Refurbishing a Tektronix TDS7104 Oscilloscope | Hacker News Reader

Refurbishing a Tektronix TDS7104 Oscilloscope

tomverbeure.github.io

152 pointsjwise066 comments

I just had claude walk me through refurbishing a Rohde & Schwarz signal generator from the 90s. It feels pretty cool getting a multi thousand dollar piece of equipment for 60 bucks even if you have no idea what you'll use it for.

Was that by any chance the R&S AMIQ that I refurbished in an earlier blog post?

And yes, it's a lot of fun bringing these bargain bin pieces of kit back to life.

My model is an SMIQ03E actually! I've grown quite addicted to going to my local 2nd hand electronics store in recent months scrounging for anything with an rs232 port!

I looked for a SMIQ for quite a while to go along with the AMIQ, but eventually settled for the R&S SFQ TV test transmitter that you can see here: https://tomverbeure.github.io/2026/06/28/Rohde-Schwarz-AMIQ-....

It can be used a vector RF signal generator with I/Q inputs. The only limitation is the reduced input bandwidth.

Very cool! Yeah I lucked out finding mine. It was in the intake section at my 2nd hand electronics store. It had a broken power button so they were selling it "as-is". A button repair and battery swap later I have a 3.3Ghz VNA. Can't complain!

That's a great find, but afaict that model doesn't have VNA capabilities, does it? However, if you also stumble on a spectrum analyzer at that second hand store, you could combine them into a scalar network analyzer.

Oops yeah you're right, it's a VNG. A VNA would be great to have though! I wonder if these agents are going to just increase the utility of ever piece of equipment people can hook them to.

A friend of mine bought an SMIQ-03b on a whim, we've replaced the mmic amplifier chip that seems to always fail, now we're trying to figure out why the ALC doesn't work. It generates signals just fine, but the fine attenuation does nothing.

Yeah that pretty close to some problems I'm working on with mine. It has an "IQCON ALC loop failure" and claude was able to diagnose that the 2013 rail is around 11.44v (outside of the 10v range iirc) and the 2015 one is zero instead of the min of 0.13v. The RF paths looks mostly healthy as far as we can tell but I've only tried 10db steps and it seems consistently 4-5db low, assuming I'm using my oscope right.

Which MMIC did you replace? And what errors did you see? I have 110 and 111.

I'm not sure which he replaced, but having started trying to grok how it actually works, it's chock full of HFET and other amplifier chips, and over-stressed RF switches.

I'll report back if we ever get it actually working properly.

Did you fix it? What was wrong?

Some parts. It's still a WIP!

Yeah 1 GHz is nice, but that tiny 200 MHz Siglent is going to be much more useful in practice for most things.

Speaking of boat anchors, I'm trying to get my company to buy a DSA91304a (13 GHz scope with SERDES debugging). ~$44K with all the options, but still Windows-XP. The Windows-7 ones unfortunately don't have the options I want. Maybe there is a hack..

As I wrote: the TDS7104 is now sitting on a shelf in the garage, next to an Agilent 54831 and some other boat anchors. My daily drivers are a Siglent scope and a puny TDS220, which is about to be replaced by a TDS3052 that I'm bringing back to life. Hobbyists rarely need more than 100 MHz BW.

I hope you're able to thoroughly test that DSA91304a before dropping $44k on it!

I hope you're aware of the TekWiki:

https://w140.com/tekwiki/wiki/Main_Page

Mine is a TDS224. Cheap, but I only use it for keeping my Eurorack synth going. 100MHz is overkill for audio. Although some modules use STM32 microcontrollers, so maybe the speed will come in handy.

TekWiki is great!

100MHz is overkill for most hobbyists use cases, and definitely repairing test equipment: it's always the power supply.

TDS220

Cute little thing. The reality is you hope that's all you need.

200 MHz is good enough for most things. And for modern computing electronics even 1GHz is too slow, e.g. monitoring USB3 or high speed PCIE data lanes.

Did you get a price quote for that keysight oscilloscope? I see that you can also rent them for $2000 a month.

