Spectrum Analyser TR4172
 

Hello all,

I wonder if anyone out there knows what the amplifier in the interface unit is, as I have no information on this device. This is the only info that does not appear in the manuals. My unit is now approx 20-30dB deaf and I would like to bring it back to a fully working unit.

Re: Spectrum Analyser TR4172
 

Hi,

I also have a sick TR4172 in my case the RF section seems to be completely useless?

However, I got the unit without the main interconnecting lead between the two units and have used a standard one to one connection lead. Could someone confirm that is correct or are the plugs wired up differently i.e. the pins from one plug connected two different pins on the other plug?

Regards,
Stuart Perry
G0LRX

Re: Spectrum Analyser TR4172
 

Hi Stuart
Are you referring to the J2-J2 connector lead or the one called J1 BUS - J1 BUS?

I have one of these analysers and mine has two main connecting leads at the back (plus the IF coax connection of course)

Regards
Jeremy
G0HZU

Re: Spectrum Analyser TR4172
 

Hi Jeremy
I'm refering to the J1 - J1 lead (I wired this up pin to pin i.e. pin1 to pin1 etc)as I think the J2 -J2 lead is for makers as from memory it has two / three rf connectors and nine pins which I assumed are wired one to one? Any help with the lead configuration settings would be greatly appreciated!

I have connected the IF BNC connectors up.

Many thanks
Stuart Perry
G0LRX
:)

Re: Spectrum Analyser TR4172
 

Hi Stuart

I'm pretty sure the J1-J1 connector is wired straight but I'll buzz it out tomorrow.


Here's some quick images of the insides to show how the originals are made.

I'll check out the screening jacket as well to see if it goes to any pins or just the connector shell.



Regards
Jeremy
G0HZU

Re: Spectrum Analyser TR4172
 

I had a quick look inside the J2-J2 connector shells and got a bit of a surprise...


The 5 'data' pins are dummies at both ends!

i.e. there is no connection to these pins at either end. Not even a termination or link etc. Just untouched as new unused pins on the insides at BOTH ends.
Note that my analyser has options 5 and 7 stamped on the back so maybe these 5 data pins on J2 get used for other options?

The RF connections are one to one on J2-J2.

I'll test J1 tomorrow as it's a bit late and also a bit fiddly. Best done in daylight.

Re: Spectrum Analyser TR4172
 

OK I just buzzed out the J1-J1 bus cable and it is indeed a 1 to 1 cable. I took some time over this and marked up the edge connector every 5 pins to make sure I didn't lose synch between cables.

It buzzes out 1:1 and also the screen shells show good conductivity from end to end and they also have conductive screen plates as per the images above (the long shiny metal plates running the length of the blue body) to ensure a solid screen connection around the outside of the whole assy and the screen of the cable.

The plates mate with the outer body of the connector on the analyser to give a solid screen connection from the screen on the cable to the analyser chassis.

There is no connectivity from the outer screen on the cable to any of the bus pins when the cable is tested in isolation from the analyser.

Re: Spectrum Analyser TR4172
 

Thank you guys I guess I must have a sick RF section?

Stuart Perry
G0LRX

Re: Spectrum Analyser TR4172
 

Hi Stuart
I'm not sure if you mean the RF front end is faulty or what is referred to as the RF Block.
The RF Block does the 2nd, 3rd and 4th mixes.

If you ever feel like scrapping it then I'd offer you something for the analyser because I'd ideally like a second unit for spares to support this one.

I've been using this one for over 22 years either at my place of work or more recently here at home and it's been a great analyser and I dread the day it finally develops a major fault that puts it out of service for good...

Regards
Jeremy
G0HZU

Re: Spectrum Analyser TR4172
 

Hi Jeremy
I'll try and spend some more time on it when I can but have a two year old to contend with she has just woken up bye.

Stuart Perry
G0LRX

Re: Spectrum Analyser TR4172
 

Hi Folks,
I have a TR4172 which is about 30db deaf ! I have full manuals and suspect that the 1st mixer is zapped. Thats the usual reason for deaf Spectrum Analysers. Anyone know a suitable substitute for the ring of 8 diodes ("D1" which only has a "house code" in the parts list). Because opposite sides of the ring are DC connected, I cannot see an easy way to test it without removing it. GUII OP's post of December 2011 sounds similar - was that problem resolved?
Regards, Brian G8MUE

Re: Spectrum Analyser TR4172
 

I've not looked at the mixer circuit yet but what I can share is what happened to my TR4172 a while ago. It suddely went deaf on one attenuator setting which meant I couldn't run the inbuilt auto cal routine any more (takes a minute or so to run) to sharpen up the accuracy for critical measurements.

So I was faced with buying a new attenuator or taking it apart to see what was inside.


