Tektronix’s TDS/DPO7000 series Mechanical (Chassis) Design is Evil

I thought the TDS500~800 series design is already frustrating to service. But TDS7140 (or DPO7000) takes the cake. Whoever the a**hole designed the chassis made it a f*cking lettuce wrap. It’s not even an onion that you can predictably guess how you’d approach it.

To get any meaningful access to the insides, you MUST first remove the plastic front panel bezel, which is a fragile part that if you didn’t get the plastic hooks right, you’ll break it when you try to force anything. The service manual is not helpful. Likely written by somebody with a ‘fuck it. somebody’ll figure it out’ attitude.

This can be seen by the service manual giving an exploded view diagram without a precise order-of-removal dependency graph, nor the decency of telling you where each hook is and which of them are slides that must not be pried open like hooks. This is basically is tricking people to break the front bezel because those who didn’t know this already won’t know until they shine a flashlight to investigate the geometry around the hooks before releasing the front bezel.

It’ll make much more sense if you see the insides of the released front trim/bezel:

Not to mention that that black (lower) Acq front panel trim was double-sided taped to the BNC panel because it’s so easy to f*ck up the plastic to get access to the insides of the unit. The bottom trim is supposed to come out first (because there are no hooks) but I wiggled the upper part out first, luckily without damaging the bottom hooks, because of the silly double-sided tape used for the lower trim.

None of these shit was mentioned in the service manual. It basically said “there are some hooks in the front panel, pry them open and pull it out. Good luck!”. Mixed with a flimsy, non-intuitive chassis design, it’s pretty much telling outsider to break the front trim.

I really couldn’t believe some f*cker who’s a delight to deal with would design a case with an insane order of disassembly so I read the vaguely worded manual over and over again to see if there’s a step I can skip to get partial access to the PC motherboard inside. I couldn’t find any so I gave up and follow the painful disassembly order and realize the design was indeed evil and fully intended to shove it up the a$s of whoever that tried to repair the unit.

The service manual is also very unhelpful that they first told you to remove all the knobs yet the rest doesn’t depend on it. The real first step is to remove the front plastic trims. Nonetheless you need to remove the carrying handle because it’s bolted to the right side of the sheet metal.

The take away is that the Mushu-chicken lettuce wrap is that the top and left side is one sheet metal, while the right and bottom side is another sheet metal in one straight piece. What’s really frustrating is that you cannot just move the top-n-left half cover in isolation because it’s sandwiched by the right-n-bottom half cover, which in turn it’s bolted down by the screws that are obscured by the front trim:

This is madness!

HP/Agilent/Keysight doesn’t do this kind of abusive design that reeks evil. I should start charging more if people sends me a Tektronix to fix if I have to deal with the damned disassembly maze.

Another bit that shows whoever wrote this damned service manual doesn’t give a shit is the instruction to take out the 2-in-1 CD+HDD bracket. A normal person who dealt with Agilent Infiniium unit would expect the 2-in-1 bracket should come out on its own in one piece. This is not the case for this damned Tektronix scope.

Here is all the information available from Tek on how to do it (which is both incomplete, misleading and confusing):

Should not have used the word slide. Should use the word pull because the HDD+CD combo bracket is trapped in the unit so there’s only one way to get it out: by pulling. You can’t slide it from the sides as the sides or the top of the CD drive bracket are covered.

Figure 6–20 is just an exploded diagram, which covers everything taken apart without explaining the order and dependencies, which is essential because this damn thing was designed in a way that you wouldn’t have intuitively guessed the order by just looking at it.

Turns out after you removed all the screws that seemed relevant, the HDD+CD combo bracket won’t come out because it

The manual is surely misleading because if you just need to take the HDD+CD combo bracket out, you don’t need to take out the 4 #0 Philips screws. This is something you can do after you pulled the CD drive bracket out, but first you have the remove the two T-15 at the front of the CD bracket, which made it look like it was the bracket of the HDD+CD bracket combo, so you wouldn’t suspect the whole thing hinges on you pulling the CD drive bracket out first.

Tektronix certainly doesn’t have good excuses for these lame design with lousy documentation. Even with cost savings in manufacturing materials, it just doesn’t justify the labor. This is just taking shortcuts in design and it shows the lack of forethoughts and purely poor judgement.


This scope is lame everywhere. It really showed the people designed it either divided it into too many teams or they simply doesn’t give a shit. There’s no reasonable excuse for this other than laziness. The hard drive interface was converted twice! It takes an IDE drive, converts it to an SATA drive, then plugs into a board that converts the SATA to IDE back again just to fit the form factor.