We've rented them in the past- always we're not quite sure we're done so keep them much longer than necessary. They're kind of a pain to pack and ship.

I have a 1.5GHz TDS7154B in addition to my 200MHz SDS1204 specifically to measure power rail noise with a low noise probe.

You just can’t characterise power rail noise well with less than 500MHz of bandwidth.

10 GS/sec is very impressive as far as I'm concerned! One sample every 20mm or so at typical transmission line speed.

That said, anything with a floppy drive scares me because it undeniably indicates how old it is. My SDS7102V is no Tektronix, but an honest 1 GS/sec and at the time, under $200 on eBay.

Love this. I knew nothing about oscilloscopes until just a few months ago, when reasearched and bought one from a science lab. It wasfor a personal side project. I gutted it out and installed a mini PC inside, and now i run AI models into it. It's like my personal AI assistance. Hooked up to OpenAI realtime voice, grok bot, claude code, etc so I can command it to do stuff even when my main computer is off.

I just mainly love the retro-moder look of it. Here's a link to how it looks, for those curious:

https://x.com/slhomme/status/1965131874734584291

I had now idea there was such a huge second-hand market for those. It was such a great opportunity to learn about this tech.

A pity you can't still use it as an oscilloscope though.

I was initially confused how you managed to get amber and green screen together, but I guess you put an LCD in ..

Give it a month and the acquisition ASICs will blow up and the caps will eat through the board traces. Can’t stand these scopes.

Nice, so I'll get even more entertainment out of it!

The equipment that's bought and sold at the electronics flea market typically isn't used as a means to an end, it is the end. There's disappointment when you buy something there and it just works. There's no fun in that.

I don't really have a use for a R&S AMIQ, but replacing all the PC motherboard caps taught me a lot about good desoldering practices. https://tomverbeure.github.io/2026/06/28/Rohde-Schwarz-AMIQ-...

Yeah agreed. I’ve fixed more Tek equipment than most people out there. In the order of hundreds of units.

These are just exceptionally bad.

I'm blanking on the exact models I used; I'm fairly certain they were tds700s or tds71xx, but a lab I ran had about a dozen units and we had nothing "fry" in the 4 years I worked there. What applications would result in these catastrophic failures?

Over the years, I've had 5 TDSxxx series scopes, and I still have a few. The common issues are:

NuColor CRT displays going bad. Bad capacitors. Dallas NVRAM units battery running out of juice

Fried electronics are pretty rare.

We had some "ghosting" on the displays on a few now that you mentioned it, but they were still 100% usable. I guess my expectations were lower since I never had to foot the bill for the lot of em (which at 20k/unit I guess I'd be upset about when one was faulty)

i thought maybe it was just something regionally here or the way people stored em, but i was in a position to refab many a few years ago and totally agree. tek had 100% repart rate and seemed like anything else (brand wise) was always in better condition internally even when way older.

I’ve seen rust in a newish (Danaher era) one.

A little while ago I bought a 386 laptop. It pretty much had everything wrong with it. Spent weeks repairing it. Best $40 I ever spent.

That feeling when it finally works is great. I spent $40 on this basket case: https://tomverbeure.github.io/2025/12/24/Repair-of-SRS-DG535.... Took me ages to figure that one out.

For a brief second, I thought this was about an original 7104.

Are there specific TDS models that you have found to be better, btw, reliability-wise?

>an original 7104

I've got one, they are very cool. I also have a 7834 (storage tube), cool in a different way. On the other hand, they are taking up a lot of space... I also have two 485s, one with a broken trigger.

Other than the very bottom end ones, no. Mostly went to Hp/Agilent instead.

Thanks for answering, not surprising. I think the transition from all-hardware to Windows-driven digital hard on Tektronix and many other storied engineering teams.

The TDS500, 600 and 700 series can expand memory by removing zero ohm resistors.

My TDS754D is waiting for the resistor and capacitor mod that converts it from a 500MHz/2Gsps scope into a TDS784D with 1 GHz/4Gsps.

Still, entering an easy to generate license key is less intrusive than removing a resistor.