What I did find was that Advantest had fitted a tiny little soft ferrite? block over the microstrip path to this attenuator and they had glued it there.

Over time the glue attacked the microstrip and one day it finally went open circuit. So the >40dB attenuator settings all died.

Luckily it proved an easy repair.

See the image below. The yellow arrow points to the start of the eroded microstrip areas.

There's no way this happened with RF overload because the signal break is on the 'attenuated' side of the attenuators and the attenuators are intact and there was no sign of thermal damage. Also, the erosion area was perfectly contained under the old glue.

You can see how tiny the attenuators are and why this analyser has max +20dBm drive level. The big black thing under the yellow arrow is the soft ferrite? thing.

It does seem to improve the attenuator accuracy up at >1500MHz so I refitted it with different adhesive once repaired. TBH I don't really know what it is made of of why it is there.


The reason I'm telling you this is because they may have used this ferrite + glue elsewhere. My analyser must be nearly 25years old so this nasty glue has had all this time to react with and eat away the microstrip.

Re: Spectrum Analyser TR4172
 

Hello everyone,

I'm another owner of a TR4172 with defective CRT. I don't want to sell it because RF part continue to works well in spite of its 25 years, so I need to repair it. Is there anyone wants to sell the visualization module (upper part)? Here is some pictures of the defective CRT:

http://www.flickr.com/photos/37815563@N08/

Re: Spectrum Analyser TR4172
 

Can't help you with a spare CRT or top unit but are you still able to use it by getting display data onto a PC or other display?

The reason I ask is there is the XYZ (option 3) to allow a remote display.

Also, you can access internal memory via GPIB so you could in theory dump the 1000x1000 point trace data to a PC and get a display on a PC screen if you created suitable SW.

I'd imagine the refresh rate would be quite slow but these are things I am considering with mine as my CRT won't last forever and I don't have option 3 fitted...

In the meantime I have controlled it and dumped 'some' data from mine via GPIB using some simple VB code when I was experimenting with the programming command set and I've also managed to get some screen plots using the freebie KE5FX toolkit via GPIB.

Re: Spectrum Analyser TR4172
 

Sorry to drag up an old thread but in recent weeks I've been servicing both of my Advantest TR4172 analysers and I found a minor issue in the dreaded and undocumented mixer/IF interface unit.

So I spent a bit of time reverse engineering it. All my findings are subject to a degree of technical risk but if anyone wants to know about the amplifer in there or how it is biased or how the rest of the interface unit is designed (i.e. what is in there and what it does) then I can post the info up on here. I can do the same for the first mixer unit too.

Both of my analysers were showing a gradual reduction (over time) in front end linearity and a gain loss of a few dB and a slight degradation in DANL.

This analyser has a very beefy first mixer using 8 diodes (+22dBm LO drive level typical) and typically gives a mixer input IP3 of +22dBm.

But my analysers had slowly faded to giving only about +10dBm IP3 and this is down at the performance level of my old HP8568B with its very basic two diode mixer.

If I had not spotted/fixed this slow decline in gain and IP3 then the interface unit would have degraded to a point where the analyser had gone very deaf. i.e. there is an issue in this module caused by corrosion over time.

I can post up all the service/repair info if it is needed. i.e. if G4UUI and G8MUE still have their analysers stashed away as BER.

Re: Spectrum Analyser TR4172
 

Since my post in Aug 2012, I have continued to use my TR4172 despite the 30db deafness. When I switch on and set the centre to 145MHz and the span to 10MHz, the baseline (the middle of the noise fuzz) is at -83db (displayed). Connecting a signal of -50dbm from my HP8654A to input 1, is just visible in the noise. I have no reason to suspect the HP which also agrees with the 50MHz reference output from the TR4172. The input attenuator operates as it should on all steps. The pre-amp at input 2 works as it should.

Without a spare diode ring for the 1st mixer, I am reluctant to open it up. (A deaf SA is better than no SA). I have recently purchased (about £50 from a well known auction site) a "simple spectrum analyser" from Hong Kong. This USB instrument does seem to be able to make useful measurements of signal amplitude and frequency with an effective bandwidth of 250KHz. Frequency coverage is 138MHz to 4.4GHz (which will no doubt tell you which Analog Devices chip is being used). I was looking at the "RF section level diagram", Fig 10-10 in the TR4172 manual, and thinking how much of this I may be able to verify?

I am very interested if you can post or mail me some of the details you have found. Here in Cyprus it gets quite warm (it was 36C last Sunday) and I have found that in a number of instruments that have "plastic foam", the foam has turned to sticky tar. (This includes professional kit from Racal, B&K and HP as well as amateur items such as my KW2000 with its mechanical IF filter supported with plastic foam). Acetone is the only thing that I have found to clean the mess but it also dissolves many wanted items!

