The MagPi Magazine - Aimed at learners - Printed edition Kickstarter

Hi

To be honest I think there is little dissent amongst those who do know about these things (yes that does include me) about the demerits of this circuit.What little there is stems from "don't know where to begin - it's so bad" to paraphrase someone's comment.

Hi Graham,

Genuine question, what would your design be ? Like I say, i'm not as technical as i'd like to be, so eager to look at an alternative and learn from the differences.

Steve

No the L298 does use npn transistors top and bottom - really. It has the same limitations as any such design. The outputs cannot get to either rail - by a couple of volts when the load current is 2A. This allows for the bases to be higher than the emitters in the top transistors. This not a problem when you are switching upto 40V. It is done because npn power transistors are generally better and cheaper than pnp.

Hi Steve

Ok genuine answer.

Someone posted http://letsmakerobots.com/node/9450 as a defence of the circuit in question. If you look down to the post by Weiss you will see a standard, proper, efficient, safe, well protected, easily available, did I mention STANDARD design which freely available to anyone. Need I say more?

OK I will say more. It could be improved by replacing the output BJTs by complentary FETs - more expensive but better. And I have to admit I wouldn't fully trust any design until I had modelled it, tried it, etc though I don't think the values were shown anyway.

Better? Well a BJT driven into saturation wastes a small voltage (Vce sat) - anything upto 0V5 whereas the FET looks like a small resistance (down to a few 10s of milliohms for the big expensive ones) so RI leads to a much lower voltage loss. Also FETs tend to turn off when abused - self-protection whereas BJTs tend to turn on - so called thermal runaway.

So this circuit is more expensive but -

Uses only 2 GPIO

Presents much less load to the GPIO

locks out turning on the source and sink on the same side

And is less expensive than blowing things

And will probably be the one shown in next months issue, so what was the point of issue 8 complete with gratuitous and wrong calculations. All that was needed was a block diagram and a comment that the resistors needed calculations.

If you need more feel free - since I retired I no longer charge £100/hr

Regards

Graham.

Hi All,

Thanks for the helpful comments.  I think the main problem is that the transistor was selected incorrectly and that probably shifted the expectations of what it would be able to achieve.  I fear that the circuit probably won't be workable in it’s current state, as a transistor able to switch the power required and still only need a low base current is going to be hard to find (without using Darlington pairs or something else - negating the hope to make it a simple circuit). The circuit did apparently work, but I can't confirm myself how well or how long it was tested for.

I'm not sure what the best option is regarding the article at the moment.  But the team is looking at the options and hopefully we can get it resolved.

A simple design which is close as possible to the principle involved would be helpful, would replacing the transistors with suitable Darlington Pairs be enough (with the calculations re-done) - I've not looked yet to see how the numbers stack up (I don't do electronics enough to have all the numbers/calculations in my head like some of you probably do).  Anyway, a fix for the current design and some words to suggest that better engineering solutions will be discussed later may be the best option (if nothing else, dropping the control pins down to 2 is worth a new circuit, and obviously suitable driver ICs can be talked about).

I've been looking at other designs (was looking at making a basic h-bridge anyway for my own projects), but yet to try any out and they wouldn't really fit into what the article was attempting at that point.

I hadn't noticed that the topic had been locked on the RPi forum, but I imagine they try to keep a tight ship there to avoid things descending into fights, or upsetting people (particularly since they have a lot who aren't used to forums).  To be fair, most people should be able to have any discussion when faced with such limits with some effort.

As for the comments about a spending money on a good text book, I totally agree there, since there is no way the information in the magazine would be as detailed or as structured as a book.  However, we hope that the magazine will inspire people to get to the stage where a textbook on the subject would be interesting and useful to them.

Also fair comment about lower level options on the kickstarter...this wasn't done on purpose, just not something we thought of at the time and we wanted to avoid making the printing complicated by splitting things.

Thanks again for being kind.

meltwater.

That second circuit is indeed already an improvement.

You could use darlington transistors for the power ones.

However, if you drive both inputs high, I think you will still see smoke.

Power fet's have a capacitive input, so to achive fast switching times, you need to be able to drive the gate to vcc and to gnd.

This makes the circuit more complex. Power mosfets usually don't saturate at a 3V Gate voltage. Most of them need something between 5V and 12V

The original circuit probably worked because the unit had low voltage motors. No idea how hot the transistors became, as they are only in TO92 housing.

@Graham

Yep that circuit looks similar to the one I had in mind (was writing my post as yours was posted it seems).  Also, I'd not tried it yet or I probably would have suggested it at the time.

The answer to why have the other one in Issue 8, it was the author's aim to try to show how they got to their solution they had used.  Obviously this is a problem with having a small amount of resource and time available to cover everything we need to.  I think the whole thing does show that often such things aren't going to be clear to beginners either, which was what was trying to be addressed (even if incorrectly).

