isolated DC/DC converter

John Larkin <jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:

> On Sun, 06 Nov 2011 07:10:49 GMT, Mike <spam@me.not> wrote:
>
>>What kind of PCB tricks will keep fast edges out of wideband amps and a 
>>250MHZ ADC? 
>>
>>Mike

OK, Thanks.
 
> Keep the switcher far from the ADC stuff; 

You mentioned it was 2 inches away. That makes it difficult, to say the 
least!

> avoid circulating currents
> poked into the ground plane 

Tough to do with several sources - transistors and diode rectifiers 
inject noise into the ground plane

> by slitting the plane if necessary; 

That can make life more difficult by shunting other noise sources into 
the area you are trying to protect

> add
> ferrite bead+capacitor lowpass filters to the supplies of the
> sensitive stuff. 

Standard stuff

>Keep the analog stuff very tight. 

Also tough to do, especially with a wideband preamp

> And slow down those
> fast edges with gate resistors.

Williams used slew rate limiting in the switching elements for his ultra 
low noise switching converter. But he was looking at the ripple on the 
output signal, and as far as I remember, he did not show the noise 
injected into the ground plane.
 
You can do the same thing to slow the switching speed down, and reduce 
the fast edges injected into the ground plane.

> Compounding the difficulty, there's the thermal design, getting rid of
> all that heat from the linear regulators. Less noise, more heat.
> Should be an interesting PCB layout.

Cutting the ground plane won't help there. It will make the heat problem 
worse.
 
> John
 
I usually figure the first 40dB is free. The next 20 dB is harder, and 
anything past that is very time-consuming and gives less and less return 
for the time and effort.

From your photo at http://www.panoramio.com/photo/61564837, you are 
running from 24V. 

60dB down gives around 24mV or so of switching spikes, depending on how 
fast you switch and where the spikes go.

Most 250MHz 12 bit ADC's use 2V p-p or less input swing. The LSB is
2/2^12 = 0.000488, or 488 uV. If you add 20dB gain from an input op amp, 
you are looking at 48uV for the LSB. So you want the spikes to be less 
than that.

Then you need another 20 * log(24e-3/48e-6) = 53.97 dB of isolation on 
top of the free stuff. That will take some time:)

Of course, the estimates are very rough ballpark. But when you end up 60 
dB short, you know you are in for some fun:)

There are some other simple PCB tricks than can be very effective, but 
they are proprietary and I can't divulge them at the current time.

So I'd go for reducing the switching spikes at the source with slew rate 
limiting, and use soft turnon diodes in the rectifier instead of 
schottky.

Mike

On Sun, 06 Nov 2011 19:43:04 GMT, Mike <spam@me.not> wrote:


>There are some other simple PCB tricks than can be very effective, but 
>they are proprietary and I can't divulge them at the current time.

Lots of people say stuff like that. Makes no sense to me: how is
sharing stuff like that going to hurt you? And if it does a little, so
what?

John

On a sunny day (Sun, 06 Nov 2011 09:43:21 -0800) it happened John Larkin
<jjlarkin@highNOTlandTHIStechnologyPART.com> wrote in
<cvadb79e2hqa79n8crrt7j8t4cj5vsf127@4ax.com>:

>OK, but a class-C resonant converter has to circulate a lot of energy
>in the L-C tank, high currents into custom magnetics. Sloshing all
>that energy around doesn't seem, to me, to have any real advantage.
>
>I do recall some of the Tek 7000-series scopes having a
>single-transistor class C resonant main power supply, off the
>rectified AC line.
>
>We should establish a PayPal fund to buy you some nice gridded vellum
>and an electric pencil sharpener.
>
>John

Somebody I know bought a rather expensive digitiser pad.
If I had one that worked in Linux maybe that would be nice to make quick sketches.
There is also something in the opera browser that allows one to share sketches live,
tried it some years ago, but even on an extremely fast link
it was very slow.
Maybe I should check ebay.