Re: Spectrum Analyser TR4172
 

Hi Brian
I'm very busy over this weekend but I'll try and find time to post up some info. However, what you must NOT do is open up the first IF module (that sits under the mixer box) and touch any of the PCB traces. Especially around the IF amplifier section where there are some little blocks of absorber stuck to the PCB.

I found the same issue with softened copper traces that I found in the attenuator section and I think the glue they used to hold down the RF absorber ages or cures badly in terms of becoming a chemical hazard.

On my IF1 unit this glue had attacked one of the printed bias resistors in the active (16mA) bias circuit to the IF amplifier and slowly eroded it. The effect was that the resistor suddenly went up in value a lot and the IF amplifier lost its power supply. The module still biased it at 16mA but the voltage at the amplifier was down at about 200mV!

Also, I think this glue has caused accelerated ageing/corrosion to the fasteners and connections in the IF1 module. They go VERY rusty like they have been underwater for a few years. Also, the connections from the SMA launchers to the PCB are done with tiny ribbons of gold? metal and the connection to the SMA centre can corrode here (the SMA centre corrodes rather than the gold ribbon) and this can break the connection or make it unreliable. There's no other corrosion like this anywhere else in either of my analysers so I blame the glue...

The good news is that the IF amplifier device survived in my analysers and you just need to solder in a SMD resistor in place of the printed resistor if yours has the same fault as mine.

But the copper microstrip on this PCB seems to go very soft. Almost like rotten paint on a car. So if you touch it or try and scrape any residues from it, it will lift up and fall to pieces like it is thousands of years old!

So the fix is easy but you have to do the whole thing like it was surgery in an operating theatre because one slip means the copper traces are gone... They can't be stuck down again like regular copper tracks because they seem to go very soft and fall apart.

I doubt that your mixer will be damaged but I did find the original manufacturer's part number for it and I think they are still available online (but not from Advantest)

i.e. this is not an Advantest part number but the real part number for the diode ring.

It's a VERY strong mixer with 8 diodes inside and the LO drive level is about +23dBm so you may be OK and your mixer will be healthy.

I'll dig out my documentation and post up the part number and a few internal pictures when I get a chance.

Re: Spectrum Analyser TR4172
 

I found my old notes on the mixer and the part number is DMJ4708 which also has an HP equivalent part 1906-0279 according to some sources.

I think Alpha (now renamed to Skyworks) used to make/sell the DMJ4708. There are a few US retailers selling new/old stock online but not for overseas buyers.

Patience is probably your best bet here. If you keep searching you may find one cheaply. Or maybe get someone in the US to buy one for you.

However, I'd be surprised if the mixer is dead.

Here's a couple of faults I would consider more likely than a blown up mixer.

IF1 amplifier lost bias voltage due to eroded (printed) resistor on the active bias board.
Corroded/broken connection to one of the SMA connector centres inside the IF1 module.

Re: Spectrum Analyser TR4172
 

Thank you G0HZU_JMR for the helpful response. Yes I did manage to get a quote for the Skyworks product but the minimum order was 100 off for about £700. Thanks for all the warnings about the PC tracks in the 1st IF. I was once involved in the offshore oil business and sometimes asked to recover an instrument which had flooded with sea water. Often this was possible but very often the repaired item would fail in 3 to 6 months through corrosion of the PCB or other components. Your post should keep me out of mischief for a while.

Re: Spectrum Analyser TR4172
 

Here's a couple of images of the IF1 amplifier device.

The IF1 amplifier is marked K in a 35 micro X package and I think it is the same as Q16 on the RF board = 2SC2150.

The first image shows all three bits of absorber and the second shows the printed resistor (arrowed red) that sits under the middle bit of absorber after the absorber is removed. This resistor had been attacked by the glue and had gone very high in resistance. It was also very faint as it seemed to have been eaten away.

The microstrip track arrowed in yellow is likely to be very soft as it also seems to get attacked by the glue. So it falls apart if touched with any tools. For this reason, don't touch ANY of the printed traces in the filter sections as these may be soft too.

You can also see how corroded the screws fasteners are. I think either the glue or the absorber sweats some form of catalyst that causes this corrosion.

This is a sealed unit and the same screws and metal tabs are used elsewhere in sealed parts of this same IF1 module but they are in sub compartments that don't have absorber or glue and the screws still look shiny.

So I don't think there's a dissimilar metal issue causing the corrosion.

