dro350 scale jitter correction
Greg
the scales on my 350 was jittering. I have already done the ground and
capacitor mods so I was a little curious about the reasoning. There
are possibly two things going on, one being that I recently found that
if a scale is mounted too close to a metal object it can provide
enough difference in the scale pickup to make the scale stop reading.
The other was something posted on Rick's website about jitter and
shielding the cables from the LED drivers. Not wanting to go the
entire shielded cable route, I grabbed a bit or scrap sheet plastic
and some aluminum tape. A couple of strips of tape on the plastic and
stick the plastic side down to the chips, and it seems I have cured
most of the jitter on that scale. The problem scale happened to be the
one that has the most length of wire exposed to the LED driver section.
Bruce Fairman
i've written my own scale fiq routine that puts a realtime stamp on each sample. this allows me to generate a scale clock input distribution map. the change noted above makes the map conform with the measured scale timing. my code also verifies the clock timing on each frame received and tallies the errors. when the above mod is in place, there are very few errors.
unfortunately, this code is in the lcd version of the dru550 code that i have been developing, so the code from shumatech will continue to do what it does (no real protocol validation). if you are interested, i can send the relevant pieces but integration into the 0.3.2 shumatech code is something i have not done.
attached is a picture of 2 examples, see the red ovals.
--Thanks,
bruce
KE6OKC
Bruce Fairman
--Thanks,
bruce
KE6OKC
Rick
That makes more sense. I looked at the ref node a lot and it was
always quiet.
Rick (via iPod)
Greg
have everything or can add it to an order. Just to be clear that is
10k on clock to ground and cap on ref.voltage to ground, correct.
Why does it keep coming up that the power supply filtering seems to be
a little on the lacking side? Seems there should have been .1 on
pretty much all devices input power and probably on all ref.voltage
lines too. I know it adds a lot of expense but seems like with all the
noise and the non-continuous rate that the LEDs are run it would be
needed.
I think with the other mods I am within the windows of averaging built
into both the 350 and whatever the new technique on the 550, Scott
said he did something new in the 550 code that makes the cheapo scales
like I'm using exhibit less flickering. There was a post about it when
he said he had the 0.3.2 firmware out.
Rick
ground and +5v plane, it would solve a lot of problems.
Rick (via iPod)
Greg
it's on the Z axis so I can probably tolerate this for a while but is
definitely something I'll work on when it comes time to pull the PIC
and solder in the DPU.
Rick
Rick (via iPod)
Greg
the filter after applying them.
I was looking at the DRO-550 board yesterday and noticed three
different size caps on the scale power line, didn't get a magnifier
out to see what the values were, but definitely a step in the correct
direction. If not already implemented I'll probably put this resistor
and cap mode on the 550 when I assemble it, need to look ver the
schematics and see what is already there.
Rick Sparber
Your shield is "electrostatic" so blocks capacitive coupling of noise. IIRC,
I used a similar trick with aluminum foil to test my theory.
Are your scales grounded to your mill or lathe? Mine are.
Rick
Rick Sparber
trampoline. If no current is flowing, the voltage between any two points
defined as ground is zero. But if current flows (or you stand on the
trampoline), then the ground points move apart. By adding a bypass cap
across each IC you will at best lock the network of ground traces to the
network of power traces. This will help but not necessarily all that much.
Power and ground at an IC will stay together but that does not mean that
they will be at the same voltage as the next IC. Any given signal out of the
first IC will be with respect to its ground. If the next IC's "ground"
voltage is different, then this difference will add or subtract from the
signal. The only solution is to add a solid ground plane just like putting
down a piece of plywood on the trampoline will hold all parts of that
surface at the same height. The plywood can bounce up and down but as long
as you stay on it, there won't be any relative motion.
I would not bother to add a ground plane to the existing 350 circuit board
but on a new board it should not be so hard. A 4 layer board could have a
power and ground plane on the inside so externally it would still be easy to
probe traces.
Rick
-----Original Message-----
From: opendro-u...@googlegroups.com
[mailto:opendro-u...@googlegroups.com] On Behalf Of Greg
Sent: Monday, May 24, 2010 1:27 PM
To: OpenDRO users' group
Greg
Knowing that some of the cheaper scales could have problems when they
were attached to the mill with metal mounts, I went ahead with plastic
mounts to isolate them from the mill. Running a wire from the scale or
the electronics to the mill would be simple enough if it is worth
doing. Not sure if my scales have the metal portion connected to
positive or negative power.
Rick Sparber
The DRO should not be tied to ground so it should be safe to tie the scales
to ground as long as they are all of the same type. Chinese scales should
have their bodies connected to the positive side of the battery. Leaving the
scales floating often does invite noise problems.
