Teardown, Repair & Analysis of a Rohde & Schwarz FSH3 3.0GHz Spectrum Analyzer

[Hugoneus]

In this episode Shahriar examines a faulty R&S FSH3 100kHz – 3.0GHz Portable Spectrum Analyzer. This exceptionally dirty unit does not power on and is missing a power supply and battery. After verifying the correct operation of the battery charger circuit and keypad the problem is traces to a damaged SOT-23-6 MOSFET device which is part of a fly-back DC-DC converter. A replacement MOSFET with a lower current rating proves to be a failure as the device fails on startup. A higher rated MOSFET resolves the problem and device begins to operate correctly.

The block diagram of the system as well as a detailed block diagram of the RF section is presented and correlated with the teardown of the RF section. The super-heterodyne down-conversion circuits and components are examined.

Watch the video here: [51 Minutes]
youtu.be/qD05ZFVplE0

More videos at The Signal Path:
http://www.TheSignalPath.com

[dr.diesel]

I'd be tempted to fab a custom metal bench enclosure.

Excellent video as always.

[ElectroIrradiator]

Ah, you posted the video in here. Your first replacement MOSFET blew up because it was a P channel device.

[Hugoneus]

Ah, you posted the video in here. Your first replacement MOSFET blew up because it was a P channel device.

I noticed that I actually showed the wrong datasheet I think. I believe that the part that I used was a N-channel, but rated only to 400mA and not 1Amp as the PMOS datasheet indicated.

[tec5c]

Another great video Shahriar. It's always the highlight of my day/week/month when a new video is posted up on the signal path.

Keep up the good work. (insert thumbs up emoticon as Windows 10/Microsoft Edge doesn't seem like it's capable!)

[PA0PBZ]

I'm most puzzled by the word "naamplaat" on the keyboard foil, which I can find no other meaning for than the Dutch word for nameplate...
A great video as always, interesting to see the noise generator just before the ADC in the block diagram, I guess that is an attempt to dither the signal.

[Wuerstchenhund]

Great video, as always! 

The MOSFET problem seems to be a regular issue with the FSH3, which btw was the first generation of handheld SAs from R&S. The FSH Series is still made, however from what I know the hardware for the old FSH3 and the newer FSH4 and above is different (the FSH3 was known for being pretty slow and having occasional stability issues, the bigger/newer models are supposed to be improved).

[dadler]

Which Mosfet did you end up using in the end?

[Hugoneus]

Which Mosfet did you end up using in the end?

I ended up using a SI3454.

[German_EE]

I know that I have asked this before,  but is there any chance you could do a short video on some of the weird PCB-based structures that we see on microwave PCBs? There are comb like areas, others shaped like fantails and lengths of PCB that seem to do nothing but I'm sure have a purpose in life.

My knowledge stops at 500 MHz or so and even then I'm struggling so a little help with the 'RF Voodoo' would be appreciated by your viewers.

[Melt-O-Tronic]

Another brilliant video!  I am learning so much, thanks to you and W2AEW.  Lately, nothing makes me happier than a new video from either of you guys, so your increased rate of production lately is profoundly appreciated.   

To tag on to German_EE's post, I'm particularly curious as to WHY distributed element circuits do what they do.  Since I read Walter Maxwell's explanation of transmission line reflections in his book, "Reflections III", the light clicked on and I'm finally beginning to understand RF (and that's why you and W2AEW are so important to me now).  So I think understanding the "mechanics" (physics) of distributed elements is another opportunity for revelation.

Also, please give Pooch my regards, plus an extra scratch behind the ear.   

[RRobot]

I know that I have asked this before,  but is there any chance you could do a short video on some of the weird PCB-based structures that we see on microwave PCBs? There are comb like areas, others shaped like fantails and lengths of PCB that seem to do nothing but I'm sure have a purpose in life.

My knowledge stops at 500 MHz or so and even then I'm struggling so a little help with the 'RF Voodoo' would be appreciated by your viewers.

The Wikipedia article is a good starting point:

https://en.wikipedia.org/wiki/Distributed_element_filter

[AndreasF]

   Excellent video!