To fix mine, I replaced the printed resistor (arrowed in red) with an SMD equivalent. I found a model for this type of transistor and modelled the stage for gain and noise figure on Agilent Genesys and found that the optimum bias point was about 2.5Vc. Biasing it at 5V Vc degraded the gain and the noise figure and I proved this with tests on the real module. With the corroded printed resistor the bias point was down at a tiny fraction of 1V because the resistance had gone quite high and this caused the stage gain to collapse. On my second TR4172 analyser the resistor wasn't as degraded and this still had reasonable performance and Vc was something like 1.5V.

The circuitry on the little piggy board is the active biasing network that biases the 2SC2150 device at a constant current of about 15mA.

Re: Spectrum Analyser TR4172
 

1906- prefix in HP-speak is either a multiple diode chip or else a matched set of individual ones.

Most of the HP diodes live on under different part numbers and in different packages on the Avago website . HP broke up into HP computers and Agilent instruments. Agilent then flogged off the old HP associates semiconductor arm and it became Avago

You can also get good mixer diodes by pulling apart old mini circuits mixers.

Some plastics and glues give off acids like HCl as they decompose, that ought to corrode most things.

DAvid

Re: Spectrum Analyser TR4172
 

I don't have much experience or knowledge about things like this but see the attached image of the insides of one of my TR4172 IF1 modules.

I can't remember which TR4172 this image came from but they were both very similar.

You can see the corrosion inside the main section and it even extends to the inner aluminium. By contrast, the input LPF section on the right still looks fairly new and shiny. So I think either the glue or the black lumps of absorber must sweat some form of catalyst for oxidation. Maybe they glued the absorber in place and immediately fitted the cover thus sealing in all the fumes from the glue as it cured? But that's just a guess.

I was expecting to see a cavity filter here but they use printed filter sections for the main IF1 filter. I measured the response on a VNA and it gave very good stopband performance.

The gain through the module was about +1dB and I measured the same on both of my TR4172 IF1 modules. I recall that the manual suggests the typical module gain is 0dB +/1dB.

The input section is a directional filter followed bu a roofing LPF, the IF1 amplifier and then the printed 2.046GHz IF filter sections. There's a load of RF absorber stuck inside the aluminium cover that fits over all this. The connections to the two SMA connectors in the main section are very fragile and had corroded to the point that one was only making a connection through touch alone. So this is one place where the module could suddenly lose gain.

Re: Spectrum Analyser TR4172
 

Thread reopened at member's request.

Re: Spectrum Analyser TR4172
 

Hi,

I’m new to this forum and I requested the moderator to reopen this thread because I have the same gain loss issue as described by Jeremy G0HZU. After some probing in my TR4172 I discovered that the gain loss was caused by the IF1 amplifier and after some googling I found this forum. In contradiction to the IF module of Jeremy mine looks ok and is slightly corroded. This is probably because there are no absorbers glued on the microstrip line at the output of the amplifier, see picture. Only in the cover some absorber is mounted. What looks suspicious are the film resistors. They are faded and have a strange colour. Jeremy I understood that you have modelled this IF1 amplifier stage and the biasing piggy board and I wonder if you want to share this information with me:). You have simulated it with Agilent Genesys and professionally I use ADS (also Keysight). I do not know if I can import a Genesys project into ADS. If this is not possible than a schematic will also be helpful.

In my case I have observed the following: the IF module is drawing 23 mA at 15V, Vbe of the RF transistor is 0.89V, Vc_cap is 6.85V and Vce is 110 mV (much lower than in your case). Because the Rc film resistor looks suspicious it is possible that it also has a very high impedance. Can you please give me the resistor value which you have soldered on the board:)? You have simulated that Vce=2.5V was the optimum but at the end you have biased it at Vce=1.5 V, is this correct? Did you change something on the biasing piggy board (resistors are also faded and have a strange courol)? One more question, I assumed that the traces on the substrate board are of gold. Did you use indium solder to mount the SMD resistor to the strip line?

Sorry for all the questions but it will be very helpful and it will save me some time.

Regards, Lex

Re: Spectrum Analyser TR4172
 

Hi Lex, I had a rummage and managed to find an old simulation of this amplifier. I think you have to first make sure the constant current biasing is operational. This uses a PNP transistor Q1 as in the circuit below.

My IF1 modules both drew about 16mA from +15V and I think the active bias fed just under 15mA into the RF amplifier BJT.

The constant current is defined by the 288R resistor and the voltage drop across it.

Constant current =(15V - VemitterQ1)/288R

Constant current = (15V- ((15V *(10000R/(10000R+4700R))) +0.65V))/288R

= (15 - ((15*0.68) + 0.65))/288 = 14.4mA

To get 2.6V at the collector the collector resistance would have to be (10.8V-2.6V)/0.0144 = approx 560R.

I chose to run the collector of Q2 at about 2.5V so I assume I added a resistor here to achieve this. I can't remember if I cut away the old failed printed resistor or not. I just used ordinary 60/40 solder but that was because it was all I had.