It would be best to directly bolt the scale to the metal of the mill rather
than use a wire but if that is the best you can do, give it a try.
Rick
-----Original Message-----
From: opendro-u...@googlegroups.com
[mailto:opendro-u...@googlegroups.com] On Behalf Of Greg
Sent: Monday, May 31, 2010 8:00 PM
To: OpenDRO users' group
Subject: Re: dro350 scale jitter correction
Bruce Fairman
--Thanks,
bruce
KE6OKC
Rick
I was confused by what you wrote. The clock and data out of the scale
are with respect to its ground. It doesn't matter what you do for
power, you can't avoid that ground connection.
Rick (via iPod)
Bruce Fairman
did an expirement. i took apart a scale and isolated the frame from the pcb. the scale continues to work and if i connect the floating scale frame to chassis ground or power supply ground, the scale reading continues. it looks like the scale does not need to have the pcb connected the the scale frame. interesting.
--Thanks,
bruce
KE6OKC
Rick Sparber
On my scales, clock and data are referenced to the, let's call it datum,
which is the frame of the scale. The positive side of the battery is tied
here too. Things are just the opposite at the DRO end. DRO datum is at -1.5V
with respect to the scale's datum. This is not a problem because the DRO's
datum should not be tied to anything outside of the box. This was fine for
the DRO-350 but given the USB connection of the DRO-550, there is the
potential of a short or excessive noise. Consider this:
The mill is connected to safety ground via the power plug.
The scales are bolted to the mill's steel body so are also at what I will
call ground.
The DRO datum is therefore at -1.5V with respect to ground.
A user comes along and plugs in their laptop. The ground shroud of the USB
connector will be floating if the laptop is running on batteries. It might
be tied to ground if powered from the commercial AC. It is also possible
that there could be a lot of power supply noise between ground and laptop
"datum" if there is no ground connection to this USB shroud.
If the laptop is running on batteries, there should be no problem.
If the laptop is running on commercial AC and has its USB shroud grounded,
then the user just shorted out the scale power supply.
And if the laptop is running on commercial AC but has no direct connection
from its USB shroud to ground, then there could be a lot of noise pumped
into the DRO and especially the scale inputs.
I know some people electrically isolate the scale body from ground. This
tends to cause a lot of noise pick up problems.
As for your experiment with the pcb, I suspect the circuit capacitively
couples to the bar's strip so it does make sense that it doesn't need a
direct connection to ground. It gets complicated when that pcb's datum is
connected to the DRO's datum while the scale is at ground. I can see where
this could cause a lot of noise problems.
Rick
-----Original Message-----
From: opendro-u...@googlegroups.com
[mailto:opendro-u...@googlegroups.com] On Behalf Of Bruce Fairman
Sent: Thursday, June 03, 2010 12:02 PM
To: opendro-u...@googlegroups.com
Subject: Re: dro350 scale jitter correction
Bruce Fairman
so, i think that disconnecting the scale pcb from the scale frame is a viable method. in my experiment, the noise level is not high enough to be a problem, but i will investigate this aspect. i tried a quick test by turning a fluorescent light on and off while holding the scale cable by the switch. since my code detects errors in the received data frames, i figured that my missed frames count would be interesting. it was boringly zero. but the noise caused the test rig to go a little nutty, so there amy be problems that were not apparent. may test setup is not very noise tolerant.
as i understand it, the scale works by the pcb generating a signal that altered by the passive pcb under the in the number strip. my searching has led me to think that the capacitive scale scheme is quite susceptible to noise.
--Thanks,
bruce
KE6OKC
Rick Sparber
-----------------------------------------------------------
6/4/2010 9:10 AM
Please see my comments below.
Rick
-----------------------------------------------------------
-----Original Message-----
From: opendro-u...@googlegroups.com
[mailto:opendro-u...@googlegroups.com] On Behalf Of Bruce Fairman
Sent: Thursday, June 03, 2010 2:07 PM
To: opendro-u...@googlegroups.com
Subject: Re: dro350 scale jitter correction
Rick, is the datum the scale frame or the neg connection on the pcb?
-----------------------------------------------
6/4/2010 9:10 AM
On my scales the scale frame is connected to the positive connection on the
pcb. You can see the details in my article on grounding and noise.
http://rick.sparber.org/Articles/DROnoise/dn.pdf
-----------------------------------------------
seems the scale frame is connected to either the neg or pos connection on
the pcb but the signal reference is neg on the pcb.
