Gen 4: Power Supply & Stepper Controller Noise and Current Traces
Brent Crosby
http://store.makerbot.com/makerbot-generation-4-electronics-kit.html
). At that time we had heard enough grumbling about the stock Cupcake
supply, so we looked around the IT hardware room and found this
supply:
Thermaltake Purepower 500w P/N: W0100RU
http://www.thermaltakeusa.com/Product.aspx?S=1170&ID=1524
After installing it, I came across Ed's article on the Thing-O-Matic:
Stepper Driver Logic Supply ( http://softsolder.com/2011/01/08/thing-o-matic-stepper-driver-logic-supply/
) (through this thread http://groups.google.com/group/makerbot/browse_thread/thread/3251111d4741c77b
).
Since we have a scope and a 50MHz current probe, I thought I would
take some data on the power supply during stepping.
Here is a link to a PDF full of annotated scope screen shots:
http://sandbox.cfontz.net/~Brent/MiseryBot/Makerbot_Cupcake_1024_Gen_4_Supply_Noise.pdf
I took data on two stepper controller v3.3 (Gen 4) boards, each in
turn hooked to our Cupcake # 1024 Y-axis.
Other stepper boards were disabled. Extruder was connected on another
power supply lead, heaters and motor disabled.
One stepper board is stock. On the other stepper board we added 22uF
MLCC caps that we had on hand across the 5v (unused) and 12v pins of
the power connector. (see the thumbnail in the PDF)
The first page is just a shot of the power supply rating plate.
The second page is the current (in green) and the voltage (in red) of
the 12v feed into a stock stepper controller. Notable points:
1) No noticeable low frequency droop on the 12v (good)
2) Peak current is only about 300ma, I figured it would be higher.
3) Current on the 12v to the stepper goes negative every cycle,
dumping some energy back to the supply.
Page 3 is the same measurement, but taken on the 22uF MLCC modified
stepper controller board. Notable:
1) The red 12v looks a little smoother (less noise, good)
2) The current is a little noisier. I am not sure what to think of
this. I am wondering if other loads in the system are being supplied
some high frequency current back from the MLCC (which have awesome
high frequency response. ref: http://www.digikey.pt/PTM/PTMPartList.page?site=pt&lang=en&ptm=8782
).
Page 3 is a zoom in on the 12v noise w/o the MLCC. Page 4 is the same
measurement with the MLCC. Noise is decreased by ~5x.
In Ed's article he saw noise on the 5v after the 7805 regulator. Page
6 shows the output of the regulator w/o the MLCC and Page 7 shows it
with the MLCC. The noise is reduced by about 2x. However, even without
the MLCC, the noise I am seeing is only 160 mV p-p, which is not too
horrible.
Page 8 is a slow-sweep of doing a step session, Notable:
1) Current when the stepper is disabled is near 0 as you would expect.
2) Current during stepping swings positive and negative, as we saw
earlier.
3) This surprised me: Current when the motor was stopped but the
driver was enabled was higher than when it was moving. It turns out
that on multiple runs, this level will vary, I assume it depends on
what phase of the microstep the motor was stopped at.
4) There is no significant droop on the 12v, only like 100mV which is
negligible in this application.
I also experimented with leaving the 22uF MLCCs on a stepper
controller and removing the electrolytic caps. This increased the
noise, so there is a synergy between the electrolytic's bulk storage
ability combined with the great high-frequency response of the MLCC.
The layout of the stepper controller could be better. I would have
concentrated the traces as much as possible on the top layer. The
bottom layer would be a ground flood, interrupted only when a trace
from the top needs to dive to the bottom layer for a bit to cross
another trace. All ground points would then via to the bottom layer
ground flood as close to the pad as possible. These layout
improvements might reduce the noise some more.
My plan is to try and take some time to take similar data on the
extruder controller board tomorrow. Even though I am curious how much
this data would change with the original supply, I do not think I am
going to go back to it to find out.