I loved your expression of joy/surprise when it powered on for the first time (probably because you didn't expect it to turn on).

[Hugoneus]

Thank you all. I do want to make more tutorial videos but I do have some important reviews which I need to complete. There is also typically a lot of tutorial style material in my reviews which I hope would be useful.

[Hugoneus]

Also, please give Pooch my regards, plus an extra scratch behind the ear.   

Will do!

[Dubbie]

Everything you post a new video I get a big groan from my wife because I like to watch it on the TV in the living room. It's funny though, for some of the repairs she even exclaims "yay" when the thing turns on for the first time. Don't stop making videos. I learn a lot from them.

[Hugoneus]

Everything you post a new video I get a big groan from my wife because I like to watch it on the TV in the living room. It's funny though, for some of the repairs she even exclaims "yay" when the thing turns on for the first time. Don't stop making videos. I learn a lot from them.

Thanks! This one had a yay moment from myself.

[G0HZU]

Sorry to be a neghead but... I wasn't very impressed with the signal path walkthrough parts of this video...

I skimmed over the repair bit to get to the part I was most interested in which was the block diagram and internal view of the RF PCB. I didn't think this was presented that well because there were lots of incorrect things said. I guess none of this really matters but the PCB walkthrough was especially poor because Shahriar got the description/location/roles of the various mixers and filters mixed up.

I can glance at that PCB and see the correct signal path very easily including the unused sections for the 6GHz version on the other side of the PCB.

A couple of other things to note. At 46:36 the loop BW of the PLL will be chosen at 30kHz mainly for phase noise but also for the tuning/startup/lock speed of the PLL.

At 46:40 the 'VCO 3X' doesn't refer to a frequency tripler or overtone. To put a tripler here would not be a good idea... 'VCO 3X' refers to a bank of 3 discrete VCOs that get selected as required to cover the total 4-7GHz range of the first LO. It looks like they used printed resonator designs. They use three VCOs here because it gives the best phase noise performance compared to trying to do this with a single VCO. This is a fairly common approach. Also the design is technically easier/safer than trying to cover this range with a single VCO. The downside is the added circuit size/complexity and the associated management and startup/locktime issues as it sweeps the full range in these three chunks. It's easy to spot these three discrete VCOs in the PCB layout as they take up a lot of space. eg see the image below where they are arrowed/circled in blue.

[G0HZU]

Dunno if this helps but here's a crudely marked up image showing the location of the filters and mixers in the basic RF path for the 3GHz range.

Obviously, I don't work for R&S and this is the first time I've seen inside one of these but I think my marked up image below shows the locations/descriptions better.

[G0HZU]

I can also describe the signal path for the 6 GHz option if anyone wants to see it?

[rs20]

I can also describe the signal path for the 6 GHz option if anyone wants to see it?

Thanks for the picture with all the different blocks overlaid -- it's nice to make observations like seeing the DC connection on the LPFs, and the lack of a galvanic connection for the BPFs (as you would expect), and the 3GHz LPF having features roughly double the size of the 6GHz LPF, etc etc. It's nice to feel like one has a clue what's going on!

I'm always happy to see more microwave magic with schematic overlays!

[Hugoneus]

G0HZU,

Thanks for the analysis. As I pointed out on the video overlay, this was a very quick overview of the board. I spent less than 30 second looking over it as I was in a rush to close it up. A thorough overview of the board was not the intention of this video.

Making YouTube videos is not my full time job, time is a major constraint. When I do a full video on microwave design I'll make sure to go through everything in detail.

By the way if you are interested in seeing VCO's with 2nd harmonic or tripplers I have built a few all the way up to 160GHz. In fact, banks of VCO's with/without multipliers to cover nearly 50GHz of tuning range.

[G0HZU]

I guess it's easier for someone like me to instantly spot these issues as I have many years' experience designing converters using filters just like these and in similar frequency bands and for similar applications. But I'm still very surprised that despite 6000 views on youtube nobody else pointed out that the PCB walkthrough got the various mixers and filters mixed up. These are classic elliptic LPF and interdigital BPFs. I think the 831MHz IF2 filters will be Murata SAW filters rather than cavity filters.