-----------------------------------------------
6/4/2010 9:10 AM
The clock and data are referenced to the scale datum which is the positive
lead although you could reference it to the negative lead since the two
power supply rails are connected together with bypass caps. At the DRO end
they do connect the negative battery lead to DRO datum. However, on my scale
the metal frame is connected to the positive battery lead and normally that
metal would be your ground. Bottom line: the grounding scheme at the scale
is not the same as on the DRO.
-----------------------------------------------
that is what the scale circuit expects, i think.
-----------------------------------------------
6/4/2010 9:10 AM
By scale circuit I assume you mean in the DRO. If so, then I agree. If you
look at the clock and data signals, you will see that they are essentially
symmetric so it doesn't matter.
-----------------------------------------------
the neg connect on the pcb is connected to ground.
-----------------------------------------------
6/4/2010 9:10 AM
On my scale the positive terminal connect to the metal body of the scale. Be
careful with the term "ground". I define ground as commercial AC safety
ground which is what the metal of your mill or lathe should be tied to. This
is not the same as DRO datum which is floating or scale datum which may or
may not be connected to ground. All of this can be very confusing without
pictures in front of us.
-----------------------------------------------
i think that is the datum. the scale frame connection s not relevant until
two scales with opposite frame connections are tied together electrically.
-----------------------------------------------
6/4/2010 9:10 AM
If that happens, the scale power supply would be shorted and the system
would die.
-----------------------------------------------
so, i think that disconnecting the scale pcb from the scale frame is a
viable method.
-----------------------------------------------
6/4/2010 9:10 AM
Maybe but if it was possible, why would the manufacturer spend the money to
connect it in the first place?
-----------------------------------------------
in my experiment, the noise level is not high enough to be a problem, but i
will investigate this aspect. i tried a quick test by turning a fluorescent
light on and off while holding the scale cable by the switch.
-----------------------------------------------
6/4/2010 9:10 AM
The killer will be a motor under load being turned off. You then get a very
large inductive kick that can couple into the scale cables and yank around
the datums.
-----------------------------------------------
since my code detects errors in the received data frames, i figured that my
missed frames count would be interesting. it was boringly zero. but the
noise caused the test rig to go a little nutty, so there amy be problems
that were not apparent. may test setup is not very noise tolerant.
-----------------------------------------------
6/4/2010 9:10 AM
What a "manly" test? Take a 24V relay and connect it up so the coil is in
series with a break contact. This will create a buzzer. Wave that around
your scale. If nothing bad happens, touch each of the relay wires in turn to
the power and datum wires of the scale. This is called a "buzzer test" and
usually crashes systems. The problem with it is the noise is severe but
uncontrolled so it is a one way proof. If your system survives the buzzer
test, all is solid. If it crashes, you really don't know how weak it is.
-----------------------------------------------
as i understand it, the scale works by the pcb generating a signal that
altered by the passive pcb under the in the number strip. my searching has
led me to think that the capacitive scale scheme is quite susceptible to
noise.
-----------------------------------------------
6/4/2010 9:10 AM
It probably is and that is why disconnecting the metal body of the scale
from pcb data seems like a bad idea.
Rick
-----------------------------------------------
Bruce Fairman
my scales are also connected to pos on the scle pcb, why this is done is not clear.
starting fron the observer point of view (the clock and data comparators), the neg terminal on the scale pcb is connected to Vss (ground from a signal perspective). this in not chassis ground since the dro is powered by a wall wart. in fact the dro is floating since the wall wart neg power supply lead is not connect to the ac power green wire.
the pos scale terminal is connected to the power supply in the dro, so it is also Vss relative.
therefore, the comparator sees the scale clock and data signals relative to Vss. the connection of the scale frame to any scale power terminal makes no difference to the above points.
in fact, isolating the scale frames from their machine is exactly the same as disconnecting the scale body from the scales' neg terminal
i got the patent for the scale technology and there is no indication that transducer has to be connected to the mechanics of the gauge.
--Thanks,
bruce
KE6OKC
Rick Sparber
-----------------------------------------------------------
6/5/2010 8:26 PM
Please see my comments below.
Rick
-----------------------------------------------------------
-----Original Message-----
From: opendro-u...@googlegroups.com
[mailto:opendro-u...@googlegroups.com] On Behalf Of Bruce Fairman
Sent: Friday, June 04, 2010 2:39 PM
To: opendro-u...@googlegroups.com
Subject: Re: dro350 scale jitter correction
Rick, this is a good discussion.
my scales are also connected to pos on the scle pcb, why this is done is not
clear.
starting fron the observer point of view (the clock and data comparators),
the neg terminal on the scale pcb is connected to Vss (ground from a signal
perspective). this in not chassis ground since the dro is powered by a wall
wart.