Not to mention that the 44-pin to SATA adapter is a ribbon inside the caddy, which means it’s not carefully designed that you can slide the drive in. You have to pull the short cable out to plug it into the drive and push the bulk back in. It’s easy to mess the ribbon cables up with this design. It just screams the people who designed it hate their job.

 14 total views

All in 1 Mini Card Reader Manufacturing Flaw (Bad USB connector)

Recently, I’ve bought this CF card reader on eBay but it doesn’t detect at all.

Picture 4 of 7

I looked closely into the connector with a loupe and realized that the mini USB conductor was molded incorrectly. The center middle pin was pushed down because extra plastic was deposited above it:

The seller refunded in full but I figured that if the connector is malformed at the molding stage, buying it from another seller is not going to make it work, and the other form factors/connector configurations are inconvenient, so I tried my luck and see if there are exact matches for the connector they’ve used. Turns out it’s a 56 cents connector (price for 1 piece) available in Mouser (UJ2-MBH-1-SMT-TR):

In bulk, this connector can be bought for $0.22. For something that’s selling for $5/pc, the Red Chinese manufacturing had to go cheap to shave a few cents that ended up turning finished products into total trash. Most people are not electronics/troubleshooting savvy enough to figure out this shit, and the labor cracking the piece up and the SMD rework can easily buy 20pcs new. I just happened to have the tools (Metcal hot tweezers) so I can desolder the bad connector in seconds, but average users do not have that luxury so the neutered USB card readers go straight into trash.

We need more products NOT made in China!

 18 total views,  1 views today

Crack open wall wart (power brick): hot air method (READ or risk more damages!)

The power brick that came with my U1620A order has a rattling noise when I shook it. I suspected a piece of soldered cracked and was wandering inside so I chose not to power it on as it might short something. So I had to crack open the wall wart, which turns out many people have little luck open it up non-destructively.

Just prying it open by force with a screwdriver will ruin the plastics. I needed to somehow soften the glue first and work into it (like using a cutter) to dislodge/cut the glue. Rubbing alcohol doesn’t work that well. The glue is quite stuff so it’s not the easily dissolvable kind.

After some experimentation, I used my hot air station and set it to a temperature right below it melts the hard ABS plastic outside (143 deg C, max air flow), and carefully pried it after taking the heat away.

WARNING: Do not pry while the hot air is blowing as the ABS plastic might be slightly softened! The hot air might overshoot in its temperature feedback) so you might accidentally remold it. When working at the borderline temperature that barely soften the ABS plastic, the second you take the hot air away the plastic cools back (hardens) enough for you to pry.

Out of practicality, I heat up one edge at a time, then focus on 1/3 of the area at a time get an entry point prying one hook open at a time. It takes a lot of patience.

After I got all the 4 sides pried open for some reason I couldn’t release the case and it felt something was stuck in the middle. Turns out the wall wart I have happened to have a center screw hidden from plain sight under the label. The label quality was so good that just pressing it around with a pair of tweezers won’t find the screw hole (because of the tension). I used hot air (at the same 143 deg C) to probe with tweezers to locate the screw hole. I cut a hole in the label for the screw hole because it’d be a pain to use hot air to get the label off in one without ruining it and be able to put it back like new.


Bonus discovery: molded plastics have some memory!

I dented the ABS plastic case during prying (before I thought of using hot air), but I discovered this magical temperature (143 deg C) actually heals the plastic when I blow it in a direction that undoes the dent! The small (like 2mm) dents magically smooth itself out and went back to rough where it was before I screwed it up (pun intended)!

Yes, for molded plastics, small dents caused by prying and bumping, heating it up to the temperature right before it softens will return it to its near original shape without external forces!

Of course, do not go past the softening temperature (143 deg C). If you heat it to a melting temperature, the plastic will lose its memory so you need to apply external force to beat it to the shape you wanted (which is a pain in the butt and it requires a lot of post-processing to get it to look like original).

 26 total views

Agilent DSO6000 series was internally called 54670/54680?

Architecturally, the DSO6000A series shares common designs with 54830 series oscilloscope and 54640 series oscilloscopes. I noticed the Acq board numbers for my DSO6052A start with 54672 and the my M/DSO6104A starts with 54684. Looks like the 5467X means 500Mhz and 5468X means 1Ghz while the X is the number of analog channels.

This is ‘confirmed’ by a slip up in the documentation (user guide) which they forgot to update the model number in their vector graphics:

I think I’m getting a hang of Agilent’s oscilloscope’s hardware to do deep level board repairs as I have various model to compare. I used If you need yours repaired, please reach me at 949-682-8145.

