HP3577A 5Hz-200Mhz VNA Teardown

[sixtimesseven]

Yesterday the postman delivered another boat anchor to my dorstep - A HP3577A VNA with (75Ohm  S-Parameter Test Kit).

Teardown photos:

• HP3577A / HP35677A/B S-Parameter Test Set
  https://www.flickr.com/photos/154275690@N07/albums/72157699120602134
  

The thing looks a bit beaten and is full of dust but it seems to work fine at first glance.
Confidence test fails at item 6a, "Receiver impeadance test" The relays seem to switch fine between 50/1M and the attenuators switch fine for both ranges. Signals comming trough a 20dB attenuator look reasonable as well. Might just be an adjustment problem...

Couldn't resist to take it apart a bit. Also the front panel boutons are a bit sticky and I would really like to get them running smooth again. Also I'm probably going to switch out the CRT with a NewScope-5 LCD.

Update: LCD ordered and switches cleaned with WD-40 Contact Cleaner. They work great now. I also installed the Sanyo fan (see link below), temperatures at the heatsink stay the same for me and the noise went down from "jet engine" to anoyingly loud oc desktop pc.

Update II:
New Scope LCD is terrific. Really nice kit and nice LCD. A bit pricy but worth it in my opinion.

Interesting links:

• Original fan datasheet (Muffin MD24B2):
  https://www.jameco.com/Jameco/Products/ProdDS/2215478.pdf
  
• Sanyo replacement fan, 39dB:
  https://www.mouser.ch/datasheet/2/471/San_Ace_120WP38_E-1286779.pdf
  
• Good Service Manual scan for 12.50$, part number 03577-90010:
  http://artekmanuals.com/manuals/hp-manuals/
  
• NewScope-5 CRT to color LCD conversion
  http://simmconnlabscom.ipage.com/store/page4.html
  
• How to clean sticky switches:
  http://www.rbarrios.com/projects/HPSWITCH/
  

[Johnny10]

I keep losing auctions on this unit.
Nice buy.
I bought an S-Parameter last month thinking I was going to win an auction on a broken 3577 unit.
Nope!

I will keep trying.

Like your photos

[sixtimesseven]

I keep losing auctions on this unit.
Nice buy.
I bought an S-Parameter last month thinking I was going to win an auction on a broken 3577 unit.
Nope!

I will keep trying.

Like your photos

Well the seller put it in for 1499$ so I made him an offer for 900$ and he accepted so

I cleaned out all some of the dust from where I could get to it, removed the two logic boards and took some more pictures -> See Flickr: https://www.flickr.com/photos/154275690@N07/albums/72157699120602134


I also wanted to backup the firmware but the beancounters at HP have decided not to put in sockets for the ROM IC's 

Now I'm just waiting on my N to SMA and N to BNC adapters to do some measurments. In the meantime I will probably have a look at the S-Parameter test set...


 

[precaud]

The thing looks a bit beaten and is full of dust but it seems to work fine at first glance.

Well I'm sorry to hear that. I have three 3577A's waiting for me to troubleshoot/repair and I was looking forward to you doing the same in advance    All of mine have non-trivial issues, unfortunately.

(j/k of course... nice that yours mostly works fine!)

Couldn't resist to take it apart a bit. Also the front panel boutons are a bit sticky and I would really like to get them running smooth again.

Easy fixes once the front panel is disassembled.

Also I'm probably going to switch out the CRT with a NewScope-5 LCD.

Huge improvement. As my eyes age, that small crt becomes increasingly difficult to live with. I need to keep a magnifier handy.

Interesting links:
-Original fan datasheet (Muffin MD24B2): https://www.jameco.com/Jameco/Products/ProdDS/2215478.pdf

50dBa ! That's one noisey fan! The fan noise and crt are what keeps the 3577A (and 3562A) off my bench. Lovely instruments crippled with noisey fans and hard-to-read crts.

The Achilles heel of the 3577A appears to center around the input relays. That was the major design change between A and B versions, a redesigned front end using different relays.