-----------------------------------------------
6/5/2010 8:26 PM
The DRO's pcb does have a ground node even though it isn't tied to the case.
-----------------------------------------------
in fact the dro is floating since the wall wart neg power supply lead is
not connect to the ac power green wire.
-----------------------------------------------
6/5/2010 8:26 PM
It is floating only if the scales are floating. Ground one or more scales
and the DRO's datum is set to 1.5V below ground. I really wonder what
happens when you plug a USB that is grounded into the DRO.
-----------------------------------------------
the pos scale terminal is connected to the power supply in the dro, so it is
also Vss relative.
-----------------------------------------------
6/5/2010 8:26 PM
Agreed.
-----------------------------------------------
therefore, the comparator sees the scale clock and data signals relative to
Vss.
-----------------------------------------------
6/5/2010 8:26 PM
The comparators are referenced to a voltage half way between +1.5V and DRO
datum IIRC.
-----------------------------------------------
the connection of the scale frame to any scale power terminal makes no
difference to the above points.
-----------------------------------------------
6/5/2010 8:26 PM
agreed
-----------------------------------------------
in fact, isolating the scale frames from their machine is exactly the same
as disconnecting the scale body from the scales' neg terminal i got the
patent for the scale technology and there is no indication that transducer
has to be connected to the mechanics of the gauge.
-----------------------------------------------
6/5/2010 8:26 PM
Patents don't always reflect lessons learned in production. At least on my
scales, the body is tied to the +side of the battery. There must be a reason
to do this.
-----------------------------------------------
--Thanks,
bruce
KE6OKC
Bruce Fairman
assume 2 cases:
A - the dro signal ground is isolated from ac power ground (no green wire)
B - the dro signal ground is connected to ac power ground
1 - scale frame connection to a scale pcb power terminal is problematic if
all of the scales connected to the dro do not have the same frame to pcb connection when the scales are attached to a common electrical base (a machine tool)
2 - when the scale frames are all attached to their pcbs in the same way, there is not a problem in case A
3 - when the scale frames are all attached to their pcbs in the same way, there is a potential problem in case A when the dro is connected to a computer (usb or serial) where signal ground is connected to ac power ground. this situation should be equivalent to shorting the scale power.
4 - the connection from the scale pcb to the scale frame is not necessary, maybe. when the scale frame and pcb are not connected, the scale works. the is empirical and it is in agreement with the sylvac patent. there is more to be done here.
5 - issues 3 is avoided in the YADRO implementation by opto isolating the external interface (serial) but YADO still has issue 1.
the picture attached is what we are discussing. or so i think.
so, what can be done to alleviate this situation?
for 1, do not mix and match scales with different power terminal to frame connections or get rid of the scale pcb to frame connection.
for 3, and in general, do not use a wall wart the has a ac power ground connection that is tied to the - wire. remove the scale frame to pcb connection or only use scales that connect the pcb - terminal to the scale frame.
--Thanks,
bruce
KE6OKC
Greg
the cables between MTA and chassis connector, slightly better. Just
found a total different solution that might be of interest.
Common wisdom had it that the inexpensive Chinese scales needed to be
isolated from the milling machine electrically because it would induce
a host of problems. I also previously found that you needed a certain
amount of distance from the metal of the mill to make everything happy
and even had to make a new set of mounts to handle this.
http://www.youtube.com/watch?v=J-0kbC5pp0A
In this video I'm holding the Z axis scale in my hand, notice what
happens to the jitter when I set the scale on the mill table. A couple
of the screws holding the scale to the plastic cover are slightly long
and stick through the plastic, notice what happens when I move the
scale so that these screws hit the table. Notice that all three scales
have reduced jitter when any one of them has the scale case attached
to the mill. Notice that I can repeat the reduced jitter every single
time I put the clip on the table, pretty definitive to me. I also
tried running the mill and running my vacuum swarf collection system
to make sure that the jitter didn't increase while those were running
(not in video). I also checked operation of the scales to make sure
they still work as expected when they are connected to the mill and
everything is good to the point where I may not even need the filter
in medium precision.
Rick Sparber
Sorry for the slow response. I live in Phoenix and my A/C died in my shop.
That is a big deal around here. It is still out but I'm getting closer to a
solution.
See more comments below.