Your circuit values for the printed resistors may be different if your circuit draws 23mA. Also, these resistances may change over time. I can't be certain but I think the resistances shown below were measured from the real board and I think both boards were the same apart from the Q2 collector resistor that had failed.

Re: Spectrum Analyser TR4172
 

I had a quick play with the simulator and tried to come up with a failure mode that fits your symptoms below.

23 mA at 15V, Vbe of the RF transistor is 0.89V, Vc_cap is 6.85V and Vce is 110 mV

If there is a significant breakdown leakage path (maybe even a short) between collector and emitter of the active bias PNP transistor Q2 the circuit will draw 21mA and will give the following readings:

21 mA at 15V, Vbe of the RF transistor 0.9V, Vc_cap is 9V and Vce is 140 mV

This matches fairly close and a lot depends on how badly the transistor has failed. I think pretty much any SOT23 PNP will work fine here as a replacement for Q2 if your circuit has failed in this way.

Could the blob of gunk that sits over Q2 (shown in your image) have caused this collector emitter leakage path?

Re: Spectrum Analyser TR4172
 

Hi Jeremy,

Thank you for the quick response. The schematic is very helpful and now I better understand what is going on. I agree that probably something is wrong on the piggy bias board. I did not mention it but VTP2 is much lower in my case, 7.31 V which is divided from the 15V and loaded by Q1 (PNP). I think I will first disconnect the biasing from the RF transistor Q2 and check all resistors values. The resistors on the biasing piggy board are most suspicious. Based on that outcome I will try to remove the blob on Q1. Tomorrow busy so this will be Sunday. Thanks so far.

Best regards, Lex

Re: Spectrum Analyser TR4172
 

Glad to be of help! The TR4172 is a very good spectrum analyser so I hope I can help you get it working again.

I earlier spotted some notes in the simulation workspace and this has the following info:

These look to be genuine voltages measured of the real circuit of the second TR4172. The collector resistor of the RF BJT on the first
one had been so badly damaged/dissolved by the glue it was almost eaten away.

However, I'm a bit confused by my notes above vs what I wrote in an earlier post about the second TR4172. It looks like the second one had a 445R resistor here although I don't know if I put this 445R resistor there or not.
I do recall measuring 1.5V here on the second TR4172 but it was 5 years ago!
It looks like it was measuring 4.13V here just before I put it back together.

I think it would be really useful if you could measure all the resistor values on your board. There's a good chance they are all healthy and it would be nice to know the correct values and the correct test point voltages.

Re: Spectrum Analyser TR4172
 

Hi Jeremy, I think that you are right that something is wrong with the PNP. I have measured the following resistor values:
R1 = 2k1, R2 = 391, R3 = 9k96, R4 = 3k28, R5 = 316

I have disconnected the RF transistor but if the PNP is broken some values can be incorrect. I did not remeasure the voltages but this is wat I have measured the last time:
VTP1 = 110 mV
VTP2 = 7.31 V
VTP3 = 6.85 V
VTP4 = 7.94 V
VTP5 = 0.89 V

Note, VTP2 is not what you expect based on the voltage divide ratio. Based on the measured resistor values and voltages the current through the RF transistor is about 20 mA. It is a pity that they used the black blob on the PNP. It will be hard to remove it without damaging R3 and R4. What do you think?

Regards, Lex

Re: Spectrum Analyser TR4172
 

It's hard to tell at the moment but maybe the B-E and C-E could be faulty in the PNP transistor.

I'm guessing that the blob is some form of hard resin. If so then I think other people on the forum will be able to offer their experience here. I can only suggest a few obvious techniques (scalpel or Dremel) that will probably damage the resistors.

If the resistors do get damaged then they could be cut or drilled away and they could be replaced with SMD versions in something like 1206 or 0805 package. I suspect the PNP will be a simple BJT. The M6 code suggests that the 2SA812 is a possibility and this version of the 2SA812 is in the gain group 200-400.
Something like a BC856B or BC857B would probably be fine here.

Re: Spectrum Analyser TR4172
 

Thanks, I have ordered the BC857B and in the meantime I will try to remove the blob. Actually I do not understand why they have used the blob unless it is a bare die bonded on the board. We will see what’s inside. Question, do you want to share your model of the 2SC5010 with me than I can do some simulations myself in combination with the BC857B. You can send me a private mail if you want to.
Regards, Lex

Re: Spectrum Analyser TR4172
 

Sadly, the 2SC5010 model (along with many others) is embedded into a huge common non-linear library file within Genesys so it can't easily be understood or extracted.

I'm not sure how accurate the model is. I can't remember if I removed the SMD caps and measured them or if I reverse engineered what they should be by using the simulator but I got the results below.

