Category Archives: Repairs

True LC900 Brake Assembly Repair

How to save a $3000 exercise bike using a $12 bearing

Mr. Determined Enthusiast’s mom called the other day saying that she needed to buy a new exercise bike, because her existing one, a True LC900, was making a really loud vibrating noise while pedaling. She had a repair person out to the house who took it apart and said that it needed a new “brake motor”. Which would cost $900 to buy and install. She figured at that price, she might as well get a new bike. People, this is how our landfills get full.

The Determined Enthusiast will not stand for a $3000 bike being thrown away because of a noise. And whatever ecosystem of manufacturer + parts supplier + dealer results in it costing $900 to replace a $12 bearing, does not get rewarded with another $3000 of revenue. Not today anyway.

The bike uses a simple motor/generator assembly as its resistance unit. Made by the Chi Hua company in Taiwan, it’s basically a three-phase permanent-magnet generator and a flywheel. That transformer looking thing on the side is an eddy-current brake, which makes sense, but the generator adds a nice touch — it allows the bike’s electronics to be powered by the rider’s own pedaling. The Determined Enthusiast approves.

Photo credit: Chi Hua

Sure enough, spinning the rotor caused a clicking sound at even intervals. Spin the thing up to a few thousand RPM and that turns into a racket.

Diagnosis

    The noisy bearing was behind the aluminum pulley located in the left picture.  This pulley had a heavy press fit (foreshadowing), and was removed using an OTC bearing splitter and a shop-made puller.

After removing the aluminum outer carrier, the bearing itself was visible:

See it? There’s a crack in the inner race at the 9:00 position.  Yet still, the recalcitrant bearing required a puller to remove (more foreshadowing).

    The bearing is an SKF 6202-2Z/C3, made in Italy.  This is a common bearing with 15mm ID, 35mm OD, and 11mm width.  The 2Z indicates double shielded, and the C3 is a slightly looser-than standard internal clearance, meaning that there’s some slop in the bearing to accommodate either temperature rise (which would close up the clearances), or maybe the misalignment inherent to the sheet metal housing held together by shoulder screws.  A heavy press fit on the shaft would also be likely to take up some of the bearing clearance (yet more foreshadowing).

    Removal of the shields revealed a bearing full of grease in good condition.  After removal from the shaft, the crack closed and the bearing ran relatively smoothly.   I punched out the ball retainer and thus disassembled the bearing, revealing nearly perfect races and balls (other than the crack).

    In operation, this bearing is relatively lightly loaded and operates in a clean indoor environment.  Some heat generation is to be expected with a brake of this type, but it’s not clear how hot the bearings actually get.  This bearing has most of the drive belt tension loading it, but all of this is well below its design spec. So why did the inner race crack?

Going Deep(-ish) on Bearing Fits

Given the heavy pressure required to remove the pulley and bearing, it seemed worth investigating the bearing fit.  Bearing fits are standardized and the choice of fit depends on a number of factors including whether the inner or outer ring rotates, the load, and other factors.  

This is a relatively lightly-loaded application, running in a clean indoor environment and probably not at appreciably elevated temperatures or at particularly high RPM – most likely the standard recommendations will work.  But to go through the exercise:  The wheel weighs about 10 lbs, centered between the two bearings.  The wheel is driven by a 10-rib belt attached to a pulley outboard of one of the bearings – Gates specifies a belt tension of 150-200 lbs for that belt, which seems about right.  You typically pedal at 120 rpm max, and the cranks are geared up by approximately 16x in two stages, for a flywheel speed of about 2000 rpm max.  So far, all of this is well within the spec of a 6202, which can handle 840 lbs static load and up to 20,000 rpm.  There is also a minimum load, below which the balls will skate and slip instead of rolling, but we’re okay on that spec as well according to the nifty SKF Bearing Select tool. 