Rick
-----Original Message-----
From: opendro-u...@googlegroups.com
[mailto:opendro-u...@googlegroups.com] On Behalf Of Bruce Fairman
Sent: Monday, June 07, 2010 1:56 PM
To: opendro-u...@googlegroups.com
Subject: Re: dro350 scale jitter correction
where is this discussion going? summary
assume 2 cases:
A - the dro signal ground is isolated from ac power ground (no green wire) B
- the dro signal ground is connected to ac power ground
-----------------------------------------------
6/10/2010 4:10 PM
I've never heard of anyone doing case B. In a good design, this is how it
should be done.
-----------------------------------------------
1 - scale frame connection to a scale pcb power terminal is problematic if
all of the scales connected to the dro do not have the same frame to pcb
connection when the scales are attached to a common electrical base (a
machine tool)
-----------------------------------------------
6/10/2010 4:10 PM
Sadly, I think some scales connect the + to the body and others connect the
-.
-----------------------------------------------
2 - when the scale frames are all attached to their pcbs in the same way,
there is not a problem in case A
-----------------------------------------------
6/10/2010 4:10 PM
agreed
-----------------------------------------------
3 - when the scale frames are all attached to their pcbs in the same way,
there is a potential problem in case A when the dro is connected to a
computer (usb or serial) where signal ground is connected to ac power
ground. this situation should be equivalent to shorting the scale power.
-----------------------------------------------
6/10/2010 4:10 PM
The circuit tells me this but I have not seen it happen. I only use the USB
port when programming the DRO so power is removed. This may be why VDRO
never worked for me in my shop.
-----------------------------------------------
4 - the connection from the scale pcb to the scale frame is not necessary,
maybe.
-----------------------------------------------
6/10/2010 4:10 PM
It would be a noise sensitivity issue and would be hard to test on the
bench.
-----------------------------------------------
when the scale frame and pcb are not connected, the scale works. the is
empirical and it is in agreement with the sylvac patent. there is more to be
done here.
5 - issues 3 is avoided in the YADRO implementation by opto isolating the
external interface (serial) but YADO still has issue 1.
the picture attached is what we are discussing. or so i think.
-----------------------------------------------
6/10/2010 4:10 PM
Some wall warts do use polarized plugs so the return side which is close to
ground could be capacitively coupled to the -V output. There are tests which
can tell you a lot about what is inside. They do this to reduce common mode
noise to the load. I have never seen a DC connection between these nodes.
I thought you were saying in case B that there was a direct connection from
commercial power ground to the DRO's circuit ground.
-----------------------------------------------
so, what can be done to alleviate this situation?
for 1, do not mix and match scales with different power terminal to frame
connections or get rid of the scale pcb to frame connection.
-----------------------------------------------
6/10/2010 4:10 PM
I would certainly not be happy to discover that a new scale could not be
used because of this problem. It would be far better, IMHO, to have two
options on the DRO board. One would be for scale + to ground and the other
for scale - to ground. If that is not possible (for political reasons), then
have an adapter circuit for scales that have their - lead tied to ground.
-----------------------------------------------
for 3, and in general, do not use a wall wart the has a ac power ground
connection that is tied to the - wire.
-----------------------------------------------
6/10/2010 4:10 PM
See above - I'm not sure what you mean here.
-----------------------------------------------
remove the scale frame to pcb connection or only use scales that connect
the pcb - terminal to the scale frame.
-----------------------------------------------
6/10/2010 4:10 PM
Watch out for noise problems!
-----------------------------------------------
--Thanks,
bruce
KE6OKC
Rick Sparber
-----------------------------------------------------------
6/10/2010 4:46 PM
Please see my comments below.
Thanks,
Rick
-----------------------------------------------------------
-----Original Message-----
From: opendro-u...@googlegroups.com
[mailto:opendro-u...@googlegroups.com] On Behalf Of Greg
Sent: Thursday, June 10, 2010 9:06 AM
To: OpenDRO users' group
Subject: Re: dro350 scale jitter correction
OK, so I tried the above cap and resistor, still nothing. Shortened
the cables between MTA and chassis connector, slightly better. Just
found a total different solution that might be of interest.
Common wisdom had it that the inexpensive Chinese scales needed to be
isolated from the milling machine electrically because it would induce
a host of problems.
-----------------------------------------------
6/10/2010 4:46 PM
It may be "common wisdom" but not with me. The scales should be solidly
grounded to the body of the machine. See my docs for details.
-----------------------------------------------
I also previously found that you needed a certain
amount of distance from the metal of the mill to make everything happy
and even had to make a new set of mounts to handle this.
http://www.youtube.com/watch?v=J-0kbC5pp0A
-----------------------------------------------
6/10/2010 4:46 PM
Wonderful way to get your point across!
