looking for spectrum analyzer / network analyzer with 1MOhm Input Choice

[Andree Henkel]

High, I´m working as electronic designer at a company designing/building LASERs, mainly used for Microscopy.

We are looking for a spectrum- / networkanalyzer to help in design / test / specification testing of the Lasers and theyr electronics.
Main task will be Laser intensity noise measurement.

We want to do
*determine RMS values of the noise say from 10Hz to 1MHz or 10MHz
*look for lines in the light spectra, to work towards spectral pure laser light sources
*measure inside the electronics to reduce / eliminate spur lines, originating for instance from DC/DC-Converters
*reduce the broadbandnoise of the LASER / its driving electronis

We look for recommendations for Analyzer with 1MOhm Input- additional to a 50Ohm Input, which would allow to measure directly in Circuit, we have 38MHz 1:1 1MOhm Scope Probes from R&S
Or with prebiased photodetektor and termination resistor (for noise measurement at a 15mW Laser and 10MHz Bandwidth typicallly 1kOhm Shunt)

Suitable devices seem HP3588A ; HP3589A ; Anritsu MS4630A

Requirements:
*below 50Hz, so we would find line interference (50Hz is line frrequency in Europe)
*better dow to 10Hz, so we could do Noise measurement starting at 10Hz
*higher than 10MHz is must for the noise / interferece measurement
*the diode type lasers often employ a HF-Modulation somewhere 75MHz to 175MHz; typically at 125MHz to prevent them from Mode-Hopping (frequency-Hopping in Laserfrequency); so upper Frequency of 150MHz, better 200MHz would be nice
*low average displayed noise level
*digital interface (GPIB will do, we have spare USB/GPIB adapters araund)
*tracking generator would be nice, so we could measure frequency response of laser-power control circuit, laser current control circuit, photodetector circuit
*Phase measurement ability would be nice to have phase information for the frequency response of the circuitry above

sorry, got a litttle long.

[chris_leyson]

HP3585A/B has 1Mrg 75 02 50 Ohm input, 20Hz to 40Mhz, 3Hz lowest RBW plus tracking gen output. Also reasonably low phase noise. Trouble is that model is probably nearly 40 years old.

[Andree Henkel]

thanks, yes 3585A / 3585B was my original search result, starting with that I found the 3588/3589 and the Anritsu

[nctnico]

The Anritsu MS4630B is also a good option. The big question is: what is your budget?

[PartialDischarge]

I’ve got the 3589A, great device, like it so much that i got a 3588a just to get spares in case the 3589a breaks. Schematics are available at artekmanuals too

[Andree Henkel]

We want first to look what are the best options specwise.
What I´ve seen so far regarding prices for 3588a/3589a from commercial resellers will fit our budget well. The Anritsu is more expensive.

We found, that we really need to work with real sprectrumanalyzer to not wasting our time hunting ghost-spurs created by Scope FFT.

There seem to be no new Analyzers around with 1Mohm, so we would need additional 1M input preamplifier to any new analyzer; I got quote for R&S FSV3 that start at 10Hz goes to 3GHz and is a demostration unit. I might even make up a case that we should have Budget for this one, which is at 10k€, because its the Spectrum Analyzer our primary costumer uses to evaluate our Lasers. But one of the older units with 1M input probably better fits our needs, we don´t need the upper bandwidth and speed of such a device like the FSV3.

[DaJMasta]

I'd definitely look into active probes, and there are some lower frequency network analyzers with spectrum analysis functionality which could be an option.  Depending on your noise floor/resolution requirements, you could also try for a 12 bit scope.  Not quite the same dynamic range, but a fair bit better than an 8 bit scope, and with some averaging and a powerful FFT (PC based or otherwise) you should get a reasonable dynamic range in that area, but likely at a lower cost.  You're above the frequency range of audio analyzers/DSAs, and your low frequency requirement is the bottom end of what 50 ohm SAs offer, so it's a tricky place for high dynamic range equipment, but if you can deal with a modern scope ADC, then your options open up a lot.

[nctnico]

We found, that we really need to work with real sprectrumanalyzer to not wasting our time hunting ghost-spurs created by Scope FFT.