[sixtimesseven]

50dBa ! That's one noisey fan! The fan noise and crt are what keeps the 3577A (and 3562A) off my bench. Lovely instruments crippled with noisey fans and hard-to-read crts.

I probably just install a quiet 12V PC fan with a 24V converter and test with the thermal cam if the 80 CFM are enough without the CRT heating the instrument.

The Achilles heel of the 3577A appears to center around the input relays. That was the major design change between A and B versions, a redesigned front end using different relays.

Do you already have some with that issue? I'm still waiting on Artek to get me the service manual so I have no idea yet what these relays are, but are they replacable at all?

[precaud]

I probably just install a quiet 12V PC fan with a 24V converter and test with the thermal cam if the 80 CFM are enough without the CRT heating the instrument.

I look forward to seeing your results. Since you've been inside the unit, I think you can feel that most of the heat comes off the various PCBs, and it takes some pressure to get air to circulate among them all. I'm doubting if a standard PC fan will be sufficient.

Do you already have some with that issue?

Yes, two of the three.

I'm still waiting on Artek to get me the service manual so I have no idea yet what these relays are, but are they replacable at all?

Replaceable, yes. Still being made, no. They were small metal-can ones made by Teledyne, the 412 series, if I recall correctly. Pull an input board and have a look.

Other HP instruments from that era (DSO's, logic analyzers  with digitizer boards) used similar relays, so its possible they may be a good source of parts if we can establish compatibility.

[rx8pilot]

I got one of these for $200 a while back - would not boot. It was the EEPROM's which I replaced with a fresh flash and it booted up.  A few days ago the big filter caps on the power supply failed - so I replaced those as well. I have done all the usual tests and it appears to be functioning as expected.

The problem is....I still don't really know how to use it. Joining this thread to learn something.

[precaud]

I got one of these for $200 a while back - would not boot. It was the EEPROM's which I replaced with a fresh flash and it booted up.  A few days ago the big filter caps on the power supply failed - so I replaced those as well. I have done all the usual tests and it appears to be functioning as expected.

Glad to hear it was an easy fix. I'm guessing it was the +5V rail caps that went? They have the most ripple current across them.

The problem is....I still don't really know how to use it. Joining this thread to learn something.

To me, VNAs are all about transfer functions. So get some devices you want to characterize and experiment. AC coupling is not an option, so be sure to protect inputs from anything with DC.

Isolation and injection transformers are a good place to start, because the 3577A has grounded inputs and output, and at some point you'll want to use a xfmr to isolate the output or insert one in a control loop. Investigate how their mag/phase responses change with load.

I use my low-freq VNAs mostly for impedance measurement. I'm a bit obsessed with Vreg output impedance and a VNA is the tool of choice for that.

[rx8pilot]

Glad to hear it was an easy fix. I'm guessing it was the +5V rail caps that went? They have the most ripple current across them.

One of the large 250V caps closer to the primary side failed short - I replaced both. They are screw terminal caps so I didn't even have to warm up a soldering iron. The EEPROM fix was quite a bit more effort to troubleshoot and execute the fix. The error message only flashes on the CRT for a second, but only when you power it up. That means the CRT is not yet displaying anything. I only noticed the message when I pressed the reset button on the processor PCB with re-boots without turning the CRT off. It allowed me to see the startup sequence. Led me right to the end of the rainbow.....

To me, VNAs are all about transfer functions. So get some devices you want to characterize and experiment. AC coupling is not an option, so be sure to protect inputs from anything with DC.

Isolation and injection transformers are a good place to start, because the 3577A has grounded inputs and output, and at some point you'll want to use a xfmr to isolate the output or insert one in a control loop. Investigate how their mag/phase responses change with load.

It looks like these can deal with 25VDC on all inputs - not quite as delicate as the HF VNA's and SA's.

That is my first attempt - an isolation transformer for testing control loop response. I am not sure how to set it up, normalize it and just get a scaler measurement of frequency response (as a first test). What is the proper setup to measure the frequency response of an isolation transformer?

I tried a few setups, but they are clearly not correct. The OUT connected to the transformer with the INPUT A in parallel - I normalized that and it is flat. Then I move the INPUT A over to the output of the transformer - but the response looks nothing like what I would expect. I did the same setup on my scope with a signal gen and it is showing that -3db should be 10hz - 6Mhz.