For a bearing of this size with rotating inner race and normal loading, both SKF website and the Machinery’s Handbook recommends a “j5” fit, which is classified as a Transition fit, meaning that it falls between Interference and Clearance fits.  The Bearing Select tool recommends a “js5” fit, which is 1um tighter (for a 15 mm shaft).  For a 15mm basic dimension, the tolerance is -3/+5um (js5: -4/+4um), meaning the shaft OD should be between 14.997 and 15.005mm.  The ID of the inner bearing race also has a tolerance, listed as -8/+0um, meaning that the bearing’s ID should be between 14.992 and 15.000mm.  Combining these, the standard helpfully tells us that the resulting fit is between a -13um interference and a +3um clearance, with -11/+1um being the “probable” fit range.

Using a recently-calibrated Mitutoyo micrometer, I measured the shaft’s bearing seat OD at 15.013um.  At least material condition (i.e. best case), this provides a -13um interference, which (barely) meets the j5 fit spec.  A max material condition however, the worst-case interference is -21um, which is almost twice the maximum allowable interference.  Reviewing the table, only the heaviest fits allow a 13um oversize for a 15mm shaft: k5, m5, m6, and n5.  (k5 allows only 12um, but my measurement could be off by a micron.)  And these heavier fits don’t appear to be used for ball bearings with shafts this size under any loading circumstance.

 Why such a tight fit?  A few possibilities:

  • Manufacturing error – the shaft was ground too large, and/or a fine-sizing operation was either omitted or eliminated to cut cost.  Probably the most likely.
  • Manufacturing compromise – they wanted to make the shaft the same size all the way across, and the wheels needed a certain size for their press and were already made
  • Poor manufacturing tolerances – the shaft grinding process wasn’t well-controlled enough to meet the -3/+5 spec, so the engineers specified a heavy interference to ensure that sufficient interference would still be achieved at LMC.
  • Design error – the engineer accidentally specified the wrong fit.  This doesn’t seem too likely given how well-engineered the rest of the assembly appears to be, and how straightforward this application appears to be.
  • Counterfeit bearings with poor tolerances – leading the engineer to specify a larger shaft to ensure sufficient interference.  Possible, but these are going into expensive products.  No one is going to see the bearings anyway, so why bother with a counterfeit when 6202 bearings are basically commodities?
  • Design compensating for previous failure – this seems possible.  Maybe Chi Hua had experienced shaft or bearing failures due to the bearings moving on the shafts for some reason.  So they called SKF, who told them a heavier interference would be okay for their relatively forgiving loads and environment, and so they spec’d it as such.  While this is possible, I haven’t found any references online to ever needing such a tight fit.

“They say that Shaft is a real bad mother—“

Pressing out the second bearing was much easier than the pulley and the first bearing.  Once it was separated from the other parts, it could be inspected more thoroughly. There’s a long gouge, roughly 180° away from the keyway, where the pulley and first bearing ride:

At first I feared that I had caused this while pressing the shaft out of the wheel.  However, the broken bearing shows marking in this exact area, right where the crack is.:

Further, the aluminum pulley also shows a gouge in the same area, and I pulled that off before doing any other pressing.  The wheel also seemed to have something in its bore in the same position.  Maybe it was casting sand that didn’t make it out, and then gouged up the shaft?  Hard to tell but it seemed like dirt or rust.

So there you have it. The combination of an excessively-heavy press fit, plus the stress riser of the burr on the shaft, then cyclically loaded over years of use, and the race finally gave up and cracked. The balls clicked each time they passed over the crack, which created the noise.

Repair

Mr. Determined Enthusiast is well aware that the right fix for this is to turn a new shaft, grinding it to the correct size and cylindricity before pressing on two new bearings in an ISO 1 clean room, while someone with a clipboard dutifully records the proceedings.

But life intervenes and sometimes things need to just get done. So the fix was simple: the shaft was stoned to remove the worst of the boogered-up nastiness (bringing it closer to the diameter spec in the process), and the whole thing was reassembled with two new bearings, using sane amounts of force on the press thanks to the now-more-correct shaft size and lack of burrs tearing up the other parts. The brake is now whisper quiet again, ready for the onslaught of raw power from Mr. Determined Enthusiast’s 80-year-old mother.