May I ask which oscilloscope you have tried? The FFT abilities of oscilloscopes varies wildly.

[Andree Henkel]

Several 8Bit Scopes: Keysight DSO-X3xxx; DSO-X4xxx; Yokogawa DLM2xxx
10Bit Scope: Rohde & Schwarz RTA4xxx

attached are three pictures, one taken with Femto Transimpedance Amplifier set to 1kOhm; Spectrumanalyzer FPC1000; the other two with same Amplifier and Scope RTA4004 set to 50Ohm at input; 15mW Laser at 1% - 0,15mW output power; DC-voltage at 50Ohm Termination measured with DMM Keysight 34461A at 20.3mV that correspons to -20dBm.
Task is to get the FFT lines up to 10MHz to 60dB below the DC-Carrier, that is below -80dBm. At 200µs/Div scope gives very good result, but at 500µs/div its result is useless.
The lines come from a DC/DC-converter operating at 2MHz and its harmonics.

The R&S surely has best sensitivity of the scopes listed above, but the Problem whatever Scope: you change time base and / or FFT Parameters and lines in FFT appear and vanish.
Lines appear, that aren´t there, if we look at same signal with spectrum analyzer.

What signal levels we are looking at?
A typical optical power of a low to mid power laser (as far as we manufacture) is 10mW to 20mW.
A biased photodiode converts that with about 0.25..0.75 mA/mW into photocurrent
However we have to specify Signal to Noise Level at say 1% power, not just at maximum power.
So we get at 1%, 20mW and 0.5 mA/mW a DC current of 10µA
We then use transimpedance amplifier (TIA) with 1kOhm and get 10mV DC Voltage; or the cheaper option biased photodiode + shunt resistor 1kOhm with the same result
We can not use higher shunt or TIA value, because this will result in unsufficient bandwidth; same is, we nannot use smaller area photodiode which could provide higher bandwidth, but we work with free beam and have to align the beam onto the diode active area.
With the TIA we can go into 50Ohm Spectrumanalyzer, with the photodiode/shunt we need 1MOhm input on Scope or Spectrumanalyzer, also if we want to search inside the circuit we need 1M or additional special probe / preamp with 50Ohm output.
With my current project it is spec, that at 0.2% of 15mW max optical power the noise is <1% of the 0.2% within 10Hz..1MHz; additionaly at 1% optical FFT lines need to be less than 0.1% form 10Hz..10MHz.

We recently bought above Femto Photo diode amplifier with switchable gain, up to 100MHz Bandwidth and 50Ohm Output, combined with lowcost spectrum analyzer R&S FPC1000 we recently got as addon as we purchased the RTA4004, but that photo amplifier alone costs 2000€ and with the FPC we don´t get low enough to make sure there is no line interference or to measure RMS Noise levels 10Hz to 1MHz..10MHz depending on spec.

The RTA4000 is really good for the RMS Noise Measurement at 0.2% Power with Photodiode + 1K Shunt, DSO-X3000, DSOX-4000 and DLM2000 where of no use because the imternal noise level was to high. Even the low cost R&S RTB2000 does beat those 3 other scopes probably because its 10Bit ADC; but for the FFT the RTA dosn´t work reliable, because I see ghost lines.
Bevore I got the RTA4000 I had to do the RMS noise Measurement using 2 different R&S URE RMS Voltmeters, but they where no longer real reliable.
Maybe, if we run into need to also specify Noise Levels 0.1Hz to 10Hz we will consider a signal analyzer, but currently our costumers don´t require this kind of spec.

[ogden]

Forget about looking for modern & decent spectrum analyzer with 1Meg input. Use SA with 50Ohm input and add active FET (scope) probe. Dunno about HP Keysight, but R&S have all you need including SA->Scope probe adapter, RT-ZA9 that you can use with FET probe such as RT-ZS30.