[Jay_Diddy_B]

Yesterday the postman delivered another boat anchor to my doorstep - A HP3577A VNA with (75Ohm  S-Parameter Test Kit).

Snip ...

Be careful with the 75  N connectors on the S-parameter test set. They will be wrecked if you mate them with 50  N connectors. The center pin is a smaller diameter on the 75 version.

I have both the 35677A 50 and the 35677B 75 test sets.

At lot of the time I use the 3577A for measuring Bode Plots on SMPS without a test set.

Regards,

Jay_Diddy_B

[precaud]

The error message only flashes on the CRT for a second, but only when you power it up. That means the CRT is not yet displaying anything. I only noticed the message when I pressed the reset button on the processor PCB with re-boots without turning the CRT off. It allowed me to see the startup sequence. Led me right to the end of the rainbow.....

Nice!

It looks like these can deal with 25VDC on all inputs - not quite as delicate as the HF VNA's and SA's.

Well I won't put DC on the inputs of mine... all three of my fixer-uppers have input stages issues, and DC is often the culprit... besides, DC shortens the useful life of relay contacts...

That is my first attempt - an isolation transformer for testing control loop response. I am not sure how to set it up, normalize it and just get a scaler measurement of frequency response (as a first test). What is the proper setup to measure the frequency response of an isolation transformer?

(Forgive if I make terminology errors, I mostly use an Anritsu VNA which uses different names for these things.)

If you want transmission response in a 50 Ohms system, or use the "Shunt-Thru" impedance technique, you need a splitter to isolate the R input from the load of the DUT. (Oh yeah, another thing you'll want to pick up at some point is a resistive splitter good to 200MHz...) An example of this is shown on p. 2-6 of the 3577A op manual.

But most often you'll want to measure transfer response from a 50 Ohm source, to include the loading effects of the DUT on the source, and into a load that you define. In the default menu, select A/R. Set inputs R and A to 1MOhm. Y the output to input R and xfmr primary. Solder a 50 or 100 Ohm variable resistor across one end of a BNC tee. That makes it easy to remove the tee to measure the R load its presenting. Maybe make some marks on the pot for certain R values. Connect the xfmr secondary to the tee, and into input A. In this setup, R=reference or stimulus, A=response. Set levels, freq range, display format, etc. Then make the sweep.

What some VNAs call "Thru", the 3577A calls "Normalize" and the Anritsu calls "X-S". At least the first two names make some sense    The basic idea is to measure the inter-channel response difference at each freq so it can be subtracted from each subsequent measurement to remove the errors. Remove DUT and connect Output to both inputs (in this situation, set at 1MOhm) and make the sweep. Then turn normalize on and voila, future sweeps will be accurate. Any change of setup requires a new Normalize sweep.

Transfer function is the same as FRA (freq resp analysis), so if you've done any of that, the same rules apply at these low freqs.

There's another clever use of the "normalize" function. Since it stores a baseline measurement, you can use it as a reference of any sort to be subtracted from other measurements. Very handy if, for instance, you want to see how mag resp and phase changes as you vary the secondary load R on a xfmr. Make the "Normalize" measurement with DUT in place at some reference R load value (say, the value of the resistor you'll be inserting in a control loop). Then make sweeps with it set at some small incremental Ohms above and below the reference value, to see how the xfmr response changes as load R varies. With Normalize engaged, it shows you the deviation from the reference, so a flat line = no change. This makes differences much easier to see than comparing overlayed curves. This will quickly show you the useable freq range of the xfmr in those conditions; the range where the mag/phase resp of the xfmr doesn't change over the load range it will encounter.

This becomes really critical when you use a xfmr at lower impedances. Unless they were specifically designed for it, the HF bandwidth of most xfmrs really narrows at lower load R's.

I tried a few setups, but they are clearly not correct. The OUT connected to the transformer with the INPUT A in parallel - I normalized that and it is flat. Then I move the INPUT A over to the output of the transformer - but the response looks nothing like what I would expect. I did the same setup on my scope with a signal gen and it is showing that -3db should be 10hz - 6Mhz.