Brent Crosby
reasonable attempt at a good of a switching power circuit layout
implemented on a low-cost 2-layer PCB:
http://sandbox.cfontz.net/~Brent/MiseryBot/Example_2_Layer_Power_Layout.pdf
Note the bottom layer flood used as a ground plane. Seriously thick
power traces on the top. Lots of small floods around the high-speed-
switching FET, inductor, and the associated caps. Lots of vias to tie
the high speed switching circuits back to the ground flood. Top
decouple caps close to chip power supply pins. All ground connections
have a very short trace to the via that brings them to the ground
flood.
Ed Nisley
> take similar data on the extruder controller board
The combined 11 A from both heaters pulls the stock MBI ATX supply down
by about half a volt to 11.7 V, as measured on a non-current-carrying
+12 V line, which suggests a certain optimism in the 22 A rating.
Check the actual DC voltage at the heater terminals along that long
trace from the +12 V input: by the time the juice gets to the H-bridge
driver, there may not be enough left of it to keep the motor happy.
Excellent work on the stepper driver!
Brent Crosby
board when heating the build platform and extruder heater (no motor
operation):
http://sandbox.cfontz.net/~Brent/MiseryBot/Makerbot_Extruder_Current.pdf
This is just for the Cupcake HBP and MK5 extruder heater (no extruder
motor), powered by the Thermaltake supply mentioned above.
Brent Crosby
to the extruder board during HBP and Extruder heater operation:
http://sandbox.cfontz.net/~Brent/MiseryBot/Makerbot_Extruder_Current.pdf
This is for a stock extruder fed off the standard 4-pin Molex HDD
style connector.
Craig Link
thru a single molex connector as well as the small trace on the Gen4
Extruder board. Initially I was using the MakerBot Relay Board 1.0
with my gen 4 electronics, but I got tired of all the clicking.
I went ahead and created the circuit here - http://www.flickr.com/photos/craig_flickr/5393405169/
to replace the relay board which also offloads 11amps from my extruder
board. I've just modeled the circuit in LTSpice and then built it on
a prototyping PCB with parts that were in stock at the local Fry's.
Since I don't have a automated build platform, I only have two copies
of the circuit on the board.
R10 and R14 just drop the 12v input from the extruder board to 5v.
For the Q1, R6, R8 and Q2, R2, R13 portions of the circuit I used a
single component for each - NTE2359 (http://www.datasheetdir.com/
NTE2359+download) which has the two 47k resistors built into the NPN
transistor.
For the mosfets, m2 & m3, I used NTE2987 (http://www.datasheetdir.com/
NTE2987+download) which can handle up to 20A.
I also put 2 schottky diodes in the circuit as snubber circuits, D1 &
D5, to help with the transient voltage spikes when the heaters where
turned off. I used NTE573 (http://www.nteinc.com/specs/500to599/pdf/
nte573.pdf) which handle 5A and a surge of 150A
I'm powering the from 4 pins ( 2x +12v & 2x GND ) of the ATX EPS12V
power connector.
I've probably done 8 hours of printing with this circuit, my Gen4
electronics and MK5 extruder and working flawlessly.
Note, awhile ago, I also had upgraded the stock 350W Cupcake power
supply to a Corsair TX650 -
http://www.amazon.com/gp/product/B000X24ISU?ie=UTF8&tag=mak545-20&linkCode=as2&camp=1789&creative=390957&creativeASIN=B000X24ISU
. This one provides +5v @ 30A and +12V at 52A off single rails
-Craig
Ed Nisley
> powered by the Thermaltake supply mentioned above.
Now, *that's* interesting: the voltage drop at the board entry is pretty
nearly the same as for the MBI supply on my TOM, despite the difference
in power supply ratings. I've always thought the wires were undersized
for high currents and this looks like confirmation.
Was the scope's ground reference point at the Extruder Controller or on
an unloaded ground wire to the supply? If the former, then I'd expect
another 0.2 V drop in the Extruder Controller ground wires: two of 'em =
half the resistance.
The voltage at the screw terminal drops to 11.4 V, so the motor H-bridge
driver is seeing (at least) a 5% change. That's equivalent to PWM=242:
enough of a change to get your attention and certainly enough to affect
the extrusion.
The HBP current drops noticeably as it heats, too...
