add a few fractions of a dB here and there - especially if you happen to grab a random cable and/or adapter from the lab's shared drawer and waste quite some time trying to figure out why the darn response seems off...

That and curves and kinks on the cable, whose influence is much more severe on a sensitive equipment such as a VNA.

[nctnico]

At frequencies over 1GHz you need to check the specs of the cable and use really short cables as well. I have a few short Huber+Suhner RG223 cables I use for measurements where cable loss must be low.

[Performa01]

I don't have a specific recommendation, but a few general hints:

• Testing the bandwidth of a scope is not a precision task in my book. Even an error up to 1 dB would not really matter that much. Take for example my measurement of the SDS2504X Plus, where a 1 dB error caused by the cabling accounts just for the difference of 570 to 605 MHz. Are we really that desperate to brag about the bandwidth of our darling scopes that such a small difference matters?
• As others have already hinted on, you want a sine, not a square as its output by some cheap toys. Dead giveaway: if there are no specifications for the harmonic distortion, then don't touch it!
• Get some 10 dB inline attenuators so you can optimize the port impedances. A quality signal generator will have a good VSWR as long as its attenuator is active (i.e. you don't demand the full signal amplitude), but the scope input VSWR might be as bad as 1.5:1 at the bandwidth limit and far worse beyond that. The resulting error is not huge, but the measurement trace looks prettier if it has no wobble in it
• A quality signal generator is "levelled", i.e. its output amplitude is measured and controlled in a way to make it as constant as possible. Tracking generators are not. You can build your own precision levelling circuit as pictured by G0HZU, but this requires quite some gear and you'll probably lose the ability to sweep the signal generator - and therefore the simple way to let the scope plot its own frequency response. Real automation might be no fun with ancient gear that only provides GPIB (or nothing at all), and it most definitely is a pain if you have to do it by hand...