Suggestions for a low-interference smpsu please. / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Suggestions for a low-interference smpsu please.

previous 1 2 3 next

Move Thread

LAN_403

WaveRider

Fri Oct 07 2011, 09:35AM

Registered Member #29 Joined: Fri Feb 03 2006, 09:00AM
Location: Hasselt, Belgium
Posts: 500

Of course, good SMPS design is needed. It is standard industry practice to use layers of linear regulation on sensitive circuits to reduce PSU interference. A low-pass filter stage on the SMPS output might give 20dB attenuation of ripple, a first series pass reg another 25dB and a final reg close to the relevant system stage another 20-40dB rejection. 65-105dB of rejection is pretty good by most standards. Also, linear regs are a lot more compact and cheaper than large inductors. Of course, there are trade-offs to consider, mainly required PSU efficiency versus permitted PSU spurs in the signal chain. Putting small linear regs near system blocks provide another advantage. They reduce inter-stage coupling of signals through power supply lines, thereby reducing the possibility of annoying, hard to diagnose instabilities in high-gain amplifier chains!

Good design of the SMPS is also crucial, but some bleed-through of SMPS switching signal is pretty much unavoidable if steps are not taken to clean up residual ripple after the SMPS. Local oscillators are VERY susceptible to PSU noise, so pay special attention to providing a clean supply to them. Multiple layers of regulation is one of the best ways to guarantee a clean LO tone to your Rx, Tx.

(Useful reg to put near system blocks with good rejection around 100-200kHz
]lp2985.pdf[/file])

Patrick

Fri Oct 07 2011, 04:32PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Im still looking for a PFM ic, but ive got a lot of homework. Im looking on mouser and TI.

EDIT: This is a high power PFM resonant circuit for high power, not sure if its too complicated for your purpose though.

]fa2ffan7621.pdf[/file]

radiotech

Fri Oct 07 2011, 04:53PM

Registered Member #2463 Joined: Wed Nov 11 2009, 03:49AM
Location:
Posts: 1546

If your institution has access to AES, this link may help

The EMI, Filter component, and Pcb layout discussion may be helpful.

GeordieBoy

Fri Oct 07 2011, 09:52PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Hi Sulaiman,

I don't post on here much these days, but here are a few sugestions from an ex SMPSU designer...

Choose a forward converter topology. At least the output current will be continuous, and this is the port that you want to be clean of rf for your application.

Avoid boost and flyback converters as the former has a discontinuous output current and the later has both discontinuous input and output currents! These are inherently noisey in terms of EMI and the flyback can be challanging to silence in terms of EMI. The magnetic gap in flyback transformers also spews out leakage magnetic flux lines!

Go for the push-pull transformer-coupled forward converter like you said. It's easy to get working and relatively easy to stabilise. Output current ripple is also at twice the switching frequency so easier to filter out.

Use an inter-winding faraday screen between the primary and secondary sides. This should be connected to the negative rail on the input side to return all HF switching hash to the power side via the shortest possible route.

Keep the switching frequency as low as possible. Yes the magnetics will be bigger but it will be quieter.

Slow down the turn-off and particularly the turn-on transitions of the MOSFETs to limit di/dt at switching instants. Don't slam the devices on or off any quicker than you need to in order for it to run cool. High di/dt is an RFI nightmare. Slugging the turn-on of the switches also gives rectifier diodes longer to recover and minimise the reverse recovery current spikes. Slowing down turn-on usually initially dramatically reduces EMI but then eventually impacts efficiency as devices linger too long in the linear region.

Keep the layout of the paths containing switched currents tight. Reducing stray inductance combined with limiting di/dt goes a long way to reducing voltage spikes from L x di/dt.

Use a toroidal iron-powder buck choke at the output. Stick inductors and gapped ferrite cores have large leakage fields. The distributed air gap of iron-powder toroid minimises leakage flux.

If you must use gapped ferrite, then use a flux band to reduce EMI.

Size the buck choke to make sure the inductor current is well into the continuous current mode at full load. Discontinuous current mode is noisey because peak currents are higher than necessary. Ringing also often occurs when the inductor current falls to zero.

Wind the buck choke toroid in a single layer winding to minimise end-to-end capacitance. Don't wind one end of the winding over the top of the other as the stray capacitance provides a great path for EMI to skip past the choke!

Use ripple steering at the output of the converter to divert the HF component of the inductor current to a ripple-steering capacitor and divert the smooth DC part to the output port.

Use multiple low ESR electrolytic capacitors in parallel to minimise input and output voltage ripple, rather than one single large uF capacitor.

If necessary use post-filtering (pi-filter) at the output side to further reduce differential-mode voltage ripple. This may need to be outside the voltage feedback loop otherwise the additional phase-lag from the post-filter may make control-loop stabilisation difficult.