 29 total views

RS-232 Stop Bits in Agilent Instruments

Turns out Agilent instruments do not use the same defaults for the RS-232 in their instruments.

54600 series uses 1 stop bits (most common):

RS-232 modules used in old 54600 series
54620/54640 series (newer 54600 series)

However other bench instruments such as power supplies (E3640 series 663X series) and 33120A arbitrary waveform generator uses 2 stop bits (fixed regardless of parity), which is usually NOT THE DEFAULT for most terminal clients:

E3640 series and 33120A’s RS-232 configuration.
Parity only trade away one data bit, so it does not affect stop bit
663X series powers supplies’ programming manual aren’t explicit about that except in code example

 61 total views

InfiniiMax AutoProbe 1 Caveats

For most mortal souls probing up to 12Ghz, Agilent’s integrated active probe system is called the the AutoProbe 1, which looks like this:

Regular full blown Windows-based Infiniium oscilloscopes takes any AutoProbe 1 probes (as long as the shape fits), but I noticed my DSO6104A (InfiniiVision 6000A series) do not take my 1152A (2.5Ghz) probes nor my fancy 1168A (10Ghz) and 1169A (12Ghz) probes.

Turns out the more compact, embedded (VxWorks) Agilent scopes that boots almost immediately. It’s called the InfiniiVision Series, which covers 1000 X-, 2000 X-, 3000A/T X-, 4000 X-, 6000 X-, 5000, 6000, and 7000 Series.

I’m not rich enough to get my hands on the X series, but I know from the architecture that 5000, 6000 and 7000 series are basically the same scope. 5000 and 6000 series looks almost identical while the 7000 series adds a giant screen and a slightly different keypad layout (the BNC ports do not align with the channel buttons and dials).

Turns out the datasheets shows two caveats:

  • 100Mhz model uses different hardware. They don’t take Autoprobe interface as there’s absolutely no reason why you need an active probe to get 100Mhz single ended. Agilent skipped the hardware for it (thus the autoprobe pins) altogether although they kept the recessed space reserved for Autoprobe so they don’t have to mold a different front bezel just for the 100Mhz models.
  • They basically take only Generation I AutoProbe I, namely the 1130 series
  • Gen 0 (not an official name) AutoProbe 1 does not work: 1152A (2.5Ghz single ended) for 54845A. These differential probes: 1153A, 1154A, 1155A, 1159A are also considered too old. They were intended to work with old Infiniiums such as 54845A
  • Gen 2 AutoProbe 1 (only 10Ghz 1168A/B and 12Ghz 1169A/B models) does not work. These embedded scopes usually max out at 1.5Ghz, with the exception that 6000X goes up to 6Ghz, which is still way below 10Ghz
  • N2800 series are Autoprobe I, but it’s Gen III (has a bigger butt extending away from the AutoProbe I hole), so it doesn’t work
  • The rest are Autoprobe II and III that’s beyond our mortal souls (and way out of the league of InfiniiVision scopes)
https://www.keysight.com/us/en/assets/7018-06782/data-sheets/5968-7141.pdf

 69 total views

E5070B / E5071B shutting itself down immediately after turning on

Got a second customer coming in today with a E507XB series network analyzer that does not turn on. Looks like it’s a common problem within the model series.

If the unit turns on without the CMOS battery but it doesn’t turn on the 2nd time after the CMOS was cleared, you have a very expensive problem which likely I’m the only person who can solve it because it’s months of effort tracing the circuit and the timings inside. I regretted chasing down the rabbit hole and spent more labor than 3 good unit’s costs for the few grands I’ve charged, but I might be able to recoup the labor in the future as more of the E507XB fails the same way.

UNLESS it’s a SC815E (there are 2 revisions that uses different motherboard), it’s NOT the motherboard or the power supply. I’ve seen some faulty SC815E doing the same on other model series, but not the VP22s (I replaced 4~5 different VP22s, they all do the same thing). It’s some good timing issues that’s hard to pinpoint to a specific module (you can replace everything on the digital side and it still doesn’t work) which I had to design, build and test a special circuit to correct it.

If you have a unit E5060 series or E5070 series VNA (Vector Network Analyzer) that doesn’t boot, I’m likely to have the exact experience fixing it. Of course I’m open to solving other problems with the analyzer as well. I offer free eval (no fix, no fee). Please email me at owner@humgar.com or call me at 949-682-8145.

 80 total views

HP 54502A Datasheet typo about AC coupling

The cutoff frequency of 10Hz on the datasheet is a typo. Better scopes at the time claims 90Hz. 10Hz is just too good to be true.