Gain 9.9dB at 2.05GHz
Noise Figure 1.9dB at 2.05GHz
OIP3 +23dBm at 2.05GHz

Re: Spectrum Analyser TR4172
 

I did a simulation myself with the BC857B and some NPN and I concluded that it must be the B-C transition. This also explains the low VTP2 of 7.3 V which I have measured. I will keep you informed.

Regards, Lex

Re: Spectrum Analyser TR4172
 

First post, so please bear with me if I write something ignorant... :)

It is apparently TR4172 ailment season. I acquired a TakedaRiken branded TR4172 a few years back for a song at a local hamfest (anyone remember those? [dang covid-19!]). I think it was $150US plus the herniated disc it took to carry both units to the car. I fully expected it to not work well, and indeed it was about 20dB deaf on input 1 and 14dB on input 2. Enough deafness that it would not pass the auto calibration routine. Still, it worked good enough for the light duty work I needed, and was still rather linear across it's operating range. Fast forward to this week, and while checking the stability of the timebase, leaving the SpecAn on overnight, I notice that the signal I had fed into it was around 20 dB less than it was the day before. Hmm... not fun! After muddling through the test flowcharts I finally decided to simply check the 1st mixer, remembering that more often than not, this is where deafness occurs. Sure enough, I discovered what G0HZU had found; lots of corrosion inside the mixer. In fact, the diode ring has pretty much crumbled apart without much chance of reviving it. The leads have basically eroded to nothing and it appears that even the package has deteriorated. Now I am stuck to find a replacement mixer or diode, and I should probably also open the IF interface to look for issues, and/or also lend a hand if anyone needs any info or measurements.
I want to give a huge thanks to everyone on this thread - I've learned something from it, and at least have a part number to investigate for the ring mixer.
On that note, has anyone ever considered or done a replacement swap with a more commonly available mixer? Would it be advisable?
Regards-
Matt

Re: Spectrum Analyser TR4172
 

And the obligatory thousand words each...

Fortunately, after a bath in the ultrasonic cleaner for a while I found that things in the mixer weren't quite as bad as I first thought. It appears that the mixer has just enough leads left to facilitate some solder surgery, so I may be in luck there. Pictures for that to come...
Unfortunately, the interface module was in pretty poor shape. They appear to have used a ceramic substrate with gold plated copper or pure gold foil. Whatever it is, one can see the lifted trace in the coupler section. This was likely the cause of the initial loss of sensitivity. I'm debating on the method to affix this, and the best I can come up with is using a sharp needle to apply a very thin layer of cyanoacrylate to the substrate and an artist's brush to lightly press down on the trace, similar to applying gold leaf.
Anyone have any better ideas?

Re: Spectrum Analyser TR4172
 

Hi Matt
Are you sure it is the first mixer you are referring to? This has 8 diodes as per the image from the manual below. On my TR4172 analyser this mixer is in a flat pill package with 4 gold legs. It looks fairly robust to me and I'm not sure how it would crumble apart.

However, if this mixer has failed then there are a few things to bear in mind:

The mixer contains 8 diodes and the local oscillator drive power to this mixer is typically +22dBm. This will make it difficult to find a drop in replacement if the original part cannot be purchased at a reasonable price anywhere.

In the diagram below I've offered a temporary bodge solution using Avago HSMS-2822 or HSMS-282C diode pairs. This 'might' work and it might cope with the +22dBm LO drive level. Otherwise you might have to fit an SMA 3dB or 6dB attenuator inline with the semi-rigid cable from the LO amplifier module.

The Avago HSMS-2822 (SOT-23) and HSMS-282C (smaller SOT-323) diode pairs are obsolete but they are cheap ($1 each?) and you might still be able to buy them on Ebay. I doubt this fudged mixer will work very well (especially at 1800MHz) and it won't have very good balance either so some aspects of the analyser performance will be compromised. I'd expect it to be better than a broken mixer if you really just want a cheap repair that restores some basic functionality.

You could opt to buy a quad diode package as this will work better at higher frequency but this is unlikely to cope with more than +10dBm LO power.

Edit:
I've changed the part number to the 15V version of the diode.