Federal-Signal EQ2B Amp Repair

The Federal-Signal Q2B is the classic mechanical fire-engine siren, essentially a scaled-down version of an air-raid siren. The EQ2B is the electronic version, and F-S was so intent on getting the sound right that they put a DSP chip in an otherwise very old-school big-iron transformer-coupled amplifier.

But the fancy-pants front end doesn’t concern the Determined Enthusiast, because this one had a blown-up power stage. Here are the basics:

Green engineering paper = I definitely know what I’m doing

Looking not at all unlike a tube amp from the 50’s, this is your typical Class AB push-pull transformer-coupled topology. Four NPN BUT92’s are used as power devices, driven by 2x TIP35c transistors. Being an automotive amplifier, B+ in this case is essentially filtered battery voltage. A separate winding delivers negative feedback to the preamp.

In this specimen, the two 25A fuses were blown. The first thing you look for in solid-state amps like this is shorted power devices (usually FETs). Even though the BUT92s are BJT’s, three of the four were shorted anyway. They were nice enough to take out the fuses before burning up the 0.025-ohm balancing resistors or the transformer windings. Score one for Federal-Signal.

BUT92 isn’t a common part anymore, but a quick trip to eBay produced a set of four for short money. While the possibility of counterfeit devices on eBay is always there, in this very-not-audiophile application I’m guessing the specs aren’t too critical.

BTW, if you find yourself into one of these, they need to be biased. Federal-Signal kindly provides the procedure, but basically you measure the bias current across the fuse terminals (fuse removed, obviously) and adjust the handy multi-turn pot for 70mA per side.

photo credit: Federal-Signal

Another note: to save wear and tear, the B+ is kept off the power stage by a relay unless the knob on the remote head unit is switched to one of the several positions that make noise. If you find yourself with a dead amp, be sure to turn it on.

Class1 “EVEN Series” Load Manager Flickering Problem

I’ll be the first to admit that the number of people on this planet who have this problem is pretty small. And even smaller is the number of those Googling this problem and finding this page. But if I can save even one fire department from the hell of this nasty little issue, it’s worth it. And if you’re a troubleshooter, you could learn something here too.

Background

This is the story of an electrical gremlin that flummoxed our local fire department for — I kid you not — 25 years. For that entire time, the emergency lights on their first-out engine had been randomly cutting out. And they could see and hear the relays on this board rapidly opening and closing (“flickering”) while this happened.

The department sent the truck to the dealer countless times, and each time the dealer piled ineffective fix on top of ineffective fix. They rewired the switches. They added auxiliary relays. They replaced the entire board. They added extra grounds. And none of it made a damn bit of difference because they didn’t have the skill and/or patience to find the root cause.

At the department, no one knew what the problem was so mythology started to take over. “It’s got too much load on it.” “We need to run bigger power wires.” “It’s worse when it rains.” “We should switch to LED lights.” “Replacing the switch helped.” Finding the problem turned out to be as much about dealing with theories as it was doing the troubleshooting.

The Method

Since this was a known tricky problem that had been looked at by several at-least-somewhat-competent people, I didn’t spend a lot of time on the easy stuff. Instead I focused on listing very clearly all of the hard evidence we had.

Mostly this was in the form of several cell-phone videos taken of the board while it was misbehaving. The indicator lights could be seen to flicker in seemingly random patterns, and relays could be heard rapidly cycling at several Hz. I studied these videos carefully, and wrote down exactly what I could see in each.

The second thing I did was research the board itself to understand how it actually works. This helped me dismiss several myths. For example, the board actually has no way to detect current flow. So being “overloaded” couldn’t be the cause. What it does detect is the battery voltage — if that voltage sags, it turns up the idle (to increase alternator output), and if that doesn’t work, it starts turning off loads (and an LED comes on saying so). But in the videos, that LED never came on. Further, the board displays the battery voltage on a display — and that voltage never fell during the videos. So the load manager wasn’t shedding loads.