Is the PC mainboard separate from all the tek ASIC etc stuff? If so, what's the interface? PCMCIA? PCI?

I'm just wondering how difficult it would be to use a more modern bit of kit in there instead of the Celeron.

Yes, it's completely separate. There is a PCI2PCI bridge between the PC motherboard and the PowerPC controller board, which has in turn an interface PCB to the acquisition board.

People have replaced the RadiSys motherboard with one from a Gateway PC. There aren't a lot of other options though: the motherboard must be an NLX type and that was a short-lived format. It's different for scopes like the Agilent 54831 (https://tomverbeure.github.io/2026/03/28/Repair-of-Two-Agile...) that I revived a couple of months ago: people do crazy things with that, making it work with higher resolution LCD panels, upgrading to completely different CPU classes, running Windows 10 etc.

It's remarkable that these scopes have both a PC and a PowerPC motherboard. I think they were a reaction to Agilent scopes: they wanted a PC front-end and reused the PowerPC infrastructure with VxWorks OS that they already had from other generations, so they just slapped an extra motherboard on it.

A bit ago a semi-retired guy came back to help with some stuff including debugging serial ports. It was very cool to see him fiddle with the knobs and be like "oh those squiggles look good" "no that doesn't look right" etc. https://i.imgur.com/R5aEXbq.jpeg

Meanwhile total mystery to everyone else how to operate/interpret the machine but a lot of fun

One totally needs a 500 MHz oscilloscope to debug a kHz range serial port signal! :-)

Family and friends aren't impressed by experiments measuring C either...

Heh, like my 100MHz scope used for audio signals.

Flebay has some used ones around $1.5k. I wonder how much this model MSRP brand new? Either way, $300 is a steal!

I think they sold for $60k back in the early 2000s.

I remember seeing the Logic Analyzer from that same series in a lab I worked in in the early 2010s, was told it cost as much as a house at the time.

This blog post shows a list price of $24,490 for the base model without options: https://www.jhongelectronics.org/2024/01/tektronix-tds7104-f....

Just wait until it needs a repair, and they tell you it's no longer being serviced. Been there.

This is one of the cases where it might be better to lease it.

Many fond memories of using these. Glad someone is keeping them going after 20+ years of service.

I recently did something similar with a TDS7154B. Rather than going with an IDE/PATA drive I was able to use the SATA Gen1 ports on the motherboard and a molex-sata power adapter to use a modern SSD.

Using SATA gen 1 would be ideal, but the TDS7104 motherboard is old enough not to have that: it doesn't support booting from a USB stick.

That's a killer scope even today after a good refurbishment. Still beats many new mid-range options for serious work.

Amazing photos!

s2idle in vertical Hz.

I went through a similar process recently with an HP/Agilent 54845A scope (1.5 Ghz, 8 Gsa combined) - a great tool if you need reasonably fast rise times and a reasonably fine time base resolution - the only issue is that it has an abysmally low sample memory for today's standards, so you get to see the trigger and "just a bit more". Well the other issue is that it pushes a lot of air... Like a lot, jumbo jet a lot. The scope luckily was in great shape and had a fresh test cert from not long ago.

I was going down the same route as the blog post (MSATA->PATA adapter), but simply imaging the HDD onto flash also didn't work.

The scope had to go into storage so I never finished this, but the SSD was being detected albeit at a lower capacity and no amount of twiddling with LBA and CHS geometry made a difference. My next step is going to be zeroing the drive, booting the scope with it as secondary, and then imagining it from within DOS/Windows. My hope is that the data will just end up at some offset. Or not even imaging it - just partitioning it, formatting it and doing XCOPY + SYS.

You may want to image the drive every so often and accessing the drive is a bit of a bastard (fewer than 32 screws on the HP but it's a pain regardless) - so it might make sense to actually get an SD->PATA adapter and mount it onto a PCI blanking plate, so that you can take it out easily. You may need a longer ribbon cable. There are also some 2.5 inch HDD brackets that mount onto a PCI blanking plate. Possibly a DIY job to replace the FDD with a plate that has the SD card mounted. Mine came with a ZIP drive.