Re: Spectrum Analyser TR4172
 

Hi, Jeremy.
Yes, it is unfortunately the 1st mixer. I thought the package had crumbled, but thankfully I was mistaken. It was just REAL dirty and cruddy (see the pic). I also had never seen up close a package quite like this and did not realize the black area was part of the package. It was only the legs that crumbled. One would think that gold plated copper would hold up to just about anything corrosive, but here we are...
I think I may have been able to save the mixer though. It's currently soaking in some denatured alcohol to remove the flux, but see the attached picture taken before the flux removal. There were basically no legs left after 35 years or so exposed to whatever evils kill these devices. All that I was able to solder to was whatever was left of the thin leads sticking out of the package. It isn't pretty by any stretch of the imagination. Hopefully I'll see some response through it on my 8753, and hopefully it will last if it works.
As for the IF amp block, I'm about to attack it and see what I can do with it. That will be another very intricate repair, but I can't see any other way to do it other than some glue.
What's worse is that there are very thin film resistor elements at the ends of the coupler traces. Thankfully the trace has not become detached from the resistor. Worst case there is that I'd have to solder in a SMT device if I break that bond. I'm not sure though... this is all new territory for me. I'm more used to large scale RF stuff... TV and radio transmitters in the several tens or hundreds of kilowatt range. T&M equipment repair is something that I've just recently delved into out of necessity, and I don't have much experience with microwave equipment on this small of a scale... usually it is a matter of swapping modules, not repairing them.
Thanks for the ideas and advice.
Regards,
Matt

Re: Spectrum Analyser TR4172
 

Wow, I just saw your post #35. It wasn't visible whilst I typed the stuff above. I can see the state of the first mixer now. I can't explain what has caused this unless there is something in the mixer box that caused the corrosion. This might be due to some sort of chemical reaction to fumes from RF absorber or glue and this might be hidden underneath.

Your directional filter also looks to be peeling away from the PCB in the IF1 module. This doesn't look good. The PCB traces on this board are very fragile.

Re: Spectrum Analyser TR4172
 

In the worst case, it would be possible to remove and scrap the square directional filter section shown in 20210302_173808.jpg and make a copy as this looks to be a separate PCB. There isn't anything special there and it just has to have the right dimensions to suit the alternative PCB material Er and thickness. Make sure you never touch the other printed BPF sections though. This 2.05GHz BPF would be much harder to replicate.

Re: Spectrum Analyser TR4172
 

I'm still on the probationary period of having to have the posts approved by a moderator.
Yeah, if I can't glue this stuff, then that would be the next step, and I do have the equipment to fab a board if need be. I had not thought of that; thanks!
I concur that these issues are likely the glue they used to glue the absorber material to the covers. It is amazing how it even tarnished the silver plated brass plate in the amp section. I'm still on the fence as to whether I should bother removing the absorber and re-gluing with something else, or to take the chance that perhaps all that could have out-gassed already has. Perhaps I'll just leave it and check in a couple of years.
73,
Matt

Re: Spectrum Analyser TR4172
 

The square filter section will probably have a very narrow response. The idea is to have a selective filter at the first IF but to also have a lowish VSWR across a much broader bandwidth. This broadband match is desirable as this filter sits directly at the IF port of the mixer.

See below for an old simulation of a similar design I did on Genesys many years ago. It has a similar IF frequency. You can see that there are two simulation responses for the filter. One uses physical models for the printed lines and the other uses the Sonnet EM engine to do an EM analysis of the actual PCB layout. The two responses agree quite well.

The slight worry will be that if you glue the microstrip back down it might be detuned slightly away from the true IF1 frequency. However, I'd still be tempted to try gluing it as you suggested. Just use minimal glue and try and get the microstrip as flush to the PCB as possible.

Re: Spectrum Analyser TR4172
 

Most excellent, sir! A huge thanks to you, Jeremy. I imagine even if I screw it up a bit, it would be better than it was. If need be, I could perhaps outboard a separate directional coupler and a sharp-tuned cavity filter - at least there's enough room in there to do so. A bodge for sure, but I simply love this old SpecAn. I don't want to retire her any time soon if I can help it.
73!
Matt

Re: Spectrum Analyser TR4172
 

If you measure the gain through the whole 2.05GHz signal path in that module you should get something like 0dB +/-1dB gain at 2.05GHz as seen in my plot in post #22. I think I saw +1dB gain here for one of my analysers and slightly less on the other. If you get within a couple of dB of this I think it means the gluing was largely successful.

It's probably best to use an SMA attenuator at each cable end before you normalise the test rig to 0dB gain at 2.05GHz as this will minimise mismatch uncertainty effects. But I guess you would do this anyway (or you could use a fully corrected 2 port VNA). The IF1 amplifier drew 16mA from +15V when I did this test.

Re: Spectrum Analyser TR4172
 

It's quite important or getting low spurii to have all mixer ports nicely terminated at all frequencies that stuff comes out of them. It's not just the wanted frequency, it's the images, LO and RF blowing straight through etc, as well as harmonics and intermod products. The absolute last place you want these things to go is back into the mixer to cause more trouble.

