3.3v <> 5v interfacing @ 15Mhz

If you run into problems, then as I live near Cambridge (work near there too), I might be willing to help you out - think of it as a xmas gift or whatever.

Cheers

PeteS

There is no way to argue with a punk. Don't go away mad, just go away.

When your accumulated income exceeds $5 million, *perhaps* I'll deign to listen to you ;-)

...Jim Thompson

isn't High typically specced as anything above 2/3 VCC.... (that'd be 3.33v)

What voltage are they running the DSPIC ?

if you power the PIC from 3.3V it'll probably work just fine

To give yourself some more margin, look at adding some parallel C to the series resistor (same peaking/matching princple as a scope probe), on those lines that need to do 5V to 3V.

What you are making is a capacitive divider (high frequency) that matches the resistive divider (low frequency), and it allows the resistor values to increase without a speed penalty (for lower power). eg If the total load shunt C is 20pF, you'll need appx 40pF.

Avoid transistors, they will be too slow.

-jg

.. When driven to the rails ?

Interesting circuit; do you have the plots of Icc vs Vin, increasing and decreasing ? And also Icc into/out of 5V and 3.3V rails ?

Limits of Vcc delta's ?

-jg

There are a few other things you might want to consider. I don't know how many of these you plan to make or how they will be used. But anytime you have external signals, it is a good idea to protect them from ESD. If you can work with a real PCB design (made by a PCB maker), you can use a part that might be better. Here is one I found at TI.

It provides 15 kV ESD protection and will do the level translation without the resistors and messy circuitry. It does not come in a 0.1" DIP, but it is available in SOIC which is 0.05" pin spacing, large by today's standards. You should be able to get a few as free samples. All things considered, something like this is the route I would take. I don't know how SD card I/Os are typically protected, but this is something you might want to take seriously.

If you want to have a PCB made for you inexpensively you can go to Olimex which I think is in Eastern Europe.

Oh, be sure to use lots of power and ground plane area, flood fill if you go two sided. This provides a good high frequency capacitor that minimizes noise on the power planes.

Let us know how your design works out!

Maybe I don't undestand what you mean by "copperclad". To me that is the same as PCB. Are you talking about mouting components to a pre-etched general prototype board or a PCB that has *not* been etched at all, or something else?

Not quite to rails.

See (updated)....

V1 = 5V, V2 = 3.3V

I don't know an answer for that. This presentation was just a lash-up. My actual ASIC implementation uses single-stage inverters instead of the 74HC04's, so my sizing ability allows very low overlaps... it's basically a form of hysteresis.

I'd guess that wildly disparate power supplies would present some slew issues and give more overlap current... but I'd hazard a guess that

...Jim Thompson

Maybe I'm dense but isn't that that pretty much the intended application for HCT parts? specified in the datasheet down to much extra current consumption when going to limits of Vi and Vcc

-Lasse

"HCT" parts were designed to receive signals from TTL parts also operating at VDD=5V.

The shifted threshold was done to minimize timing skew when making the translation.

Over time they've come to be misused as logic level translators, even though there are specific parts for such purposes.

Unfortunately many lurkers here presume that "getting away with it" is good engineering practice :-(

Me, I have to deliver good stuff every time... otherwise I'm unemployed.

...Jim Thompson

look at olimex.com they are pretty cheap, 160x100mm two layer pcb with solderstop and silk for something like 33$

-Lasse

The venerable Alison etched in runes:

. . .

Not at all. Have a look here:

and go to the 'Plot & Go' service. They'll do one-off prototypes very cheap and good quality too.

RTFM and think. Give it a try.

HCT/AHCT is good only for unidirectional 3.3V->5V translation. In this case it is the most convenient and least expensive solution.

Specialized IC can be used for bidirectional level translation, when different voltage levels required, like 1.8 3.3V, etc.

Bare copper on fr4. You can usually get scraps from any pcb house for free. Only one side needs copper. You can cut it with a knife, hacksaw or tin snips to the required size.

There are many ways of mounting the components. I use plain hot glue most of the time, but also screw components such as heatsinks to the board if needed. You can cut the plane to isolate components if needed.

See Terry Ritter's page for an example of Manhattan style:

Since the copper is not cut or etched, it makes a much better ground plane than a pcb, and you can get excellent analog and digital performance up to about 1GHz. However, fast risetimes do not do well with copper wire interconnections. For example, the Motorola EclInPs with 160ps risetime is a problem. You really need stripline or microstrip to handle these signals.

A big advantage is you can easily move parts around if needed to add new circuitry. But complex circuits with a wide data buss would be difficult to wire.

After a few days, you will see your fingerprints etched into the surface. The copper may eventually darken due to corrosion. Brasso may clean the oxides off, but I'd rinse the board carefully to remove any trace of acids.

The technique is ideal for testing new components or circuit concepts. People have made entire high-performance radio receivers using this method. Each circuit sits in its own shielded box for good isolation. When you want to try a different mixer or amplifier, it's easy to pop a new board in and test it under actual conditions.

Regards,

Mike Monett

Antiviral, Antibacterial Silver Solution:

Three things:

a) Vih is typically 0.7*VCC (that'd be 3.5V). Minor difference, but see..

b) A 3.3V 5V circuit contains two power supplies, each with its own tolerance. If 5V goes +5% (5.25V) and 3.3V goes -5% (3.135V) at the same time, the situation gets worse. Vih is now 0.7*5.25V = 3.675V, while the 3.3V sender can generate 3.135V max.

c) You're missing that a 1 generated by a 3.3V device is not 3.3V. It's Voh. Typically CMOS outputs tend to climb up to almost VCC, but not all do. And certainly some devices trade lower Voh for a faster timing spec (ie faster device without silicon change).

A fool who thinks that this might be related to the timing problems of the GP... ?!

Regards, Marc

I understand what you are saying, but your first few posts were not at all clear and only appeared to be attacks on others. I am not trying to place any blame for this flame war, as there is plenty of blame to go around. I would suggest that you might consider how your posts appear to a third party reading this. Take it from someone who knows, this is not the best way to promote yourself.

BTW, sort of like in the movie Chinatown, "Let it go Jack, it's just the Internet!"

Ok, this is not addressing my question. I know what copperclad is. I don't understand how you are using it.

Are you saying that you are just using the board as a mechanical mount and not to make connections? How do you make connections with point to point wires?

Ok, I guess this is a bit more clear.

This seems very difficult to me, but I do digital much more than analog and the number of connections with multi-pin packages is much higher.

One of my favorite movies... in my DVD collection ;-)

You're right, I should ignore the flack.

...Jim Thompson

I have to admit that I didn't expect the cross conduction to be that high in the HCT at 3.3V input levels. Actually it made me rethink my position. The HCTs can be used as 3.3V -> 5V level translators, however the user must be aware of the power consumption.

BTW, the maximum of cross conduction appears to be at 0.9V input (5Vdd), and it is about 4mA.

VLV