LOLbooster: Output Stage (Part 2)
Having selected the LMD18200 as the h-bridge driver for the LOL booster, what remains in designing the output stage is to select a heatsink, and to draw up the auxiliary circuitry.
No sooner have I posted this schematic than I’ve already discovered an error: I had inadvertently swapped the RAIL_A and RAIL_B lines. (The reason for the mistake will become clearer when I post the input stage.) The schematic of REV A.1, posted below, replaces REV A; it swaps RAIL_A and RAIL_B, and updates the logic table to reflect the changes.
Also, I’ve update the bill of materials to reflect the need for heatsink mounting hardware.
The schematic is below; Once the total design is complete, with a PCB layout, I’ll make the original EAGLE files available for download. This schematic is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License. Please read that carefully.
LOLbooster Output Stage REV A.1
And here is the bill of materials. I give Newark part numbers, not because of any relationship I have with them, but because their catalog appears to be the most complete, and their prices the lowest in general. ($5 flat-rate shipping helps, too.) That said, Newark as a whopping 152-day lead time on the LMD18200: You’re better off sourcing it elsewhere.
| Label | Value | Description | Newark Part No. | Newark Price | Notes |
|---|---|---|---|---|---|
| Total: | $19.305 | ||||
| R1 | 4K22 | 1% or better | 94C5096 | $0.043 | |
| R2 | 100K | pull-up | 38K0329 | $0.012 | |
| C1 | 10nF | 5%, 35V, ceramic | 97M4044 | $0.160 | |
| C2 | 10nF | 5%, 35V, ceramic | 97M4044 | $0.160 | |
| C3 | 1uF | ceramic | 97M4165 | $0.261 | |
| C4 | 330uF | 35V, aluminum electrolytic | 70K9694 | $0.192 | |
| C5 | 100nF | ceramic | 98K1025 | $0.057 | |
| IC1 | LM18200T | h-bridge | 41K2745 | $14.760 | 152-day lead time (!!) |
| X1 | screw terminals | 14N5685 | $0.270 | ||
| H/S | 6399BG | 5.4ºC/W or better; rated to >=12W | 99K0085 | $2.390 | |
| H/S mounting hardware | $.500 | local hardware store | |||
| thermal compound | $.500 | local computer repair shop |
Now, for the commentary on the schematic.
For the heatsink, I have selected the Aavid Thermalloy 6399BG. This model has sufficiently low thermal resistance and mass to keep the LMD18200 quite cool even at peak currents and in still air. Users wishing to use forced-convection cooling will have the option of figuring out for themselves how to mount a fan; in most cases, it shouldn’t be necessary to use a fan at all.
The LMD18200 requires a minimum of external circuitry—but it does require some. C1 and C2 are precision ceramic capacitors necessary for the internal charge-pump: They ensure the h-bridge can switch quickly.
The LMD18200 redirect directs a tiny proportion of the current running to the rails to the current sense output. R1 converts this current to a voltage such that approximately 3.1A to the rails produces 5V on the current sense line; this will make current monitoring very easy for the (forthcoming) fault-detection circuitry. C5 is a filter capacitor for the current sense line.
The thermal warning line is open collector: It must be pulled up to logic level high via R2. Any large resistor will do fine there.
Finally, C3 and C4 are the power line filter capacitors as specified by the data sheet. These need not be very precise.
The brake, direction, and PWM lines on the LMD18200 are being used in a manner someone inconsistent with the labeling, and so deserve a little attention. Basically, the h-bridge will be operated in one of three modes: Either rail A will be held high while rail B is held low; rail A will be held low while rail B is held high, or all power to the track will be cut off. To achieve this, the incoming DCC signal will be processed by the (forthcoming) input stage into a single square wave signal, and fed, not to the PWM line, but the direction line, to take advantage of “Locked Anti-Phase” control (per the spec sheet). In these modes, PWM must be high, and brake low. The third mode is for when a fault has been detected: In case of an overcurrent, or over temperature, or loss of DCC signal, we want to shut everything down. In this case, PWM must be brought low, and brake must be high. So, I’ve just labeled PWM as “enable” and brake as “not enable”, and I’ll be sure that the fault detection circuitry always sends opposite signals to each.
So, that was easy. Next time, we’ll discuss the input stage. As always, comments, criticism, and advice are always welcome.
Don, two questions:
1. is output voltage adjustment a feature of the pre-amp, or achieved by adjusting the input voltage (on pin 6)? I’m concerned with both the 12/14/18 selection for N/HO/O and the fine-tuning adjustment to synchronize multiple boosters, as well as wondering if the supply needs to be a regulated one, or you’re going to have a separate regulator circuit for the supply so that cheap ones can be used.
2. Is this going to work for both “common rail” (all boosters have one of the two rails tied together with no ground connection) and “booster common” (all boosters share a common “ground” line, which isn’t necessarily the same as any safety ground; it can be a floating ground) DCC wiring methods? While booster common is highly recommended, lots of older layouts are wired for common rail, and the standards do support that approach.
Ken
Ken,
Thanks again for the questions!
1. The output voltage adjustment will be a feature of the LDO regulator. I have selected an adjustable regulator, and so choosing between N and HO voltages will be a matter of setting a jumper (cheaper than a switch) Incidentally, I’m not planning to support O: The heat calculations were performed for worst-case HO performance; and I’m not sure 3A is really big enough for O. My focus is on N and HO. But I’d love to hear reasons why I should consider adding a third jumper for O
Anyway, since there will be an on-board regulator, you’ll be able to use cheaper unregulated supplies. (Yay! I just saw that a regulated 15V 4A supply can cost up to $40!)
There will also be a trimpot for making small voltage adjustments: At the “N” setting, you will be able to adjust between 12 and 14VDC; at “HO” from between 14 and 16VDC. Such adjustments will probably require a DCC voltmeter, though.
2. This setup will probably not work for “common rail” wiring. H-bridge drivers are a way to get + and – 12V (or whatever) across the rails from a single-rail power supply. They do this by alternating which rail is hot and which is ground. As I understand it, “common rail” wiring requires that one rail only be the hot rail (and alternated between +12V and -12V (or whatever)), and the other (the so-called common rail) be held at ground. This requires a completely different output stage design based around an class-AB push-pull amplifier (vs. an h-bridge), which necessitates a two-rail (read: AC) power supply. I’m trying to avoid two-rail power supplies for reasons of cost (extra components to rectify and filter the current) and reliability (having a strong AC signal on the PCB is a good way to cause unpredictable problems via non-obvious routes).
But I could be wrong. I’ll read up on “common rail” wiring tonight and see if perhaps there is a way. But I doubt it.
Actually, since DCC is a square wave, you can do a pretty close voltage measurement using a DC meter to measure each rail to the booster ground and adding them (assuming you measure close to the booster where the signal is clean; the further away you get, the more ringing there is and the less accurate this would be). This is a method Digitrax suggests. You should also have address 0 selected and the throttle at zero to ensure an even wave form (no zero stretching).
I haven’t posted anything on this yet, but I’ve been doing some measurements on my Command Station and track (checking loss and similar), and saw 12.3 on the RRampMeter and 12.56 peak-to-peak using the DC meter (unloaded). Loaded with an 18W lightbulb (which drew 0.77A) it was less accurate, with 11.7 on the meter, and 12.77 peak-to-peak, suggesting some ringing had set in.
But I’d certainly recommend an RRampMeter (or equivalent DCC meter) to anyone doing much with DCC. It’s a very handy tool to have.