John Larkin <jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:

> On Sun, 06 Nov 2011 19:43:04 GMT, Mike <spam@me.not> wrote:
 
>>There are some other simple PCB tricks than can be very effective, but 
>>they are proprietary and I can't divulge them at the current time. 
> 
> Lots of people say stuff like that. Makes no sense to me: how is
> sharing stuff like that going to hurt you? And if it does a little, so
> what?
 
> John

Why on earth would you say the following?

> And if it does a little, so what?

That is unbelievable. Why should I risk anything for someone who could care 
less how much damage it would cause.

You also use the same thing, John. Proprietary is exactly what it means.

Mike

Mike <spam@me.not> wrote:

> You also use the same thing, John.

> Mike

Here's a recent example. Not even proprietary:

  Subject: Re: Fast power LED driving
  Date: Sat, 20 Aug 2011 05:38:35 -0700
  From: John Larkin <jjlarkin@highNOTlandTHIStechnologyPART.com>
  Newsgroups: sci.electronics.design
  Xref: z sci.electronics.design:846558

  [...]

  I have  a cute and cheap way to drive mosfet gates  really  fast and
  hard. I'd post it publically, but Thompson would steal it...

  [...]

  John

There are many others.

Mike
 

On Sun, 06 Nov 2011 21:24:40 GMT, Mike <spam@me.not> wrote:

>Mike <spam@me.not> wrote:
>
>> You also use the same thing, John.
>
>> Mike
>
>Here's a recent example. Not even proprietary:
>
>  Subject: Re: Fast power LED driving
>  Date: Sat, 20 Aug 2011 05:38:35 -0700
>  From: John Larkin <jjlarkin@highNOTlandTHIStechnologyPART.com>
>  Newsgroups: sci.electronics.design
>  Xref: z sci.electronics.design:846558
>
>  [...]
>
>  I have  a cute and cheap way to drive mosfet gates  really  fast and
>  hard. I'd post it publically, but Thompson would steal it...
>
>  [...]
>
>  John
>
>There are many others.
>
>Mike
> 

I was ragging JT. He won't post anything for fear that I might benefit
from it. As someone else noted, I had already posted using NL37WZ16s
as fet gate drivers.

John

On Sun, 06 Nov 2011 20:39:36 GMT, Mike <spam@me.not> wrote:

>John Larkin <jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:
>
>> On Sun, 06 Nov 2011 19:43:04 GMT, Mike <spam@me.not> wrote:
> 
>>>There are some other simple PCB tricks than can be very effective, but 
>>>they are proprietary and I can't divulge them at the current time. 
>> 
>> Lots of people say stuff like that. Makes no sense to me: how is
>> sharing stuff like that going to hurt you? And if it does a little, so
>> what?
> 
>> John
>
>Why on earth would you say the following?
>
>> And if it does a little, so what?

I guess it depends on how close to 100% selfish you can get.

John

On Sun, 06 Nov 2011 21:24:40 GMT, Mike <spam@me.not> wrote:

>Mike <spam@me.not> wrote:
>
>> You also use the same thing, John.
>
>> Mike
>
>Here's a recent example. Not even proprietary:
>
>  Subject: Re: Fast power LED driving
>  Date: Sat, 20 Aug 2011 05:38:35 -0700
>  From: John Larkin <jjlarkin@highNOTlandTHIStechnologyPART.com>
>  Newsgroups: sci.electronics.design
>  Xref: z sci.electronics.design:846558
>
>  [...]
>
>  I have  a cute and cheap way to drive mosfet gates  really  fast and
>  hard. I'd post it publically, but Thompson would steal it...
>
>  [...]
>
>  John
>
>There are many others.
>
>Mike
> 

And why would I "steal" it?  More likely that I'd note that Larkin
infringes one of my patents or those of my clients.
		