So the RF lowpass on the input might be arranged to terminate the LO (and harmonics) leaking out the mixer input port, or maybe an attenuator is used to sacrifice some sensitivity for cleanliness. The LO port in posh analysers may get a circulator and load to dump anythng coming back, or maybe a beefier LO drive is used with a broadband pad.

Termination of the IF is an obvious one. Something important in high performance comms receivers as well.

David

Re: Spectrum Analyser TR4172
 

After having to try a few more times to get this guy soldered in there, it would appear that I have gotten the mixer to once again mix. I don't have a spec an that goes high enough, but I do have a counter that does, and when applying a 2.1GHz +12dBM signal to the LO input, I see it on the output. When I switch on a 10MHz 0dBM signal to the RF in and a I get the 10MHz offset on the counter. I buttoned it up and checked the RF and LO inputs to the mixer, and it would appear that perhaps I have a blown input attenuator in the mixer... the VSWR is rather poor.
All unused ports terminated, of course.
Now to tackle the amp...

Re: Spectrum Analyser TR4172
 

Well, not so much luck with the amp.
I went ahead and looked at the input impedance, and as expected, she's a bit off. At least I'll know if it gets better after I attempt a repair.
Just for kicks I measured the LPF response.
Then on to check the gain. I see about -45dBM through it without power, and that drops several dB when applying 15V. Current draw is only 1.085mA. Checking voltages on the transistor, and I see my 15V on the green dot in my picture, but when I place the probe onto the red dot, the voltage starts at around 300mV and drops to around 160mV within just a couple of seconds.
I'm done for the day... time for some grub in my pie hole.
Thanks again for the tips, guys... I'll get this thing going one way or another.
Regards-
-Matt

Re: Spectrum Analyser TR4172
 

I can add some performance info about the mixer if that helps.

The 2HI performance and IP3 performance of this mixer across HF and into VHF was very special in its day. When integrated into the analyser the analyser boasted an input IP3 of +22dBm with 0dB attenuation at the front panel. Obviously there is also a permanent 3dB attenuator there but this is still really good performance from a mixer. The 2HI performance was even better. A good spectrum analyser will manage a 2HI of +50dBm down at HF/VHF. Most barely manage +30dBm. This high level mixer with its 8 diodes was at least 20dB better than the classic HP8568B in this 2HI test.

I can remember testing the mixer in isolation when my TR4172 was failing at IF1. I recall it managed an input IP3 of +25dBm. I think the conversion loss was quite poor because of the input attenuator and the RC compensation. I think it was something like 13.5dB loss through the mixer at HF/VHF. The mixer performs best with at least +13dBm LO drive and the distortion figures improve as the drive level is increased to +20dBm.

Note I wouldn't advise running the analyser if you ever remove the mixer block or disconnect the LO cable because this will leave the large (expensive + unobtanium) LO drive module unterminated and this might stress it.

It does look like your IF1 amplifier block is faulty...

I had a rummage and found a plot of the input LPF in my first TR4172. See the plot below. This shows just under 2dB loss at 1800MHz. There is very little in the way of IF rejection in this filter and that is why I think there is also a stub filter.

Re: Spectrum Analyser TR4172
 

One other thing, I keep meaning to mention that it seems odd to see that big chunky brown axial resistor on the LO port of the mixer. There will be +22dBm across 2GHz to 4GHz here and it seems odd to have that large type of shunt component fitted here. It isn't shown in the manual and I'm quite sure it isn't there in either of my TR4172 analysers.

What value is the resistance? It might be there to stabilise the LO port impedance but it seems an odd package choice. However, Takeda Riken (Advantest) do seem to fit some odd components in odd ways.

Re: Spectrum Analyser TR4172
 

I have read about this SpecAn's legendary sensitivity, image rejection, and noise figure. These are all parts of the reason why I really don't want to retire her, although if I do, her guts will go to those needing to maintain another.

Duly noted. I was worried that the YIG was falling in output at first due to my prior experiences with the High Priced equipment - namely my 8753D, but thankfully that turned out to be well within spec. I'd hate to think of the bodgerry I'd need to come up with or the expense I'd have to incur to rig a different one or have this one repaired.

That is very helpful; I need to verify that the poor response I see is not due to some shady adaptors; I calibrated the VNA to the N connectors, and this includes the N to SMA adaptors I used to gointa and getouta.

Re: Spectrum Analyser TR4172
 

It's a 1K 1/4 watt resistor. The ultrasonic bath handily removed the rust, grunge, and the paint.
The point at which it fastens is key... I imagine it's at a node along the stripline that is ideal for that complex impedance. I'm figuring it kills some parasitic oscillations or tunes the line a bit sharper to help narrow things down, but I'm just assuming... when it comes to MMIC tech, striplines, and all of these microwave kinds of craziness, it's all FM to me.