Found the specs from the service manual:

Don’t be fooled by the -3dB cutoff and ignore how wide the transition band can be (depends on the filter type and the order). Turns out this model has a very primitive filter that AC couple mode still messes square waves below 3kHz up despite the specs says the -3dB is at 90Hz. You better have a 30+ fold guard band for old scopes!

Remember square wave pulse train in time domain is basically a sinc pulse centered at every impulse of the impulse train in frequency domain superimposed. Unless you have a tiny duty cycle (which is not the case for uniform square waves, they are 50%), the left hand side of the sinc function at 1kHz fundamental still have sub-1kHz components that can be truncated by the AC coupling (high pass filter).

 136 total views

HP 54003-61617 Probe = HP 10017A

I have a 54003-61617 probe lying around and I never got a chance to find out what bandwidth so I rarely used it. After some digging, thanks to searchable PDFs, I found on 1986-04 edition of HP Journal (Archived copy here) that 54003-61617 probe is equivalent to 10017A:

By the way I noticed HP Labs still had the old HP Journal PDF files hosted on the website except without indexing: https://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/YYYY-MM.pdf, replace YYYY with year and MM with month.

HP Journal (hparchive.com has an excellent collection) is an excellent source of hard-core electronic engineering education materials, better than anything you can get in colleges because electronic circuit design (not IC designs) is not an academic research area anymore. As of 2000, the only way to get into this area is to work at companies (apprenticeships) instead of formal training like classes. There’s a little problem though because this specialty (analog electronics) is so disorganized, very often even veterans in analog electronics has blind spots like not taking advantage of math tools/thinking enough (they tend to be very good at back-of-envelope calculations).

Finding the specs for 10400A series (10017A) is not easy either because the datasheet is not on Keysight (the new name for HP’s instrument division). It’s listed in an 1998 catalog “How to Select A Probe” kindly hosted by Marc Mislanghe (who passed away in 2014) as HPMemoryProject.org that listed the specs of 10017A in attempt to find an approximate ‘modern’ substitute:

HP 10017A (54003-61617) mini-probe has the following specs:

  • Attenuation: 10:1
  • Input Capacitance: 8pF
  • Input Resistance: 1MOhm
  • Bandwidth: 300Mhz
  • Compensation range: 9~14pF
  • Takes up to 300V
  • 1 meter long cable

Even older record, Operating Note Part No. 5955-6270 courtesy of HParchive.com shows:

I bought a bunch of snap-on ground lead alligator clips (MP2+MP3) and grabbers (MP7) for the probe

Figure 2. 100XXA probe with hook tip or grabber

The snap-on alligator ground clip actually has two parts: the alligator clip (MP2, 5061-1258) and the snap-on ground lead (MP3, 10006-61301) and they are screw-mated :

I happened to have bought a pack of multiple new ground leads and grabbers (MP7, 10017-69501) for this model series, more than what I’d need. They’ll fit miniature probes models 10017A, 10018A, 10040A, 10041A, 10042A.

These accessories will also fit 10021A, 10022A, 10026A, (10027A?), 10032A, 10033A as well.

 149 total views

Demystifying comparisons between eBay managed payments and the old way Hint: it's still the old final value fees (before cap) + payment processing fees (analogous to Paypal)

Most often when we sell on eBay, we receive Paypal payments, so we ended up paying eBay’s commissions (also called the final value fee, FVF for short) and Paypal processing fees as a percentage of the sale.

Now eBay pushes out Managed Payments (MP) which combines payment processing fees with the eBay commissions (FVF) because eBay now manages the payment gateway. The rest is every time you sell something, you get a payout (sales – fees) deposited directly to your bank (it used to be collecting paypal balances then withdraw it).

They have a different calculation formula which they claimed the sellers are better off in most cases, but should we take eBay’s word for it? Regardless of whether you are enrolled in managed payments, the fee percentage for each sale depends on:

  • eBay store subscription level
  • category

It’s impractical to do the side by side fee structure comparison to see when we are better off for each sale, plus we cannot easily switch between the two plans.

It’d be very helpful if we can put the managed payment fee structure on the same form as the conventional eBay+Paypal fee structure, and figure out under what conditions we are better off or worse off with managed payments.

Initially I was ready to do the derivations to put both plans on the same scale, but I spotted that managed payment (combined) percentage is simply the vanilla (non-managed) eBay category final value fee + 2.35% payment processing fees! That’s how they’ve calculated the combined managed payments percentage!