By the way if you are interested in seeing VCO's with 2nd harmonic or tripplers I have built a few all the way up to 160GHz. In fact, banks of VCO's with/without multipliers to cover nearly 50GHz of tuning range.

I think quite a few people would be interested in seeing info about oscillator design up at those frequencies

However, I don't think it would be a good idea to use a VCO + tripler to generate 4 to 7GHz for the first LO on this type of analyser as it would demand so much more PCB space for the required filtering and also it would generate a lot more internal spurious issues and also the close in phase noise of the synthesiser would be higher with a tripler approach.

[Hugoneus]

I guess it's easier for someone like me to instantly spot these issues as I have many years' experience designing converters using filters just like these and in similar frequency bands and for similar applications. But I'm still very surprised that despite 6000 views on youtube nobody else pointed out that the PCB walkthrough got the various mixers and filters mixed up. These are classic elliptic LPF and interdigital BPFs. I think the 831MHz IF2 filters will be Murata SAW filters rather than cavity filters.

I agree. I am an ASIC designer myself and a quick glance at a layout gives me a lot of information.

I think quite a few people would be interested in seeing info about oscillator design up at those frequencies
However, I don't think it would be a good idea to use a VCO + tripler to generate 4 to 7GHz for the first LO on this type of analyser as it would demand so much more PCB space for the required filtering and also it would generate a lot more internal spurious issues and also the close in phase noise of the synthesiser would be higher with a tripler approach.

I also agree that a trippler is not a good idea on this board (and was surprised myself when I thought it was implemented). The mm-wave and THz oscillators have to use multipliers as it is not possible to reliably make fundamental VCOs at those frequencies with devices which have fT/fMAX <300GHz.

PS: I am shooting a VNA repair which I think you will like.

[Melt-O-Tronic]

VNA repair . . .

Getting a better understanding of how VNA's function is exactly what I've been wanting to learn.  Can't wait.  Feel free to make a 12+ hour video with plenty of detail if you want.   

[Adal]

Hugoneus Thanks for your video!
I have a FSH-3 that has the same damage that you show. Unfortunately mine have the PCB carbonized. I attach some pictures of my machine in order to ear your suggestions for fix my spectrum analyzer. For the service info, I use the FLUKE scopemeter 190, they use the same  battery charger power control.
Best regards. 

[tonyM]

Hello everyone,

I have TSMU unit, which has the same RF-block as FSH3.

It is broken, i.e. it can't measure correctly. I will be very gratefull, if someone help me to repair it.

My TSMU-RF-board PCB is older than FSH3, just a little difference at the input (other SPDT RF-switches=AS179-92LF, SMD marking S79) and other/older TCXO. Attenuator IC is also missing.
2xRF-switches + 1xBFP520 are making the circuitry of Pre-Amplifier, placed between missing Attenuator-IC and 3GHz-LPF. Pre-Amplifier can be switched ON or OFF/by-passed by the software.
RF-transistors used are: BFP420 (SMD marking AMs) and BFP520 (SMD marking APs).

Below is the progress until now on my attempt to repair the unit.

Due to the info, found in FSH3/6 service manual and also in this thread, I've made some investigations and measurements. I think there is a problem with PLL-synthesizer: it can't lock correctly.

I made that decision by measuring the signal at the output of IC=HMC434 (divide-by-8 module, placed between 1st-LO and SYNCON-synthesizer, i.e.
between 2xBFP520-transistors and IC=uPB1509GV/divide-by-2 or 4 or 8 module) and the signal at the output of IC=HMC433 (divide-by-4 module, placed between 2nd-LO-3200MHz (1xBFP420-transistor) and 3rd-LO-800MHz (1xBFP520-transistor).