Although I rarely do this, as part of my research I called Tech Support at Class1 (now owned by IDEX). As expected, they had no clue — they couldn’t even find a manual. But I read their tech support rep every part number I found on the board, and to his credit he looked them up. One of them came up in his database as the circuit board itself, and he actually sent me the schematic! (I’d call him out by name for praise here, but they’d probably discipline him.) This saved me a ton of time.

The Insights

Although the flickering in the videos was mostly random, it wasn’t completely so. Specifically, some channels that were ‘on’ flickered ‘off’, but channels that were ‘off’ never flickered ‘on’. Also some channels stayed on and never flickered — and these were always the same channels in every video.

The schematic gave me more insights:

  • Several relays (#3, #4, #7, #10) have no electrical connection to the LM logic board.  They are controlled only by their cab switches.  The LM does not know if they’re on, and it cannot turn them off.  
  • Some relays (#1, #2, and #5) turn on with the MASTER switch, but have no individual switches.  (I believe that #1 and #2 were intended for ‘white’ (aka ‘scene’) and ‘amber’ (aka ‘respond’) lights.  #5 was intended to be used for A/C and is always on unless it needs to be shed.) 
  • The remaining relays (#6, #8, #9, #11, #12, #13, #14) come on with the Master, but also have individual switch inputs, meaning that they also need to be turned on with their individual cab switches.

It was this last group that was flickering. And that group uses different circuitry than the rest of the channels. Specifically these channels have 74HC125 tri-state buffers (U40 and U41) which “mix” the signals from the load manager logic and the switches on the driver’s control panel. Further, the panel switch inputs are active low (i.e. grounded to trigger), and pass through 1N4008 diodes which serve as protection against applied voltage.

The specified input voltages for the 74HC125 are 0-1.4v (low)  or 3.2-5v (high).  Any voltage between 1.4 and 3.2v results in ‘undefined’ output behavior.  This can include being “on” or “off” randomly, or rapidly oscillating between on and off.

To be safely “low”, the voltage has to be below 1.4v. But recall that the pulldown current passes through a 1N4008, so there’s a .7v diode drop.

Finally, this board was made as a drop-in replacement to fit on a relay board made by another manufacturer (Federal-Signal). That board has its own diode protection on the inputs, adding further diode drops and effectively eliminating the already small .7v margin.

The Root Cause

So the root cause of the problem was that the input channels on the board, combined with the relay board it’s attached to, created a system that’s extremely sensitive to the quality of the switch wiring, to the point that even reasonable imperfections in switch contacts or grounding would put the logic devices into their unstable regions. In other words, there is no right way to install this thing and have it work reliably, unless you have absolutely perfect grounding of the switched-on inputs.

Now What?

There’s a lot more to this story, but Mr. Determined Enthusiast is a busy guy and this is already pretty long. If you’re struggling with this problem, hit me up in the comments and I’ll be happy to help.

Sentinel 300P Battery Charger Repair

One day the backup generator at our local fire station was throwing alarms about a dead battery. The battery charger was this beautiful thing in a stainless box with CANBUS control and alarm outputs and…some error message about no AC power. The charger would turn on with a battery connected, but not with only AC power connected.

Turns out this beautiful thing costs about $1000, and has essentially zero parts or service support from the company.

As expected, the charger is basically a switch-mode power supply with a microcontroller directing current flow. Generally speaking with devices like this, the things that fail are a) big power-handling devices, and b) anything close enough to AC power to get fried.

With no schematic, the first thing I tried was to look for fried components, but no joy in this case.

When tracing schematics, a trick I like to use is to scan the solder side of the board (for single-sided) with a flatbed scanner, then flip the image in software so I’m looking down from above. Then I can draw the components on in their “correct” orientations. After that I can trace out the schematic.

I didn’t save any of those pictures, but I did save the pictures of my hand-drawn schematics. The board has tens if not hundreds of components on it, but most of them are after the isolation transformer — and roughly speaking that’s the side that works. So I focused on the high-voltage side.

As you know if you’ve gotten this far, a SMPS basically rectifies AC to a high-voltage DC rail, and then switches that at high frequencies through a transformer. In this case, the 150v rail was okay, but we weren’t getting any switching happening. So I started there.