The analyzer wasn't set up to measure A/R. And the scope isn't showing you the attenuation of the input signal by the xfmr input impedance at that secondary load R. Set it up as described earlier and try it.

Hope this helps.

[sixtimesseven]

Yesterday the postman delivered another boat anchor to my doorstep - A HP3577A VNA with (75Ohm  S-Parameter Test Kit).

Snip ...

Be careful with the 75  N connectors on the S-parameter test set. They will be wrecked if you mate them with 50  N connectors. The center pin is a smaller diameter on the 75 version.

I have both the 35677A 50 and the 35677B 75 test sets.

At lot of the time I use the 3577A for measuring Bode Plots on SMPS without a test set.

Regards,

Jay_Diddy_B

Oh I did not know that  Thank you I would have destroyed that connector. Now I have to order new adapters 

[sixtimesseven]

I probably just install a quiet 12V PC fan with a 24V converter and test with the thermal cam if the 80 CFM are enough without the CRT heating the instrument.

I look forward to seeing your results. Since you've been inside the unit, I think you can feel that most of the heat comes off the various PCBs, and it takes some pressure to get air to circulate among them all. I'm doubting if a standard PC fan will be sufficient.

Well I have to agree with that. I interpreted the pressure to be in mm not in inches. 0.28in of H2O is a lot 
Looking at Delta AFB 120 x 120 x 38 MM SERIES fans (https://www.mouser.ch/datasheet/2/632/AFB120x120x38mm-515857.pdf) the obvious replacement would be the: AFB1224HE. But I probably try the AFB1224ME which has a bit lower pressure (0.25inch) and CFM (~92)  but then the noise drop of 12dB to 38dB is worth the risk to me.

[precaud]

Looking at Delta AFB 120 x 120 x 38 MM SERIES fans (https://www.mouser.ch/datasheet/2/632/AFB120x120x38mm-515857.pdf) the obvious replacement would be the: AFB1224HE. But I probably try the AFB1224ME which has a bit lower pressure (0.25inch) and CFM (~92)  but then the noise drop of 12dB to 38dB is worth the risk to me.

Well if it works, 12dB reduction is huge and would be well worth the effort! Keep us informed, please.

[Berni]

I am just in the process of repairing one of those exact CRT modules in my HP 4145B since the NewScope5 is a bit pricy

It uses the same switches on the front panel and mine also have an inconsistent feeling. Some are soft and some are clicky. I am guessing that all of them ware supposed to be clicky but the springs got bent enough with use that they don't provide the buckling action anymore. But i prefer the soft feel of the worn out ones. I was guessing if jabbing the spring with a Flathead screwdriver would bend it enough to "break" the clicky action of the switch.

[bson]

An excellent workhorse - I use mine quite regularly, mainly for anti-imaging filters.  The nice thing about the 4-port model (I hardly ever use the S param set) is it can be used with my LeCroy active probes and the ADPPS power supply.  This way I can probe at the output of each filter stage separately, and if something odd is going on quickly track down what is invariably some mistake on my behalf.  (The same power supply also allows using the probes with counters and other instruments, just brilliant.)

Doing something about that jet turbine fan would be a huge improvement!

[precaud]

I am just in the process of repairing one of those exact CRT modules in my HP 4145B since the NewScope5 is a bit pricy

Is it just the switches, or is the display quality becoming worse too? At some point we should probably start a separate thread about repairing those modules and extending their useful life. The NewScope5 is very cool but, as you say, pricey.

I thought it would be worthwhile to mention a few things for those readers not familiar with the 3577A. Measurement-wise, it set a new standard of accuracy and performance for VNA's in its freq range. The main thing a VNA needs to be able to do is measure inter-channel magnitude and phase differences over a) a wide freq range, b) offering a wide range of suitable filter bandwidths, and c) over as wide a dynamic range as possible. These areas are where the early VNA's had limitations.