Adapter with probe:

[David Hess]

High impedance buffer amplifiers are available to use oscilloscope probes with 50 ohm instruments.

https://www.mathews-engineering.com/store/p1/High_Impedance_Buffer_Amplifier.html

There are oscilloscopes which likely meet all of your requirements; I know LeCroy makes some.  They can do network analysis because their FFT function returns the phase response instead of throwing it away.

https://www.edn.com/design/test-and-measurement/4441000/1/Measure-frequency-response-on-an-oscilloscope
http://www.testunlimited.com/pdf/an/55W_8815_2_2009.01.07.10.50.04_2765_EN.pdf

[RFDUK]

Just for interest and to add to the vintage equipment list useful for this app  ....

TEK 7L5 was the HF spectrum analyser trend setter of 35 years ago, commonly used for down converted phase noise measurements. Has a 1M input option.

http://w140.com/tekwiki/wiki/7L5

[Andree Henkel]

The 1M input amplifier is looking good; the R&S probe adapter is looking good as well. However most times they should be used with 10:1 probes, resulting in in my opinion unnescessary signal to noise ratio degradation. But in combination of the amplifier with the DC-Block I probably can also use R&S ZP1X 1:1 1M 38MHz probe; also prebiased photodiode + shunt would work in combination with that amplifier; still at 200€ for diode+shunt + 500€ for Preamp much cheaper than Femto TIA @2000€
Or with the R&S Probe adapter I could probably use R&S RT-ZPR20 power rail probe 1:1 50Kohm 2Ghz; less noise pickup by probe and low enough capacity + high enough input resistance for most in circuit measurements.

since there were no other suggestions we will have to decide:

a) "new" 50Ohm input analyzer; selection: R&S FSV4 10Hz-4GHz 18k€ new / 11k€ demonstration unit if available; +? FSV-B9: Tracking generator for FSV4 + 2k€ additional Femto TIA + 0,5k€ 1M Preamp

b) used HP3589A spectrum/network analyzer 10HZ-150MHz with 50Ohm/1Mohm; tracking generator included; about 4k€ with new calibration

I just requested prices from R&S for the FSV-B9 option, the Probe adapter and the powerrail probe.
The Powerrail probe I´ll likely try to get into budget anyway to work with the RTA4000.

We will not consider buying another scope from another vendor. We now have 2 Keysight, 1 Tektronix, 1 R&S scope in engineering, 1 R&S and a lot of yokogawas in manufacturing and we will not add Le Croy to this mix. Reason for this:
*each vendor is cooking its own soup with the proprietary probe interfaces, so each special probe we only can use on 1-2 scopes
*we have enough scopes around, what we need is spectrum analyzer / networkanalyzer capabilities without the possible Problems shown, we just want to make sure, we only see lines that really are there
*the 2 Keysight and the Tektronix have been around here for years. The yokogawa preference in manufacturing is based a lot on that special optical measurement equipment is bought from yokogawa.
*the RTA4000 was bought recently because it provides lowest noise level of the current R&S scopes and true 0.5mV/Div range setting; regarding FFT capabilities RTE or RTO would have been better choice but they are more noisy. We also had a 10Bit S Series of Keysight here for demo, but it was not as good regarding noise, but more expensive compared to the RTA. Maybe the 12 Bit Le Croy and the Tektronix 12Bit MSO6 are comparable with the RTA regarding noise.

[joeqsmith]

The 3589A is getting up there in age.  You should consider reliability and service if you are planning for the long term.

I have owned one for several years and had to pull it apart over the summer to do some maintenance.  I use it a fair amount and over all it has been very reliable.   Basic cleaning, had a connector come loose once,  replace the battery has really been about it.   Using LabVIEW with it is a snap. 

I don't have the S-parameter set but have used it with an external coupler.  There are some more options available when using the test set.

While it has the ability to select a 1M input, I seldom use it.  Sometimes my projects get out of hand so normally I play it safe and buffer the input.   

[rhb]

The FFT "spurs" may be self noise of the DSO.  I've been investigating that the last few days.  Set the instrument to the minimum V/div setting, stick a 50 ohm terminator on the input  and take a look at the FFT.  Arguably a 1 Meg termination would be better, but I didn't have one.