-----------------------------------------------
In this video I'm holding the Z axis scale in my hand, notice what
happens to the jitter when I set the scale on the mill table. A couple
of the screws holding the scale to the plastic cover are slightly long
and stick through the plastic, notice what happens when I move the
scale so that these screws hit the table. Notice that all three scales
have reduced jitter when any one of them has the scale case attached
to the mill.
-----------------------------------------------
6/10/2010 4:46 PM
Makes sense to me.
-----------------------------------------------
Notice that I can repeat the reduced jitter every single
time I put the clip on the table, pretty definitive to me.
-----------------------------------------------
6/10/2010 4:46 PM
The clip is a high inductance connection and poor at high frequencies. The
fact that it helped so much tells me the noise is rather low frequency and
may even be related to clock and data.
The way I think about this is to picture the power, datum, clock, and data
all as points rigidly held together at the scale as if mechanical points.
Then put a floppy spring between each point and the corresponding point on
the line receiver in the DRO. The line receiver is trying to see clock
relative to either power or datum but is looking through those floppy
springs. Same for data. When you ground the scale's datum, you stiffen the
corresponding spring and make it easier for the line receiver to see the
information. At best, signal quality is awful.
-----------------------------------------------
I also
tried running the mill and running my vacuum swarf collection system
to make sure that the jitter didn't increase while those were running
(not in video). I also checked operation of the scales to make sure
they still work as expected when they are connected to the mill and
everything is good to the point where I may not even need the filter
in medium precision.
-----------------------------------------------
6/10/2010 4:46 PM
I would sure be worried if you came to a different conclusion.
Rick
-----------------------------------------------
Greg
I tired one other thing after all that testing... It was to measure
the voltage difference between scale case and machine table. I'm not
sure that my reading was correct but it read 45 volts AC. Checking the
mill to power ground showed zero volts which is correct since that
earth ground is connected directly to the mill with a screw. So I put
a 0.1uf cap between the scale and the mill and got the same results as
I did with the clip lead. Those caps are only rated to 35 volts and I
let it run for several hours without any of them popping so it should
be fine. Put a cap on each scale so there are three points to spread
the coverage. If I put the DRO into low precision, I get results that
I can tolerate without needing the filter. The filter has been
modified in 0.3.2 to have some smoothing applied which is affecting
very slow movement like when you are sneaking up on a position, so
without the filter will be great until this smoothing is modified so
that the user can select between low, mid, and high smoothing values.
When I decided to test the limits of this filter I found something
very surprising, if I twisted the handwheel back and forth really
fast, I could get it to where there was no reading change on the DRO,
in jitter mode it was a huge amount of what it thought was jitter so I
think there is some room for improvement in the math and limits of
this new jitter correction.
Rick Sparber
Try this: disconnect the wall wart from the DRO but leave it plugged into
the wall. Use your meter to measure the voltage between each DC output lead
and ground at the power plug. I suspect you will read the same high AC
voltage. This is the common mode noise I told you about. I look forward
learning what you find.
Rick
-----Original Message-----
From: opendro-u...@googlegroups.com
[mailto:opendro-u...@googlegroups.com] On Behalf Of Greg
Sent: Thursday, June 10, 2010 8:23 PM
To: OpenDRO users' group
Subject: Re: dro350 scale jitter correction
Bruce Fairman
you are using a dru550, right? it might be interesting to see how the daughter card scale interfaces behave. they have a different circuit than the original 350 board.
--Thanks,
bruce
KE6OKC
Bruce Fairman
my comments are interspersed below.
--Thanks,
bruce
KE6OKC
On 2010 Jun, 10, at 16:45, Rick Sparber wrote:
> Bruce,
>
> Sorry for the slow response. I live in Phoenix and my A/C died in my shop.
> That is a big deal around here. It is still out but I'm getting closer to a
> solution.
>
> See more comments below.
>
> Rick
>
> -----Original Message-----
> From: opendro-u...@googlegroups.com
> [mailto:opendro-u...@googlegroups.com] On Behalf Of Bruce Fairman
> Sent: Monday, June 07, 2010 1:56 PM
> To: opendro-u...@googlegroups.com
> Subject: Re: dro350 scale jitter correction
>
> where is this discussion going? summary
>
> assume 2 cases:
> A - the dro signal ground is isolated from ac power ground (no green wire) B
> - the dro signal ground is connected to ac power ground
> -----------------------------------------------
> 6/10/2010 4:10 PM
> I've never heard of anyone doing case B. In a good design, this is how it
> should be done.
not intentionally but if the wall wart is a switcher, it can happen. at one time, i had the mindin8 shell and cable shield connected to frame ground, independent of the signal ground. not sure if it made a difference.
i think this will happen with a desktop machine connected to the ac mains but not laptops.
