ESR vs a few uF, and good practice for mixed signal systems

rwgast_logicdesign

So not having a scope it makes it hard to see exactly whats going on as far as getting clean power signal. The 52khz switching regulator ive been working with requires a 22uf bypass cap on the input side. The data sheet says to use an aluminium electrolytic. To me this is kind of strange to me as the rule of thumb is ceramic bypass caps where ever possible. what would be the reason for this, is 52khz to low of a freq for ceramics to filter?

Also im wonder what my better option is to use two nice 50v 10uf caps in parallel on the input to get a lower esr, or to use a generic 50v 22uf. Im thinking the 2uf is with in the 10% tolerance that most caps are rated so maybe a lower esr is more faveorable here?

SRLM

Can you post a link to the datasheet?

rwgast_logicdesign

Of course , sry bout that

http://www.google.com/url?sa=t&source=web&cd=2&ved=0CDEQFjAB&url=http%3A%2F%2Fwww.ti.com%2Flit%2Fds%2Fsymlink%2Flm2574.pdf&ei=nl7RUPDDGpSr2AWeyIG4AQ&usg=AFQjCNFDhLpwIFDUjZ3RfgBC-C-R1u5yqA

davejames

rwgast_logicdesign wrote: »

Also im wonder what my better option is to use two nice 50v 10uf caps in parallel on the input to get a lower esr, or to use a generic 50v 22uf. Im thinking the 2uf is with in the 10% tolerance that most caps are rated so maybe a lower esr is more faveorable here?

Dude - throw the two 10s in parallel. Aluminum electrolytics typically have a +/-20% tolerance anyway. That "missing" 2uF ain't gonna be no bother.

RE: "...rule of thumb is ceramic bypass caps where ever possible..."? Ever see a 22uF ceramic cap? They're big in comparison!

SRLM

See page 14, item 4 Input Capacitor:

An aluminum or tantalum electrolytic bypass capacitor located close to the regulator is needed for stable operation.

A 22 uF aluminum electrolytic capacitor located near the input and ground pins provides sufficient bypassing.

<empahsis mine>

So, you can go with tantalum if you want.

INPUT CAPACITOR (CIN)
To maintain stability, the regulator input pin must be bypassed with at least a 22 μF electrolytic capacitor. The capacitor's leads must be kept short, and located near the regulator. If the operating temperature range includes temperatures below −25°C, the input capacitor value may need to be larger. With most electrolytic capacitors, the capacitance value decreases and the ESR increases with lower temperatures and age. Paralleling a ceramic or solid tantalum capacitor will increase the regulator stability at cold temperatures. For maximum capacitor operating lifetime, the capacitor's RMS ripple current rating should be greater than <EQUATION>

IIRC, you typically don't want too low of an ESR in certain high frequency applications because then the capacitor will be conducting too much current and will fail sooner. But that could be totally off.

Mark_T

Decoupling for the input side of a regulator may be less important than output-side if that input rail isn't used for anything else.

If the circuitry used +/-15V analog and you used the regulator from the +15V rail to general a logic supply, say, then that
input decoupling would be much more important to avoid noise on the analog rail (in fact I would avoid that topology
myself)

Using several 10uF ceramics in parallel would be my prefered approach, I think you get better ripple current ratings
that way (can be more expensive than Al elect caps of course)

PS haven't checked if that chip works from 15 to 5V, just making a point(!)

davejames

...when all else fails, RTM:

http://www.onsemi.com/pub_link/Collateral/HB206-D.PDF

Section 8 in particular.

Mark_T

davejames wrote: »

...when all else fails, RTM:

http://www.onsemi.com/pub_link/Collateral/HB206-D.PDF

Section 8 in particular.

We are talking about switching regulators, that's section 9.

rwgast_logicdesign

The forum just ate my post that took 30 minutes!! Here we go again! errrrrr >:-\ *Very P.O'ed*

Mark_T

You had mentioned it being a bad idea to connect the regulators to a higher voltage supply that was running analog devices. So I figured I should probably give some more details about my bot's power system, so I do not run in to issues later down the road that will need a scope to track down. At the moment I'm running the bot just fine, but the more circuitry added the more chances I have of dirtying up the logic signal with noise.