This makes life a lot easier. Since I can factor out the 2.35% that applies to the whole sum (which also include shipping and sales tax) regardless of the fee cap, this works exactly the same as Paypal (which charges 2.9%) and we are getting a 0.55% discount in payment processing fees.

For managed payments, since we’ve already separated out the payment processing fees, the fee cap applies to the vanilla final value fee portion which is equivalent to the old eBay final value fee. Keep that in mind.

The part that has changed is the fee cap. The old way caps the commissions/FVF directly regardless of product category, yet the new way (managed payments) caps the sale price that are charged commissions. This means the fee cap goes up or down a little depending on the final value fee percentage class applying to your sale.

eBay set the (commissioned) sale price cap to closely match the realized commission cap in the old way, for example non-store subscriber who will have their raw final value fees capped at $750 will see the same cap for the most common 10% categories (12.35%-2.35% = 10%) because the (commissioned) sale price cap is $7500, which $7500*10%=$750.

Big industrial equipment also have the same cap regardless because $15,000*2% = $300.

For non-store subscribers, 10% is the anchor (iso-fee-cap) category. So you are a little worse of with books (price cap +2%) and better off with musical instruments (price cap -6.5%). 

For store subscribers, you get a bit more break over the fee-cap (lower) because your final value fee percentage is lower than the anchor (likely they chose a breakeven point at 10% when determining the sales price to cap final value fee over. Just easy numbers, not rocket science):

Managed Conventional Discount*
Common (9.15%) $2,500*9.15% = $228.75 $350 (Small stores)
$250 (Big stores)
$121.25 (Small stores)
$21.25 (Big stores)
Heavy gears (1.5%) $15,000*1.5% = $225 $250 $25
Books (12%) $2,500*12% = $300 $350 (Small stores)
$250 (Big stores)
$50 (Small stores)
-$50 (Big stores worse off)
Guitars (3.5%) $2,500*3.5% = $87.5 $350 (Small stores)
$250 (Big stores)
$262.5 (Small stores)
$162.5 (Big stores)

I call Basic/Premium ‘small-stores’ and Anchor/Enterprise ‘big-stores’.

So here are the observations, which is all you need to know:

  • Under managed payments, small-stores gets a better deal than big-stores, because the advantage of the $100 lower fee cap with big stores are eliminated with variable fee caps.
  • The breakeven line for small stores ($350 cap) is 14% fees ($350/$2500), which the highest category is 12% so far. This means small-stores are ALWAYS better off switching to managed payments.
  • The breakeven line for big stores ($250 cap) is 10% fees ($250/$2500), which the only category above that right now is books (12%). Big-stores selling BOOKS are worse off with managed payments.

So in other words,

  • everybody is better off with managed payments (fee-wise) EXCEPT big-stores selling books
  • under managed payments, there’s no final value fee advantage for being a big-store

* Remember you got 0.55% discount over the payment processing portion of the fees too and is not shown here since we’re just talking about savings in vanilla final value fees.

As far as books (12%) is concerned, if you are a big-store owner, your raw commissions cap raised from $250 to $300 because $300 = $2500 * 12%. But if you factor the 0.55% discount, if the sale price is $x,

     $$ (0.12-0.0055) x = 0.1145x > 250 $$ $$ x > 2183.4 $$

Since the raw commissions stops at $300 ($2500 * 12%), the additional $50 cap increase starts to get offset (and turn out positive as the processing fee discounts outweighs the commissions cap increase) when the payment processing discounts ABOVE $2500 covers all of it:

     $$ 0.0055(x-2500) > 50 $$ $$ x>11590.91 $$

So the range of sale price x which the only setup (books for big-store sellers) can do worse is

     $$ 2183.4 < x < 11590.91 $$


There are some ambiguities (technically incorrect documentation) on eBay’s website which implied vanilla final value fees (portion) are charged for sales tax. I made a sale and checked the numbers and it’s not true. Only the 2.35% payment processing fee portion is charged against the sales tax (like paypal for handling extra money), the category final value fee (in my case 9.15%) is not applied to the sales tax.

They actually meant “the 2.35% payment processing fee portion” when they said “managed payment final value fees”. This is also part of the reason why I wrote this article, because they do not use the language that conceptually separate the two portion of the combined final value fees (vanilla final value fee + payment processing fee) in managed payments, thinking that they are simplifying the math for sellers, without realizing if the two concepts really fused into one, they’ll be shortchanging sellers over sales tax.


The $0.3 fixed per-transaction fee applies to both managed payments and the conventional way (Paypal also charge +$0.3 fixed per-transaction), so nothing has changed.

 229 total views