Take dynamic range. Up until then, VNAs had amplitude accuracy of +-0.5dB and phase accuracy of +-1.5º over a 60dB dynamic range (below full-scale), and only at mid-band IF's and frequencies. Max freq range was in the 10-30MHz range, and smallest IFBW was 3Hz (and not very stable). HP 3570A and Anritsu MS420 are examples of this.

The 3577A comes along and betters everything by an order of magnitude, with worst-case amplitude accuracy of +-0.08dB and phase accuracy of +-0.5º over the same 60dB dynamic range, extending it to amplitude accuracy of +-0.25dB and phase accuracy of +-1.5º over an 80dB dynamic range. And doing it at ALL IF's and frequency ranges. And offering a stable 1Hz IF and 5Hz - 200MHz useful bandwidth in the process. Remarkable.

It was an incredible work of engineering, both analog and DSP. And HP sold a ton of 'em.

The fact is, the newer VNA's don't perform any better over this same freq range. 25 years later, and it is still as good as it gets accuracy-wise.

My main beefs with the 3577A are:
: The noisey fan is almost unbearable.
: The display is too small and crowded, with poor contrast. Causes eye strain for me.
: Scattered user interface. Control grouping reflects the underlying firmware organization, not the best ease of use.
: Like most instruments from that era, its big and heavy. But it was best possible at the time.
: No built-in power splitter. A silly omission.
: No AC coupling option on the inputs. Another silly omission.
: Output and all inputs are bonded to chassis ground, requiring isolated/diff probes and/or use of output xfmr for ground-isolated low-level measurements.
: 1 Hour warmup time for stated accuracy. The instrument will spend 1/10 of its life warming up for use 

But considering that you will have to pay royal sums to find something that improves on these things, the 3577A is an incredible bargain.

EDIT: and worth mentioning... the 3577A has no firmware bugs I am aware of... I've never encountered one.

[sixtimesseven]

An excellent workhorse - I use mine quite regularly, mainly for anti-imaging filters.  The nice thing about the 4-port model (I hardly ever use the S param set) is it can be used with my LeCroy active probes and the ADPPS power supply.  This way I can probe at the output of each filter stage separately, and if something odd is going on quickly track down what is invariably some mistake on my behalf.  (The same power supply also allows using the probes with counters and other instruments, just brilliant.)

Doing something about that jet turbine fan would be a huge improvement!

May I ask which active probes you use exactly?
Thanks

[rx8pilot]

Well I won't put DC on the inputs of mine... all three of my fixer-uppers have input stages issues, and DC is often the culprit... besides, DC shortens the useful life of relay contacts...

Did you build some AC coupled adaptors? Like you, I would prefer not to test my luck! Addressing this now could save any 3577A owner a lot of trouble.

: No built-in power splitter. A silly omission.

Is it worth buying a commercial splitter for under 200Mhz or is it easy enough to DIY? Searching around for some info - they only seem difficult when the BW goes much beyond 500Mhz.

[Berni]

I have built a 3 way restive slitter before out of SMA connectors meeting in a triangle and some 0402 resistors joining them in the center and it seamed to measure just fine on the 3GHz network analyzer at work. It did get a bit wobbly past 2GHz but otherwise seamed pretty flat with a reasonably low SWR. Tho i did not measure how accurately it splits power and phase.

So for 200MHz i think a DIY one could work pretty well.

[Cubdriver]

I am just in the process of repairing one of those exact CRT modules in my HP 4145B since the NewScope5 is a bit pricy

It uses the same switches on the front panel and mine also have an inconsistent feeling. Some are soft and some are clicky. I am guessing that all of them ware supposed to be clicky but the springs got bent enough with use that they don't provide the buckling action anymore. But i prefer the soft feel of the worn out ones. I was guessing if jabbing the spring with a Flathead screwdriver would bend it enough to "break" the clicky action of the switch.

Berni, actually the 'soft' ones are those that are working properly.  They're the bog-standard leaf spring switches that HP used in skazillions of instruments back in the day.  The leaf spring is supposed to have a slight curve in it that lets it bend smoothly, but in older instruments they get sticky and don't want to bend properly, resulting in the feeling you describe.