So first of all the bot is powered by a single 19.2NimH battery pack. I had originally thought of running the motors on there own battery but, decided I wanted a single power source. In the future I hope to build an autonomous charging system, and I also wanted the ability to regulate almost any voltage I may need. It seemed to me that having a single source would greatly simplify those tasks, and I've seen other bots including commercial bots like the Roomba use a single big battery. The reason I started down this Robot path was to learn about all kinds of systems and integrate everything in to a single massive system. It seemed like a really good project as far as a hands on way to grow my electronics and programming knowledge. One of the singal most important aspects of this to me is learning about analog electronics and good power supply techniques in mixed signal systems.

Ok so lets get down to business hopefully you can tell me if my power design is going to be problematic, down the road and if so what my options are to fix it.

Like I said above there is a single 19.2v battery. This battery is connected the fly back diode set up for each motor, my circuit is the standard four reverse biased diodes per motor. The battery is also connected to the motor supply pin of an h bridge chip. The motors have the standard three .1uf caps going to ground and to there casing, the h bridge also has .1uf caps between the power and ground for its logic, and motor supply inputs. I had also planned on putting a big electro cap on the positive/negative of the battery somewhere between a 470uF and 2200uF, I'm not sure how much this will help but I figure it couldn't hurt right? So besides regulators this is the only thing I had planned to hook up directly to the battery. I'm not sure if an h bridge chip or fly back diodes are considered analog circuitry or not, so the comment you made above (about connecting regulators to a rail supplying analog, post #6) spooked me a little bit.. If I am going about this wrong maybe theres a way to use opto couplers and still have a single battery or something along those lines?

Next thing is I just wanted to make sure the rest of my ideas are correct. I plan to have 3 power rails, all regulated by switchers with the recommended ripple filter shown in the data sheet attached to the output. These three power rails will be a 5v digital rail, a 3.3v digital Rail, and a 5v Analog rail.The digital rails will only be used to supply things like Propellers, the AVR, ram, real time clocks, digital sensors etc... The 5v analog rail will be used to supply all the analog sensors and any other analog circuitry such as op amps, comparators, 555s etc. More importantly the 5v analog rail will be used for the analog ref and ground to for any ADC chips and the the AVR's built in analog peripheral. Since im doing most things on proto board the idea is to route a nice star topology and for the digital ground and do the same for the analog grounds, after the digital and analog signals have had there grounds routed out nicely I will then use the thinnest piece of wire I can to connect the two ground star topologys together.

Hopefully im doing the best I can to keep noise down and isolate the analog and digital signal as much as possible. Like I said above this is all about learning and I know there's a lot of care that needs to be taken in mixed signal systems so by all means any of you bust my chops if im doing something wrong!

rwgast_logicdesign

Come to think of it this brings another question to mind, my motors have built in magnetic encoders. I can read them on a logic analyzer and for the most part I get a solid looking signal but sometimes the signals timing looks a bit whacky. Unfourtantely there's no data sheet on the motor im aware of. Would the encoders be considered analog or digital. If there just using hall sensors going low and hi this would be analog correct? Im not if thats actually how they work, im just guessing, I would assume there the same as any other motor with a built in encoder. I guess what im asking is for the sake of the power system do I want to power the encoders off a digital, or analog rail. Im obviously reading a digital value from the encoder but that doesn't actually mean its a digital circuit correct

kwinn

Most of the enccoders I see are digital optical encoders. They seem to be the most common and economical type. There are also Hall Effect digital and analog encoders as well as magnetic encoders that produce sine wave outputs (like a miniature alternator).

rwgast_logicdesign

Well all i can pick up on from what ive read and exprementation is that these are magnetic encoders, that want a 5v input (although 3.3v seems to work), the signal lines have built in pull ups, and are readable with a logic analyzer. Would it be safe to say they are definately digital signals, not just a low and hi analog value?

The way you phrased your post or the way im reading it makes me belive theres a difference between magnetic and hall effect encoder, is that correct?

agfa

The encoder output of my Faulhaber motors, as viewed with a scope, is a clean 0-5 volt square wave when supplied with 5 volts.

agfa

Dr_Acula

What would be your thoughts on adding another supply rail - a bog standard 7805 linear reg which you use to run all the sensitive A to D converters etc? Cost is a few cents, and it will save a lot of grief thinking about inductors and filtering etc. If you only run the sensitive things on this rail the losses from the linear reg are negligible.