Zip Zap
From: Craig Link <cr...@moonrock.com>
To: MakerBot Operators <make...@googlegroups.com>
Sent: Thu, January 27, 2011 1:40:43 PM
Subject: [MakerBot] Re: Gen 4: Power Supply & Stepper Controller Noise and Current Traces
You received this message because you are subscribed to the Google Groups "MakerBot Operators" group.
To post to this group, send email to make...@googlegroups.com.
To unsubscribe from this group, send email to makerbot+unsub...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/makerbot?hl=en.
Brent Crosby
whatever board I am looking at. I soldered test pins on the back of
the Molex, and clipped the scope leads to those pins.
Light bulb filaments draw more current when they are cold, and I
wondered if the MK5 heater and HBP might draw more current when cold
than when they are warm.
http://sandbox.cfontz.net/~Brent/MiseryBot/MK5_and_HBP_Warm_Up_Motor.pdf
The first page is a slow trace of the current drawn by the MK5
extruder heater over the full warm-up from room temperature to 220 deg
C. The current is essentially flat, which indicates that the
resistance of the heaters is constant over the temperature range.
The second page shows a similar trace for the HBP. The HBP current
does go down over temperature indicating that the resistance increases
with higher temps. I guess that is expected with copper. That does
make me wonder what material is used in the commercial power
resistors.
The third page shows the current drawn extruder board when the
extruder motor is on or off. Nothing too exciting. The motor running
does not seem to increase the noise at the 12v connector. There are
some potentially nasty spikes when the motor is switched on and off. I
wonder if the extruder controller includes some catch diodes to shunt
these spikes to the rails?
The fourth page is a zoom of the current drawn extruder board when the
extruder motor is on. There is ~160 Hz pulsing. I am guessing that is
the motor commutation.
Regarding the power supply. I think the Thermaltake power supply may
be a true two rail design. There is a slightly different voltage on
the 6-pin PCI-X connector and the 4-pin 12v add-on connector, compared
to the drive loom and motherboard 12v connectors. So at this point, I
am thinking of using the 6-wire to power the extruder board, since it
will keep all the other 12v clean, and it should reduce the voltage
drop on the extruder board a bit.
Ed Nisley
> ground point is the ground pin right next to the power pin
So then the voltage applied to the heaters & motor is as shown.
I just added a 2.5 A dummy load to the +5 V supply in my TOM, which
dropped the voltage from 5.25 to about 4.83 V. Both are within the usual
10% tolerance, but a somewhat higher minimum load should improve the
stability. The 30 ohm resistor on the Motherboard provides a tiny 600 ma
base load
The supply voltages are surprisingly interlinked. Some measurements from
an older project:
http://softsolder.com/2010/12/30/atx-power-supply-vs-heatsinks/
I suspect some of the random glitches arise from load dumps when the
heaters switch off: the transformer flux changes faster than the power
supply control loop can handle, so the circuitry sees out-of-spec
voltages for a very short time. After all, ATX supplies were never
intended to support high-current devices that blink on and off at PWM
rates!
> that is expected with copper.
The resistance of pure copper increases +0.39%/C. The resistance change
is 0.9/4.2 = 21.4%, a bit less than you'd expect for the observed (115 -
22) = 93 C rise, which should be 36.2%. On the other paw, the HBP trace
isn't pure copper, it's not uniformly heated, and we're looking at it
through a fairly large interconnect resistance. Close enough.
Real power resistors use something like nichrome wire and have a tempco
under 100 ppm: you'll never see the variation on the scale we're working
at. [grin]
> potentially nasty spikes
The "current spikes" are measurement artifacts: the current through an
inductor is continuous. Most likely, the scope is catching a mild case
of the tremors from those all-too-real voltage spikes.
The H-bridge uses FET body diodes to direct the current around the OFF
switches, which should avoid the worst of the inductive voltage
transients. Paths 2, 4, and 6 in the bewildering graph on page 4 of the
A3949 data sheet show this happening.
http://www.allegromicro.com/en/Products/Part_Numbers/3949/
There's also the SMS12 transient suppressor on the OUTA/OUTB terminals,
but it has a guaranteed limit of 21 V at 1 A that would be off-scale
high.
> using the 6-wire to power the extruder board
Strong agreement!
Verily, it is written: one careful measurement trumps 1000 expert
opinions. Keep up the good work...