Thread: power supplies with minimal radio interferance?

Great news Mickz, its good to see that the work you have put into it has had such a massive payoff for you!

And thanks for all the information you have put up here, it should help a lot of us in our projects.

Hi guys, thanks for the words.

The problem with the optional M4 case is that it’s too small to accommodate
any feed through caps. The ATX and Power connectors protrude through the opening.
It’s also made of tin metal which has a high RF resistance, particularly at higher frequencies.

You can leave the lid off the shielded box providing there are no wires within 4 inches
of the opening. I drilled 4 x ½” holes the first time for a test; unfortunately this creates
audible fan noise. I cut out the opening correctly and once again unless you run a lead over
the opening there should be no problem with RF.

As a final note, I will say again that the latest PCB is really well made, the supply runs way
down to a low voltage and regulation is really good. The USB settings, voltage and
temperature feedback works well. There is a programming problem with the standby timer
(Hard Off) when resuming from suspend. If you are a certain percentage into the timeout
period and you resume it will drop the 5V standby as it starts.

This can cause it to retry a start, and timeout, or just reboot with the typical “Windoze did
not shutdown correctly” blah, which is no big deal.

The first one is a problem if you’ve already started the car, as the IGN has to be turned off to
restart the PSU, another reason why I have an on/off switch in the lead going to the IGN
terminal on the M4.

I know some of you are into programming microcontrollers; this link is the one I started on
another forum. I tried to keep the description simple when describing the unique problems
encountered when removing the OEM audio-CD system from the Honda Accord.

The link shows how I over came them. Hope it's not off topic to post this.

http://www.driveaccord.net

Note: the pictures show the top off the PC case, it was fitted when installed.
It doesn’t matter now that it in the boot.

Edit:
I had already added two ferrite beads and 2 x .001 chip bypass caps on the leads coming from +12 lines on the M4 board,
removing them made no apparent difference once it was in the RF enclosure.

What can I say? FYGJAU! (so say we all).

The others (driveaccord) were also blown - good to see.
I like those typical words "....it's easy - now we know how"! (aka "easy, but tricky"?)

So when are you joining the brain drain?

A bit of info if anyone is interested:

I previously mentioned filtering the leads that run from J8 on the M4-PSU to the MB USB and On/Off sockets.

I tried the old idea of using a shielded cable and grounding the braid at the exit point and
case close to J8. This was only partly successful as there was still some residual noise from
the cable (to be expected).

I figured that the USB data lines were likely to be low impedance and not that high in
frequency spread so they should tolerate .001 feed-through caps. They do!

After running the data lines, Thump and On/Off lines through feed-through caps there is once
again no trace of PSU noise.

I guess that this kind of highlights one issue, the cost.

If you have access to the caps through a mate with spares etc, then its easily doable.

But, if you have 9 caps for the ATX and drive cables, then more for the usb and power leads, then you are looking at....14, 15, 16 caps or thereabouts?

Going by the prices on Mouser, even taking the bulk discount on the cheaper ones ($3.60 each), you are looking at $54 AUD in caps plus the extra needed for some sheeting (I haven't priced it yet).

I'm definitely not having a crack at the work you have done Mickz, its absolutely stellar.

But it sure begs the question of what testing the manufacturers have done, when the cure for their power supplies costs more than the power supply itself.

I know that its a case of supply and demand, but if you could offer a power supply that was properly filtered, in a case, for car computer use at a bit of a premium, people would still buy it?

I know that given the choice of fixing my M4 doing soldering work that I will freely admit, I'm not the neatest at, versus paying an extra 20-30 premium on buying a new power supply that has all this done, I would go the premium new model!

Yes it all add up. However the thing with PSU noise, unfortunately, is that it’s not just confined to the M4. Switch-Mode supplies have been a pain in the *!#^ for communications from day one.

The problem is really noticed when GPS is added to the Car-PC. Some people with external
antennas on GPS and FM may not really notice it, or just assume that loosing GPS satellites
occasionally is par for the course.

There are premium supplies out there specifically designed for low noise. They can be
expensive when compared to the M4. One mentioned a few pages back is around $230.
Some don’t appear to have all features specific to running a CAR-PC that some of us need.
So for around $70.00 AU extra you end up with an RF shielded PSU.
The case PCB is around $20 and 14 caps are required, so you’re looking at around $70.
The M4 in AU is around $120 at the door.

I agree that it’s not for everyone, however, if you need a low noise supply or want to mod
one you already own, then it's pretty easy to do and there’s nothing to modify or change on the PSU so any warranty is not affected.

Don't worry Chester - I don't see it as an attack on MickZ (nor IMO will he).

But it points out the shortfall of the product itself.
As Mick wrote, SMPS have been a pain since day one - manly the advent of the PC.
They had typical AC power factors of 0.5-0.6 when most world Standards require power factors of at least 0.8.
They were responsible for huge voltage distortions ad interference.
Yet who required the "cleanest" voltages? The computer industry - yet THEY were the ones creating the dirt!! (And then came CFLs. ...Mick may even know of a Sydney company that produced a pf=1.0 CFL last millennium, but CFLs were too political to stop imports and prevent corrupted voltages supplies! Nothing much has changed...!)

But clean SMPS exist. I figure if companies could produce 1kW dc inverters with 50khz switching with psophometric filtering 20 years ago - ie, no audible frequency noise between ~300Hz - 3kHz - plus the "normal" dc filtering, then surely today's companies ca do the same for a ****y little dc-dc converter. (Granted, 1kW from 48VDC; equivalent to 250W (300VA) from 12V DC.)


And the costs you (Chester) outlined is exactly why manufacturers need to fix their design - even if only to meet Standards.

Beside which - depending on local Regulations - they risk fines, damages, product recalls etc.

A bad enough product or incident can cause the whole industry to be investigated and tightened.

I understand where Chester is coming from and I’m actually leaning to his point of view; it really annoys me having to spend more money to fix a problem that should not be there in this day and age. My project has blown out in time and money because of the quality of commercial equipment,

Amazing how everything I design, build, program and hack together with basic tools works perfectly.

I think part of the problem is we want SMALL and highly efficient supplies in the car and “Hi frequency variable duty cycle” switching convertors (broad-band transmitters) work well for this but at the expense of noise. The other thing I notice in some of the lower noise supplies for 12v DC is their efficiency is often quoted in the 70% to 80% range and most likely a result of the frequency and wave shape used in the inverter.

As usual OldSpark is spot on in his appraisal of the quality of current SMPS.

Originally Posted by Mickz

Amazing how everything I design, build, program and hack together with basic tools works perfectly..

That must be an Australia thing!

LOL! Just don't read the forums I write of - they counter any brags of national talent with national sheep - but dumb sheep.

Purpose built converters can be 95% efficient, but up to 85% is more the norm. It will vary with load. (The 1992 1200VA inverter was circa 85% efficient using 50k (or 100k?) Hz.)

But noise to the extent you have mentioned - IMO, UNFORGIVABLE!