USB Isolated DC to DC problem

[oliglaser]

Ok, this one is driving me nuts now, I've been going round in circles for ages.

Basically I need an isolated, very quiet +/-5V supply for a USB scope I am designing.

I initially decided on a Murata +5V to +/-5V 2W supply, with +/-3V LDOs to sort out the ripple. Unfortunately, I can't find any ideal high speed, fet baased opamps for such a low supply, so that's that one pretty much out.

So I thought the ideal voltage for the opamps would be +/-5V, as there are a few suitable candidates there. The trouble is, the DC to DC is unregulated and there is no headroom voltage to use LDOs. So I thought I'd get a higher voltage DC to DC, like +/-9V, but then the losses from that and the LDOs will probably take me over the USB limit of 500mA (I've not done any exact calculations, but it doesn't look good)

So I thought the *ideal* solution would be to make my own DC to DC, and set it for a voltage of about +/-6V so I can use the LDOs to provide +/-5V and 3.3V for the MCU.

Two questions:

1. Has anyone got any experience with isolated USB supplies that could give me some advice?

2. Can anyone point me to a decent switching IC and the necessary transformer (I can find a few ICs, but the transformers seem to be more elusive)

Any help at all would be appreciated, as I've been battling with this for days and don't want to waste too much time/money prototyping stuff that's not going to work.

Oli

 

[M_in cadiz]

LT1533 from Linear.com is quiet and Wuerth make a range of "Flex" smps transformers

 

[keith1200rs]

I would have thought that there would be plenty of good CMOS opamps that will do the job at +/-3.3V. JFET ones tend to need too many volts.

What bandwidth do you require?

Keith.

 

[oliglaser]

Thanks for the replies

The bandwidth I require is around 20Mhz for single shot. The ADC will run at around 40Msps. I could "get away" with a bandwidth as low as maybe 5MHz, but this would be far from ideal.

But, I would like to provide Equivalent Time Sampling if possible, so a much higher bandwidth would be preferable.

The reason I looked at FET based was to provide a much higher impedance than the source that the opamp "sees" which is 1 Megaohm. It doesn't necessarily have to be FET, anything with >10 Megaohm should be okay to prevent level shifting/attenuation.

I'm beginning to think that just filtering the +/-5V supply from the DC well would be the simplest/cheapest option. I can reduce the ripple from ~100mV to <1mV with an LC filter. I have done a few tests in spice to confirm this (with ESR, inductor resistance included) With the PSRR etc of the opamp, surely this would lead to acceptable performance from an 8-bit ADC?
I'm not looking to create the worlds most sensitive scope, just a well priced, usable one.

 

[keith1200rs]

A few possible high speed, low bias current opamps from Analog Devices for +/-3.3V or +/-5Voperation: AD8033 AD8065 AD8067

There will be others from other manufacturers - you just need to use their opamp selection tools and input what is important to you.

Keith.