> -----------------------------------------------
>
> 4 - the connection from the scale pcb to the scale frame is not necessary,
> maybe.
> -----------------------------------------------
> 6/10/2010 4:10 PM
> It would be a noise sensitivity issue and would be hard to test on the
> bench.
well, if the scale frame attached to the machine, is connected to the + scale power, seems like that would introduce electrical noise from the machine (and other ac circuits) into the dro power supply and maybe couple to other parts.
> -----------------------------------------------
> when the scale frame and pcb are not connected, the scale works. the is
> empirical and it is in agreement with the sylvac patent. there is more to be
> done here.
>
> 5 - issues 3 is avoided in the YADRO implementation by opto isolating the
> external interface (serial) but YADO still has issue 1.
>
> the picture attached is what we are discussing. or so i think.
> -----------------------------------------------
> 6/10/2010 4:10 PM
> Some wall warts do use polarized plugs so the return side which is close to
> ground could be capacitively coupled to the -V output. There are tests which
> can tell you a lot about what is inside. They do this to reduce common mode
> noise to the load. I have never seen a DC connection between these nodes.
>
> I thought you were saying in case B that there was a direct connection from
> commercial power ground to the DRO's circuit ground.
i have seen switcher wall warts the have a resistor between - and green wire.
> -----------------------------------------------
>
> so, what can be done to alleviate this situation?
> for 1, do not mix and match scales with different power terminal to frame
> connections or get rid of the scale pcb to frame connection.
> -----------------------------------------------
> 6/10/2010 4:10 PM
> I would certainly not be happy to discover that a new scale could not be
> used because of this problem. It would be far better, IMHO, to have two
> options on the DRO board. One would be for scale + to ground and the other
> for scale - to ground. If that is not possible (for political reasons), then
> have an adapter circuit for scales that have their - lead tied to ground.
is this possible given the dro power supply circuitry? i think the only way to avoid the problem is to have independent supplies for each scale (batteries).
> -----------------------------------------------
> for 3, and in general, do not use a wall wart the has a ac power ground
> connection that is tied to the - wire.
> -----------------------------------------------
> 6/10/2010 4:10 PM
> See above - I'm not sure what you mean here.
> -----------------------------------------------
> remove the scale frame to pcb connection or only use scales that connect
> the pcb - terminal to the scale frame.
> -----------------------------------------------
> 6/10/2010 4:10 PM
> Watch out for noise problems!
i am going to try some tests with the floating pcb. seems like it can't be much worse that having the + scale power connected to the machine frame which may be a noise generator.
> -----------------------------------------------
>
part of the reason that the scale reading jitter is that the Shumate dro code does not check any protocol timing so the noise appears to be clock transitions and the data value assembled is bogus. all the current code needs are clock edges to assemble a data value, without regard to the validity of the clock timing. and the averaging filter is only useful for a smooth set of data values, so when there are spikes in data values, the averaging is pointless. actually, a limit rate of change filter would help here, discarding the wild values.
> --Thanks,
> bruce
> KE6OKC
>
Rick
Please see my comments below with the 5:05 time stamp.
Rick (via iPod)
On Jun 10, 2010, at 9:29 PM, Bruce Fairman <b...@phlogiston.com> wrote:
> Greg, maybe try the scales with batteries in them and no connection
> to 1.5v from the dro. this should allow the scales to be connected
> to the mill and not have the dro power be a factor.
---------------
5:05
This would solve the problem of mixed scale datums. Not so sure it
improves signal quality.
-------------
> you are using a dru550, right? it might be interesting to see how
> the daughter card scale interfaces behave. they have a different
> circuit than the original 350 board.
---------------
5:05
That would be a good test.
Rick
-------------
Rick
5:05
Bruce, please see time stamped comments below. Rick
-------------
Rick (via iPod)
5:05
It would depend on what else was connected to dro datum.
-------------
5:05
I completely agree about the tower being grounded. Just not sure if
there are some exceptions out there for laptops.
-------------
>> -----------------------------------------------
>>
>> 4 - the connection from the scale pcb to the scale frame is not
>> necessary,
>> maybe.
>> -----------------------------------------------
>> 6/10/2010 4:10 PM
>> It would be a noise sensitivity issue and would be hard to test on
>> the
>> bench.
> well, if the scale frame attached to the machine, is connected to
> the + scale power, seems like that would introduce electrical noise
> from the machine (and other ac circuits) into the dro power supply
> and maybe couple to other parts.