                                        ...Jim Thompson
-- 
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

On Sun, 06 Nov 2011 13:12:48 GMT, nico@puntnl.niks (Nico Coesel)
wrote:

>John Larkin <jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:
>
>>On Thu, 03 Nov 2011 11:51:16 -0500, legg <legg@nospam.magma.ca> wrote:
>>
>>>On Wed, 02 Nov 2011 23:45:15 -0700, miso <miso@sushi.com> wrote:
>>>
>>>>Given base charge storage, if the driver rise/fall time is fast, the 
>>>>design could have shoot through.
>>>
>>>...and given magnetizing energy, emitters will be pulled above and
>>>below the rails. Shottkys can discourage this. The driver still may
>>>not be close enough to the rails to prevent base current and
>>>conduction in the switch, CE, to the opposite rails.
>>>
>>>Consider using a driver or controller intended for the function, with
>>>schottkys. 
>>>
>>>RL
>>
>>I want to avoid custom magnetics, and I want it to be cheap, in case
>
>These guys put a nice transformer together for me rather quickly and
>for a reasonable price (competing with off-the-shelf transformers):
>http://www.custom-transformers.co.uk/


In the US, these guys are great:

>http://www.minntronix.com/

but specifying and qualifying custom magnetics is still a nuisance.

John

On Sat, 05 Nov 2011 22:45:34 -0500, "krw@att.bizzzzzzzzzzzz"
<krw@att.bizzzzzzzzzzzz> wrote:

>On Sat, 05 Nov 2011 16:01:56 -0700, John Larkin
><jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:
>
>>On Sat, 05 Nov 2011 17:01:23 -0500, "krw@att.bizzzzzzzzzzzz"
>><krw@att.bizzzzzzzzzzzz> wrote:
>>
>>>On Sat, 05 Nov 2011 10:36:42 -0700, John Larkin
>>><jjlarkin@highNOTlandTHIStechnologyPART.com> wrote:
>>>
>>>>On Sat, 05 Nov 2011 12:15:02 -0400, Spehro Pefhany
>>>><speffSNIP@interlogDOTyou.knowwhat> wrote:
>>>>
>>>>>On Sat, 05 Nov 2011 10:59:48 -0500, the renowned Vladimir Vassilevsky
>>>>><nospam@nowhere.com> wrote:
>>>>>
>>>>>>
>>>>>>
>>>>>>John Larkin wrote:
>>>>>>
>>>>>>>>Vladimir Vassilevsky wrote:
>>>>>>>>>John Larkin wrote:
>>>>>>>>>
>>>>>>>>>>This
>>>>>>>>>>
>>>>>>>>>>http://www.panoramio.com/photo/61564837
>>>>>>>>>>
>>>>>>>>>>might work for powering a thing I'm doing.
>>>>>>>>>>Anybody got ideas for the base driver device?
>>>>>>
>>>>>>>>>You might want to check IR2085S
>>>>>>> Actually, I ordered some IRS2153D's, a very simlar part.
>>>>>>
>>>>>>So, getting a self oscillating thing to work seems too complicated ?
>>>>>>(http://www.abvolt.com/misc/self_osc.jpg)
>>>>>>
>>>>>>> A couple of bux of parts cost is not a big deal here. Getting it to
>>>>>>> work the first time is.
>>>>>>
>>>>>>Use a PIC.
>>>>>
>>>>>It should be pretty easy to throw some randomness into the frequency
>>>>>with a PIC for a spread-spectrum converter. 
>>>>>
>>>>>
>>>>>Best regards, 
>>>>>Spehro Pefhany
>>>>
>>>>I did a box that used a bunch of the LTM8023 bricks. I built a
>>>>TinyLogic schmitt oscillator and used the triangle to FM all the
>>>>switchers a little. It spread the switcher noise spectrum beautifully.
>>>
>>>What exactly did you do?  Dither the voltage?  
>>
>>
>>Just squirt a little triangle current into the Vadj pin. That dithers
>>the duty cycle. How much it wobbles the output voltage depends on the
>>triangle frequency and the output filter caps.
>
>That's what I suspected, but doesn't that also dither the output voltage?
>>
>>http://www.panoramio.com/photo/61713825
>>
>>http://www.panoramio.com/photo/61713828
>
>Impressive.

Sure is, great fun, textbook stuff. I just noticed, there seem to be
some Gibbs Ears on the spread-spectrum plateau.

John