In this first schematic Q3 and Q4 are big MOSFETS, and they’re acting as low-side switches grounding the transformer primary. U12 is a MOSFET driver connected to the gate of both. A quick lookup of the datasheet gave us the pinout, and so we can see where the power and drive should be coming in. With the drive being shorted to ground we already have a problem.

This next schematic shows the PWM controller chip that’s supposed to drive the FET driver. There’s a whole science to making switch-mode supplies run efficiently, but you don’t need to understand that to see what’s going on here. There’s some feedback from the transformer coil, there’s an optocoupler feeding signal from the logic (probably regulating the output), power, ground, and a bunch of passives. The chip gets power directly from the 150v rail, but also from “D31”.

Here’s the power supply for the chips above. There’s sort of a belt-and-suspenders approach here, with Q15 regulating power from the rail, but Q14 also providing power via a rectified winding from T1. These get “mixed” by D35 and D17. Give the high numbers on the left side, I could imagine their original jump start didn’t work, and they added the left side later. Anyway, if I ever figured out exactly why this was needed, I’ve forgotten, but note that the output voltages from Q15 are too low to drive U12.

In general I’m not one to throw parts at a problem, but we can make some educated guesses here. The FET driver is definitely toast, the PWM chip might or might not be, and Q15 is definitely suspect. The passives all look good. So the smart thing to do is replace every active in the area, including the big FETs because they see a fair amount of stress.

Total parts cost was about $40, including the big FETs. I did this SMD work with a regular iron, just by flowing solder over all of the legs and then heating them all at once. Hot air would’ve worked too.

After replacing these parts, the charger fired right up and is still working today.

So why did they die? Our area gets a lot of power disruptions, so a surge seems like the most likely culprit. To be honest, I was pretty impressed with the amount of filtering I found on the incoming AC lines (as it should be at this price), but apparently it wasn’t enough.

ISG X380 Thermal Imaging Camera (TIC) Charger Repair

Firefighters use the TIC (Thermal Imaging Camera) for a variety of tasks, from rescue to finding fire in walls. When our local fire department’s TIC stopped charging, the issue was rapidly found to be a broken spring contact on the charging cradle.

Charging contact assembly removed; the mangled pin is the middle one.

To their credit, the designers of the cradle mounted the contacts in a removable part and supplied a quick-disconnect.

What they didn’t do is prevent the company from being sold to Scott (the SCBA people), who’s website screams “go away” with its lack of information about these units and the inevitable “contact your Scott dealer”, who they know damn well have zero interest in finding you a small part from a charging cradle for a TIC that was discontinued during the first Clinton administration. Nine out of ten fire chiefs will just buy a new cradle, or an entire new TIC, which suits Scott just fine.

To which the Determined Enthusiast says: no thanks.

I was worried the spring pins would be overmolded or heat-shrunk in, but they’re not. They’re a light press fit. My normal MO here would be to measure up the spring pins and buy new ones from DigiKey. Sadly, they’re weird, in that the pin part is abnormally long.

The ones I ended up finding were from Everett Charles, part number P2532. They’re beefier than the originals (which is a good thing) and they don’t fit the original plastic piece. So I made one.

Here’s where the spring pins hit on the TIC.
Scan of the bottom of the part after inverting colors and changing brightness/contrast

A great reverse-engineering tool is the flatbed scanner. They’re great for tracing the weird profiles of parts like this one. Just put a ruler down on the bed with the part and scan away. If you have access to a CAD system like SolidWorks, you can use the Sketch Picture functionality — it has a handy feature which will calibrate the size of the picture to an actual length in CAD (which is what the ruler is for). Hint: before importing into CAD, use image editing software to change the brightness/contrast and/or invert the colors of the image so you can easily draw over it.

I’m old-school, so I CNC milled the part from some ABS I had around. But you could definitely 3D print this part too. Then press the pins in by hand and reconnect the wires and you’re good to go. Better than new.

Want to make your own and need the CAD? I got you…grab it below. Need some pins? Hit me up in the comments and I’ll see if I can hook you up.