The series of articles which this one is from is a great project and I think that it is great they have taken a lot of time to try to explain what they are doing and how they are doing it.  Even if it is just in their spare time.  Highly commendable I think.

Hi

You need to look for low-threshold or "logic-level" FETs which typically switch with 2V.

I wouldn't worry about turn times - the gate resistors will be only 100R or so and so the FET will switch in a uS or so.

The Weiss circuit is good enough really though.

If both inputs are driven high then smoke will happen. To avoid this you need a new topology - still two inputs, now called enable and direction. Or a couple extra transistors arranged to ensure that a turned on input transistor at one side disrupts the signal in the other side, and vice versa.

All more expensive but properly put together almost bomb-proof and much cheaper than an accident.

You might ask why professional designs don't seem to bother so much about bomb-proofing. Well in my experience the final product will be on a pcb and driven by well-tested software, and if there are a few disasters during experiment (sorry design) well what's a few components when you are paying/charging £100/hr for your labour? And proper lab power supplies tend to protect experimental hardware, and did I mention design software, modelling and simulation?

MagPi is a fine idea, however one doubtful article where I know something seriously damages credibility in those where I don't - areas which includes Linux (ok I know something), python, and one or two (well almost all really) other things.

Graham

Oh, and just found out that this forum does not like iPads so a couple of posts have disappeared including the one where I said I now understood and agreed with MK about his assertion of 6W dissipation in a non-saturated (the condtion I failed to note) bottom transistor.

I think you have probably shown why the simple circuit was aimed at, and by the time you have got to the bomb proof stage, an IC is the obvious answer (that is why they sell them).  However, you don't try to teach someone calculus by getting them to write out Fast Fourier Transforms, they have to understand the parts which make them too.  So yes, more circuits and solutions are needed to cover this topic fully - it is not something which can be covered all at once or over a few pages.

Of course we would welcome a clear and well written article which covers this topic in detail for a future issue from someone who has the skills and expertise to do so.  Graham obviously you do know this topic well and are keen for the right information to be presented on the topic and are keen to highlight where we have gone wrong so far, so I would welcome an article from yourself (or anyone else for that matter who is keen to help others learn).

There are many similar topics like this which would be a massive benefit to the community if clear stage by stage explanations can be provided to help beginners (and hobbyists) gain a deeper understanding of them.

An interesting point was made previously about why go through the steps rather than give the solution and be done with it.  There are a lot of reasons for this, firstly being a community magazine, we are happy for authors to take their own approach to their articles, afer-all what we think is a good way to present something may not be the right way for everyone, so the diversity which people bring to the table is very important.

Also, the articles (particularly this series) are a reflection of the process the author is going through (a step or so behind his current stage) so there is no full step of instructions for what he is doing.  I think more importantly though, this reflects the real world of engineering, real projects unfortunately don't usually come with a step by step guide and part of the role of an engineer is to turn the real world problem into such a guide so that others can make what they have made (it's not a perfect process as many of you know full well).

Finally, the reason why magazines do split such things up is because not everyone is interested in the same things.  If I'd brought the latest issue of Stuff, I would be very annoyed if it had all the pages filled with the iPad, I don't have one and although I don't mind reading about it, I also want to read things which interest me too.

We welcome useful input to the magazine, we maintain a live draft of the coming issue (linked from our site) to allow comments and feedback as we work on it, and are keen to have more people on the team of all levels of skills and knowledge to help us continue to improve and grow.  We are lucky to have a great team who work very hard and do whatever they can to help.

I think I'll draw my involvement with this to a close now, as I have a better idea of what the issues are, what actions we could take on the article and most importantly what the people here (a section of our readers) would like as we move forward.

Many thanks for your input on this,

Meltwater. (on behalf of the MagPi).

Steve,

You wrote:

> Also, i'm not sure what the point is of doing two articles on programming languages (Cecil and ADA) that are hardly in the sphere that the Pi is aimed at.

I disagree with this.  I think teaching kids of all ages to program is squarely within

"the sphere that the Pi is aimed at", and doing articles on programming languages

is squarely within the sphere of teaching programming.  Certainly a Computer Science

degree at any reputable school requires a class surveying multiple programming

languages.

There seem to be hardware engineers that would like to see more emphasis on

hardware topics, and software engineers that would like to see more emphasis

on software topics, but that's to be expected.  Byte Magazine used to have a

popular column by Steve Ciarcia on hardware, and Jerry Pournelle on software

(including surveys of programming languages such as Ada), which worked well.

It is generally accepted that knowing at least the highlights of several programming

languages, even obscure ones, makes you a better programmer in whatever language

you end up using.