The signal at IC=HMC433 output is OK: exact 800 MHz, with strong level. But the signal at IC=HMC434 output is not OK.
Due to RF block diagram, 1st-LO-frequency is between 4031...7031 MHz, and after HMC434 it must be between 503...879 MHz.
So:
1. if I want to measure signals with frequency=925...960 MHz (=signals from BTS->Mobile at 900MHz-band), 1st-LO-frequency after HMC434 must be between 619 and 624 MHz. In my case this is not true, I don't observe increase in levels of this spectrum piece.
2. if I want to measure signals with frequency=2110...2170 MHz (=signals from BTS->Mobile at 2100MHz-band), 1st-LO-frequency after HMC434 must be between 767 and 776 MHz. In my case this is not true, I don't observe increase in levels of this spectrum piece.

For the measurements, I'm using Tektronix NetTek Spectrum Analyzer YBT250, with frequency range 30...2500 MHz.

I also found, that 1st-LO is made by 3xVCOs: 1xBFP420, 1xBFP520, 1xBFP520, and each one is responsible for a spectrum piece of about 1000 MHZ (VCO-1=4000...5000, VCO-2=5000...6000, VCO-3=6000...7000 MHz). Each VCO is switched-on by applying Vcc to respective BFPx20-transistor, and the exact frequency depends on Vtune, applyed to the varicaps (a pair for each VCO, Vtune rail is common for all VCOs). I observed the above facts by using ROMES software to measure at 3 different frequencies: 500, 1500, 2500 MHz, which in turn pulls into operation the respective VCO. Scanning with ROMES software at full band 15...3000 MHz didn't give any increase of levels for spectrum 503...879 MHz at HMC434-output.
It seems that VCOs are working (Vtune is ramping to about 20V when full band 15...3000 MHz is selected for measuring), but to be 100% sure, I need spectrum analyzer with higher bandwidth- at least up to 7...8 GHz.
So, at this moment, I don't know if the problem is in HMC434.

The only thing, which I can try at the moment, is to swap the places of HMC434 and HMC433. In such way, I can test if HMC434 works properly: at HMC433's place (after 2nd-LO-3200MHz), the output of HMC434 must be 400MHz.

Any suggestions?

[tonyM]

I just swapped HMC434 and HMC433.
At HMC433's place (after 2nd-LO-3200MHz), there isn't any strong 400 MHz-signal at the output of HMC434.

So, the problem is in HMC434.
Do you agree with me?

[tonyM]

I just swapped HMC434 and HMC433.
At HMC433's place (after 2nd-LO-3200MHz), there isn't any strong 400 MHz-signal at the output of HMC434.

So, the problem is in HMC434.
Do you agree with me?

Just changed HMC434.
Result: my TSMU is fully operational again

[dianzimi]

I have a fsh3 ,it is not work and missing  some resistence  and  capacity  ，show the picture，can you tell three resistence  and one capacity  value   ???thank you
I can not load the picture ,can you tell the email??my email is dian.zi.mi@163.com

[hgl]

In 2018, I looked into the R&S TSMU and also saw the excellent FSH3 teardown video.
A German dealer had gotten a whole pallet of them from a bankruptcy and offered them for about 25€ on Ebay.  I then bought 2 and started a thread in German microcontroller forum.

https://www.mikrocontroller.net/topic/454625

The RF boards of TSMU and FSH3 look the same but the assembly is different.
The rest of the construction is completely different.

[tonyM]

I have a fsh3 ,it is not work and missing  some resistence  and  capacity  ，show the picture，can you tell three resistence  and one capacity  value   ???thank you
I can not load the picture ,can you tell the email??my email is dian.zi.mi@163.com

First, help us identify PCB version of your FSH3.

My TSMU has older PCB, slightly different from FSH3. It has:
- missing Downconversion section for frequencies above 3GHz on the back side of the PCB
- missing Attenuator at the input signal path
- slightly different input signal section
- different model of 10MHz-TCXO


Newer TSMU PCB is the same as FSH3 PCB. It has:
- Downconversion section for frequencies above 3GHz on the back side of the PCB, but without components (populated in FSH6)
- missing Attenuator at the input signal path



FSH3 PCB is the same as newer TSMU PCB. It has:
- Downconversion section for frequencies above 3GHz on the back side of the PCB, but without components (populated in FSH6)
- Attenuator at the input signal path



All above PCB versions have the same Tracking generator part on the front side of the PCB, where the missing parts have to be placed:


Please check and identify your PCB.
And also provide the numbers on the front and on the back side of your PCB, marked by the blue rectangles in the above TSMU-ver.2 pictures.