I added the following software extras to the OS which I would recommend to anyone with an old scope like this that has an Ethernet port:

- An SNTP client so that you don't have to worry about the time and date (https://www.timesynctool.com/) - open source, win95 and up, just disable all logging to disk so as not to introduce extra writes and drive wear.

- TightVNC (1.3.10 I think) - a version that works on older Windows. Accessing it while the scope app is running slows things down significantly, so this is only for configuration changes and whatnot, also this particular scope writes directly to the VGA framebuffer, so you don't actually get to see the scope output via VNC!

- Shared some folders from the scope. This required a significant amount of fuckery to mount an SMB1 drive elsewhere without ruining one's existing Samba config, but it helps quite a bit to be able to exchange files with the scope.

Hardware extras that proved useful in a lab setting:

- One of those cheap VGA to USB frame grabber adapter things - they are detected as a video device, don't need drivers and you can watch your scope with VLC or your OS's built-in camera app.

- A remote-controlled power outlet. Speaks for itself - no need to touch the scope to shut down and restart an experiment from wherever you are.

- A GPIB to USB adapter if your scope has a GPIB/IEEE 488 port. Current Linux support is in a bit of a state, but anyway, this will allow you some degree of automation when reading the math / stats / other data, and will sometimes allow you to do more than the scope's UI will. I know for example that with GPIB I can control the signal on the AUX port on my scope, for which the GUI has no options.

- Not tried yet: this scope uses a PS/2 mouse (remember that they are not hot-pluggable) and a DIN keyboard. This bad boy: https://github.com/rasteri/HIDman - will let you plug in your wireless USB dongle and control your scope with your everyday keyboard and mouse.

The above significantly improves the "workflow" so that you hardly need to touch your scope or even swing your chair, or can use it remotely, and applies to most of the late 90s - early 2000s Windows scopes.

Not sure if you saw it, but before doing the TDS7104, I refurbished two Agilent 54831 scopes: https://tomverbeure.github.io/2026/03/28/Repair-of-Two-Agile.... For the one that still had a HD, copying the image from to an SSD worked without any issues. However, I also had a 54825A where it didn't work (I sold it, no blog post about that one.) So it's clearly hit-and-miss.

I found the 32 screws of the TDS7104 less of a pain than the dismantling the Agilents which had less screws.

> I found the 32 screws of the TDS7104 less of a pain than the dismantling the Agilents which had less screws.

Yeah sliding the whole thing out of the shell and the (curses) front panel ribbon that gets in the way, and the best part is that if it pops out on the front panel end, the tab is held in place by a metal prong that you have to bend (break away) to undo the tab to secure the ribbon. The prong is not screwed down, it's riveted. Breaking that thing away was the best thing I could do without ruining the panel. I ended up also gluing that ribbon to the chassis so that it doesn't get caught again.

Edit: I also bought mine to use it, hence all the extra bits around it.

My 54825A had a power-up problem: if you unplugged the AC cord for a few hours, pressing the front-panel button wouldn't power on the scope. But if you plugged in the AC for about 5 minutes, it would work fine again.

That sounds like an easy thing to fix, but the whole power-on architecture was a baffling Rube Goldberg device from the button on a small support PCB to the front panel PCB to the PC plug-in card to the power supply. I've never seen anything like it.

I disclosed the flaw to the buyer (the price was very low) and he was excited to have something to work on. :-)

There are exceptions to the "Winscopes" from the same era in the 1 GHz+ space (RTOS like VxWorks but with Ethernet and modern goodies) that are not HP/Tek/R&S, but I'm not listing them because I'm out on an active hunt for one ;-)

> not HP/Tek/R&S

That doesn't leave a whole lot of alternatives, Mr Lecroy.

...sorry, not HP/Tek/R&S/Lecroy ;)

$50 bucks would be a lot for that amount of work.

It's a hobby, not work.

$300 bucks is at least three grocery trips for two people, shopped wisely. That stands for work to me.

Reply on news.ycombinator.com