Below is a link to a YouTube video posted years ago by a guy showing how to remove and reform the springs.  I've recently read that the issue can be caused by old lubricant gumming up and preventing the springs from moving properly, and that it can be fixed by flushing the switches with alcohol to clean out the grease rather than taking them apart as shown in the video.

I've flushed a few of them with denatured alcohol and been successful in getting them working again, though some were stubborn and ultimately needed to come apart.

I'd try an alcohol flush first (you may need to hit them two or three times) before going to the trouble of pulling the springs.  Easy stuff first, right?

Here is the video link:
http://www.rbarrios.com/projects/HPSWITCH/

Good luck with it.

-Pat

[Berni]

Oh thanks! I will give that a try.

Not only does my HP 4145 have this problem but also my HP 8566B, all use the same switches. I have expected them to be harder do disassemble but yeah will definetly try cleaning them first since doing this on every single switch could take a while.

[sixtimesseven]

I am just in the process of repairing one of those exact CRT modules in my HP 4145B since the NewScope5 is a bit pricy

It uses the same switches on the front panel and mine also have an inconsistent feeling. Some are soft and some are clicky. I am guessing that all of them ware supposed to be clicky but the springs got bent enough with use that they don't provide the buckling action anymore. But i prefer the soft feel of the worn out ones. I was guessing if jabbing the spring with a Flathead screwdriver would bend it enough to "break" the clicky action of the switch.

Berni, actually the 'soft' ones are those that are working properly.  They're the bog-standard leaf spring switches that HP used in skazillions of instruments back in the day.  The leaf spring is supposed to have a slight curve in it that lets it bend smoothly, but in older instruments they get sticky and don't want to bend properly, resulting in the feeling you describe.

Below is a link to a YouTube video posted years ago by a guy showing how to remove and reform the springs.  I've recently read that the issue can be caused by old lubricant gumming up and preventing the springs from moving properly, and that it can be fixed by flushing the switches with alcohol to clean out the grease rather than taking them apart as shown in the video.

I've flushed a few of them with denatured alcohol and been successful in getting them working again, though some were stubborn and ultimately needed to come apart.

I'd try an alcohol flush first (you may need to hit them two or three times) before going to the trouble of pulling the springs.  Easy stuff first, right?

Here is the video link:
http://www.rbarrios.com/projects/HPSWITCH/

Good luck with it.

-Pat

Nice link. How did he get the key of the switch (I mean the actuall boutton on top of the switch)? I tried to remove the keys since they have yellowed and I tought I could try the Peroxid/UV trick on them... But I didnt manage to remove them.

I had really sticky switches and all I did was spray them with WD-40 contact cleaner spray and they work perfectly now.

[precaud]

Did you build some AC coupled adaptors? Like you, I would prefer not to test my luck! Addressing this now could save any 3577A owner a lot of trouble.

Yes, I had a couple old Tek input capacitance normalizers which I repurposed. I used 100nF caps, which give -3dB @ 3.18Hz. There's probably room for 220nF's in there if I find that cutoff is too high. But I doubt that will be the case. If I need to examine xfer functions below 10Hz, an FFT-based DSA is a much better tool, anyway. Sweepers take forever at such low freqs.

: No built-in power splitter. A silly omission.

Is it worth buying a commercial splitter for under 200Mhz or is it easy enough to DIY? Searching around for some info - they only seem difficult when the BW goes much beyond 500Mhz.

Probably not worth buying. I made mine, using closely-matched 1/2W axial MF resistors I had on hand. It's good beyond the 30MHz upper limit of the Anritsu VNA I mostly use. I've never used it with the 3577A. And given Berni's excellent results with 0402's, making one is a no-brainer.

For power supply/PDN stuff you probably won't be working above 10MHz, anyway. When you do, come back and tell us about it! 

[Cubdriver]

I used one of those crook-necked lab squeeze bottles to flush the springs with alcohol - it lets you direct the flow and get some force behind it so it gets into the area under the spring.  Actuating the switch as you squirt it seemed to help, too.

Ideally you'd gently (so as not to dislodge the springs) blow them out afterwards with compressed air, but my compressor was down when I did the ones on my 8662A so I just let them air dry.

-Pat