Where are the missing parts on your PCB? At the input signal section (marked in the above pictures as "Pad")?

[pdenisowski]

The FSH3, which btw was the first generation of handheld SAs from R&S.

The FSH3/6 series was a very popular product: one even made a trip to the International Space Station (although it did have to be slightly modified - see link to article)

https://www.rohde-schwarz.com/file/n184_fsh3.pdf

The FSH Series is still made, however from what I know the hardware for the old FSH3 and the newer FSH4 and above is different

Yes, that's correct.  In addition to the FSH4/8/16/20 we also currently manufacture an even newer (!) portable spectrum analyzer, the FPH.

[dianzimi]

I want  to  know  Attenuator component   value  at the input signal path

[tonyM]

I want  to  know  Attenuator component   value  at the input signal path

Do you mean the components in the red rectangle bellow?

[dianzimi]

yes

[dianzimi]

yes

[tonyM]

I think these 3 black parts (excluding input capacitor) at FSH3 input signal path are resistors, which are making Pi-pad attenuator.

As you can see, TSMU (ver.1 and ver.2) lacks Pi-Pad attenuator. Only one Zero resistor (see pictures below):



If you put Zero resistor in the right place (as in TSMU-ver.2), I think your FSH3 will work (if next components/ICs are OK), but the measured levels will not be correct.

As I don't have FSH3, I will propose the following experimental approach for calculating the Pi-pad resistors values (and calibrating the instrument):
1. Put zero resistor at the input signal path
2. Plug-in signal generator with known level at RF-input
3. Check the level, measured by FSH3, and calculate the difference in dB
4. As you know the difference (and input impedance=50 ohm), use calculator https://www.pasternack.com/t-calculator-pi-attn.aspx to find the values of the resistors.

[hgl]

There is an interesting detail, the pin 2 of the synthesizer chip is isolated on the TSMU version 2. (Not properly visible in the photo, you can see only a small piece of brown foil under the pin). This is not the case with version 1 and FSH3.  On the photos you can not see whether the synthesizer chips of FSH3 and TSMU are really the same.

[dianzimi]

  who  knows   the components  detail  information  in the red rectangle bellow？？[img][img]

[tonyM]

  who  knows   the components  detail  information  in the red rectangle bellow？？[img][img]

As others nearby with SMD marking "512", probably it is SPDT switch QPC7512: https://www.rfmw.com/products/detail/qpc7512-qorvo/676832/

[dianzimi]

Do you know the components in the fsh6 pcb，with SMD marking "534"?

[tonyM]

Do you know the components in the fsh6 pcb，with SMD marking "534"?

Could you provide Hi-Res pictures of FSH6 PCB?

[dianzimi]

ok[/img]

[dianzimi]

[tonyM]

SMD parts with marking "534" are SPDT switches up to 6GHz.
In order to check if next sections are operational, you can bypass the input secton (Attenuator + Amplifier) by removing its first and its last SPDT switches, and soldering thin semi-rigid coaxial cable:

[tonyM]

Only attenuator by-pass (by the same way it is originally done in TSMU-ver.2):

[scopeman]

I can also describe the signal path for the 6 GHz option if anyone wants to see it?

I realize that it has been a while but I would love to see the FSH6 even though I think I know how it works. I do have both a FSH3 and a FSH6 both with blown attenuator chips. It's been a while but I think that the one in the FSH3 that I have is a SOIC packaged part but I don't remember the part number. I think it was made by Macom.

After I get my E4407B fully squared away perhaps I'll work on the two FSH's assuming I can find the attenuator chip for the FSH3.

Thanks!

Sam
W3OHM