Most of the scope makers seem to be unwilling to hire anyone who actually knows how to compute a proper FFT.  GW Instek have a pretty good  one in their spectrum analyzer app  for the MDO, but the instruments have significant self noise and terrible shielding.  The FFT in the other members of the GDS-2000E line is usable, but just barely.

The LeCroy DDA-125/LC684DLX FFT is also properly done, but not very flexible.  But instrument self noise is much lower than the Instek.

I'm on an EMI witch hunt and finding that even a Keysight 33622A emits a ridiculous amount of noise from the LCD.  I find this really weird as I can detect none from my computer monitors which are 15 years old.

[Performa01]

The FFT "spurs" may be self noise of the DSO.  I've been investigating that the last few days.  Set the instrument to the minimum V/div setting, stick a 50 ohm terminator on the input  and take a look at the FFT.  Arguably a 1 Meg termination would be better, but I didn't have one.

Can you quantify the spurs you're seeing - or even show a screenshot?

Is there a significant difference between terminated and open input?

[Andree Henkel]

Hy, I have some questions regarding the differences between HP3588A and HP35889A

I understand, that the xxxx9 is a network analyzer.
It does complete S-Parameter (it delivers all 4 complex numbers s11,s12,s21,s22) measurement, provided you combine it with the external S-param testset.
As I understood the base unit includes tracking generator, choice of 1M input and capability to measure complex s11 including phase information.
It then can display phase plots, group delay plots and smith chart for s11

The capabilties of xxxx8 are somewhat unclear to me. Different sources of info are partially contradictory; so basically I have 3 questions
Q1: does it have 1M input choice also?
Q2: does it have a tracking generator?
If Q2 is yes
Q3: can I measure phase of s11?; I sure can measure amplitude of s11 (even without tracking generator, but then I would need additional white noise source)

[PartialDischarge]

Hardware-wise both are the same, except for the CPU board. Both have 1M input, and tracking generator, but the 3588 will not make log-sweeps and will not give you phase info, so other options like polar, smith chart and Re-Im are also missing. The menus are not exactly the same also, they have tweaked some things that appear on the x9 and not on the x8

Just for the log-sweeps the 3589 makes a difference, some images of it, filters and a 10000x resistive divider. It can be connected via GPIB using this utility (http://www.ke5fx.com/gpib/readme.htm) although I've been too lazy and still rely on pics

[Andree Henkel]

Thankyou, so well yes in case we go for the vintage unit, I´ll go for the xxxx9. While the xxxx8 would be all needed for noise measurement and spur detection, the xxxx9 could be really helpful if we are investigating things like stability issues caused by insufficient phase margin in current control / power control loops. The used xxxx9 aren´t that more expensive than the xxxx8.

I´ll present both variants the vintage and the new FSV4 to my teammates and manager of development, then we will decide which way to go in budget planning.

For short term I´ll order the 1M buffer amp linked to work with the R&S FPC1000 low cost Spectrum analyzer we have already.

[Andree Henkel]

The FFT "spurs" may be self noise of the DSO.  I've been investigating that the last few days.  Set the instrument to the minimum V/div setting, stick a 50 ohm terminator on the input  and take a look at the FFT. 

I doubt, that the spurs shown in one of the 2 Screenshots of my RTA are caused by internal noise sources, if so they should likely show up in both screenshots. Even worse the actual existing spurs are missing in one screenshot while the other displays them correctly. The only change in setting I actively do between the two pictures is changing the timebase.
So yes, in my opinion more likely the origin of the problem in case of the RTA is some bug in the FFT implementation. Changing timebase sure has effect on the aquisition process and the intertwined parameters off FFT processing. But correct FFT implementation needs to take care for that.
All other scopes around here are more noisy than the RTA.

I´ll try to update Firmware and look if that helps. Result of that I´ll report here. If it doesn´t help I´ll report the issue to R&S.

[aaronlow]

I know this is a very old topic, but the ideal instrument for this (even though it is discontinued) is the Rohde & Schwarz FMU36 Baseband analyzer.  It has a 30 MHz bandwidth, a selectable 50 or 1M ohm input impedance and can measure I, or Q, I/Q or balanced I and/or Q.

I am sending this because my FMU36 has failed its calibration and R&S cannot repair it anymore.