>
---------------
5:05
All connections can introduce noise relative to some reference point.
The killer with this system is that we have more than one reference
point. Going back to the mechanical model, say we are both standing on
the floor. You wave your arms and I look at them. Since we are both
solidly on the floor, I see all arm motion exactly right and make no
mistakes. But what if you stood on a bed and jumped up and down while
waving your arms. If I just looked at your arms, I would see a complex
combination of arm motions and whole body motions. This is because we
are not both standing on the same floor. Circuits are the same way.
With the scale bouncing around relative to the line receiver's
reference, the line receiver deals with the sum of a valid signal and
the bounce or noise. The solution is simple in theory - pick one
ground and have both scales and line receiver use it. This is easy at
the machine end - just securely bolt all scales to the body of the
machine. This mass of metal is our reference point. You should see
extremely little voltage between scale bodies under all conditions.
Say the mill's motor generated a large current spike that arced over
into the mill's head. This would have zero effect on the voltage
BETWEEN ground connections on the scales. Going back to the mechanical
model, if we are both standing in an airplane aisle and you wave your
arms, I clearly see that motion even though the plane is moving in the
air relative to the ground.
Now, once we establish the scale bodies as ground, the next step is to
lock the DRO's datum to this same ground. This would be tricky given
the present design but not impossible. I think I would bring the wall
wart power in at the mill, connect the minus lead directly to the
mill's steel, and then cable over to the DRO. In this way the noise
from the wall wart is dumped into the system's single reference point
where it can do no harm.
The present system is a multipoint grounding scheme with no control
over the voltage between points. It is just begging for noise problems.
-------------
5:05
I don't have the schematic in front of me but IIRC, the existing scale
power supply would be fine for scales that have their minus power lead
tied to their body. We would need to add a second supply that puts out
-1.5V relative to DRO datum and then level shifts the clock and data
up to the normal voltage range. It would need another power supply but
we are talking about very little current here so it could be a simple
charge pump. The signal level shifter could be essentially a stack of
diodes properly biased such that signals in both directions are passed.
A battery could be used instead of the power supply. It is more of a
personal preference.
-------------
---------------
5:05
My experience with jitter filters is that the more you use, the more
confused you get with the output. The best approach is to remove the
jitter and not use any filter. It gets back to the fundamental problem
that the filter can't tell in zero time what is good data and what is
noise. If you move the z axis quickly it might look like a single bad
output so the filter tries to smooth it with previous values. This
just makes the resulting display sluggish.
Greg
negative and earth ground. Instead of bringing that negative to earth
through the mill frame, I think I may just buy a better quality of
power supply, one that uses a 3 connection AC cord and has the
negative tied to earth. I don't mind spending the extra $50, might
even buy an open frame power supply and mount it in a box so that I
know I can modify it if needed.
As far as batteries in the scale, can't do that since I made custom
covers to help with the mounting, this removed the battery tabs and
set some plastic straight down on top of the circuit board.
Rick Sparber
Amazing! Once in a while I guess right.
I would not buy a new power supply. There is nothing wrong with the one you
have. You just need to provide a path from the output of the wall wart back
to ground. Grounding the scales does that just fine.
I once spent a lot of time chasing this same problem for a very high
performance circuit. In the end I bought a power supply used for medical
applications with extremely high isolation between input and output. Yet it
is a simple law of physics that to transfer power across a transformer, you
will have this noise path. You can model it as a non-ideal current source.
If no current flows, the voltage will rise to the values you have seen. I
should mention that it really is not 46 to 48V AC. That is what your meter
displays but it assumes it is looking at a sine wave. It is far more likely
a series of spikes.
Just for grins, connect that meter to the wall wart again and read the
voltage. Then put a 1 meg across the meter and note the voltage. Repeat with
a 1K. I think you will see a dramatic drop in voltage. If all is linear,
expect the voltage to drop by a factor of 1000 between the 1 meg and 1K.
Rick
Bruce Fairman
i tapped the holes that normally have screws going into the plastic cover, and i made a new cover from fiberglass filled epoxy. the lcd and buttons are also removed and i did not put any caps at the scale. this mod worked without problem.
now, as an experiment, i set it up so the scale pcb was no longer connected to the scale frame. the scale still works a expected.
so, i am thinking that the scale frame does not need to be connected to the pcb. this eliminates the shorting issue and allows the scale frame to be electrically connected to the machine which is connected to mains ground.
moding the scale with the buttons, lcd, and plastic housing such that the pcb is not connected to the frame is a little trickier, but it can be doe.
--Thanks,
bruce
KE6OKC