Put the whole SMPSU in a metal enclosure to reduce radiated emission.

Use high permeability ferrite common-mode chokes for the input and output side filtering to reduce conducted emissions from both ports. Most RFI in the HF range is usually common-mode noise, not differential mode current ripple from the buck inductor.

Don't place ground-plane copper fill underneath switching nodes with fast dv/dt. Stray capacitance from high-voltage switching nodes causes large current spikes at turn on and turn off.

Keep the two alternative paths for the output current through the two output rectifiers as close as possible to each other on the PCB. The larger the distance that currents are redirected across the PCB when they switch the more radiation. This applies to any currents that switch repeatedly between two paths.

Use dissipative RC snubbers across output side rectifiers to damp any ringing caused by transformer leakage inductance.

Although modulation of the switching frequency is a common technique to achieve EMC compliance commercially, I wouldn't recommend using this for a radio receiver supply as it tends to spread the switching noise over a wide range.

You'll find loads more tips like these in power electronics books, but I hope these help.

-Richie,

Patrick

Fri Oct 07 2011, 10:13PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Everthing Richie said is gold, pure gold, im printing this right now!

Print Button Engage!

Sulaiman

Sat Oct 08 2011, 08:57AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

Richie, thanks for all the hard-earned experience.

The main changes to my design will be

- a powder toroid instead of a gapped-ferrite buck inductor
- choosing a slower than considered turn on/off scheme
(I was going to use TC4421s just because I have them)
(and to reduce stress on the TL494)
- snubbers for the rectifiers
- going to read-up on 'ripple steering'

Thanks!

Although I could use the smps for all of the low voltage supplies
I'm building the rig in two (recycled 19" rack) units,
one unit is a self-contained QRP (low power) transceiver
with an onboard 12V 7AH battery and charger
(and 2x 5W solar panels)
so I use multiple linear regulators.
one is this inverter, a hf PA, atu/amu, swr etc. and maybe receive pre-selection, with external battery power.

If this inverter proves too troublesome then the pa unit will have an external 24V battery!
(could be charged/maintained as 2x 12V in parallel by the QRP unit)
P.S. books cover so much, I really appreciate the knowledge of what are the key points which come from experience.

P.P.S. I just had a quick read on 'ripple steering'
.. it looks like it's great if you can spend a lot of learning time
I think I'll give it a miss for now.

Thanks you all, time to build it, I'll post here with results
........eventually

Pinky's Brain

Sat Oct 08 2011, 09:43AM

Registered Member #2901 Joined: Thu Jun 03 2010, 01:25PM
Location:
Posts: 837

wrote ...

Slugging the turn-on of the switches also gives rectifier diodes longer to recover and minimise the reverse recovery current spikes.

What reverse recovery? His voltage&current is low enough to use Schottky diodes.

Shrad

Sat Oct 08 2011, 09:57AM

Registered Member #3215 Joined: Sun Sept 19 2010, 08:42PM
Location:
Posts: 780

Hi Richie, and Sulaiman

do you mind if I use extracts of this discussion later in a kind of faq I'm slowly building in the dark? credits will be put where they should..

Sulaiman

Sat Oct 08 2011, 10:18AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

Pinkys Brain .. I have/will use a common-cathode dual-schottky TO220 rectifier MBR2080, 20A 80v for 28V 5A output.
Does that mean snubbers wouldn't help?
I bought the switching transistors and dual rectifier based on price and over-engineering ...
(I like through-hole components, TO220 package where possible, no pcb etching, no smd and definitely no programming)
probably ok to 100s kHz, whereas I'll use 20 to 40 kHz,
and 'slow-down' the semiconductors.

Shrad, it's a public forum, and I don't mind,.....Richie?

P.S. Richie, you can't take all the gems of experience with you to oblivion ... it's your moral responsibility to publish all the knowledge and experience from the inner-guild of smps designers.
and answer all my questions before I ask,
and design my smps for me ....
oh, and if you're free could you build and test it for me too? ;)

P.P.S.
the TL494 datasheet gives about 4% minimum deadtime
at 20 to 40 kHz, thats 1 to 2 us
... suggest a good time constant for rise and fall anyone?
I'll start with Cgs=3n3 and Rg = 100 Ohms.
EDIT ... TL494 doesn't have totem=pole outputs....

Pinky's Brain

Sat Oct 08 2011, 10:44AM

Registered Member #2901 Joined: Thu Jun 03 2010, 01:25PM
Location:
Posts: 837

Sulaiman wrote ...

Does that mean snubbers wouldn't help?

I don't have the experience to make any statements regarding that ... I'm just saying your diodes don't have reverse recovery as such (a little junction capacitance sure, but no reverse recovery).

previous 1 2 3 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob