Talk:Emitter-coupled logic

From Wikipedia, the free encyclopedia

Contents

[edit] Comparison

See Talk:Transistor–transistor_logic#Comparison for a comparison of ECL, TTL, and a few other logic families. --68.0.124.33 (talk) 13:24, 26 April 2008 (UTC)

[edit] First sentence

The first sentence reads "In electronics, emitter coupled logic (or ECL) (sometimes called 'current mode logic') is a logic family which uses transistors to steer current through gates to compute logical functions" (emphasis modified, link added). My understanding is that a gate is an entire logical function, such as a NOR, so it is not appropriate to say that current is steered through gates. Let me suggest a re-write:

In electronics, emitter coupled logic (or ECL) (sometimes called 'current mode logic') is a logic family in which current is steered through transistors to compute logical functions.

Any reaction? Gerry Ashton 00:39, 8 July 2006 (UTC)

Yes, thats better and makes more sense! 8-|--Light current 00:48, 8 July 2006 (UTC)
I have made the change I proposed. I have also created a redirect from "Current switch emitter follower logic" to this article. This name was commonly used at IBM in the 1980s and 1990s for ECL. I think it's too obscure to bother mentioning in the article, but in case someone reads it and does not know what it means, they can search for it and end up at "Emitter coupled logic". Gerry Ashton 01:03, 8 July 2006 (UTC)
Current switch emitter follower is a new one on me! 8-)--Light current 01:04, 8 July 2006 (UTC)
It helps if I get the title right. Also, there is no way to put footnotes in a redirect article, so I documented some patents that use the term "Current switch emitter follower" in talk:Current switch emitter follower logic.

If IBM called ECL Current switch Emitter Follower logic at any time I am ashamed! IBM invented current switching in the early 1950's and called it Current Switching. It is definitely not an emitter follower as an emitter follower has its output at the emitter. I designed circuits for IBM from 1958 through 1990.UPCMaker 12:44, 26 August 2007 (UTC)

I designed ECL and related circuits at IBM from 1978 until about 1992. ECL was usually called current switch emitter follower, or CSEF for short. It consists of a current switch to perform the logic and an emitter follower to provide the output. Neither "current switch" nor "emitter follower" alone is a sensible name, but all four words together do make a sensible name. --Gerry Ashton 16:07, 26 August 2007 (UTC)

If you are saying that the ECL output had emitter followers added then I apologize. I assumed that you were referring to the conventional ECL current switch with collectors driving the output. Sorry!UPCMaker 19:06, 26 August 2007 (UTC)

[edit] ECL also called CML?

I've finally had a bit of time to look into this and would like a bit of feedback on how we could improve the intro. I assume that the word "emitter" in "emitter couple logic" refers to the emitter followers in the output structure (shown here). Since CML does not have those emitter followers, I'm at a loss to explain how CML would be another name for ECL. However, thinking about it further and looking at their output structures, I see how this confusion might have started, since ECL is CML plus an emitter follower. This book confirms that concept, pointing out that "ECL utilizes CML". As such, I'd like to propose we reword the intro slightly to make it clear that ECL "makes use of current mode logic" rather than implying that the names are interchangable. Also, could anyone do up some diagrams of the output (and maybe input) structures for this page and the Current mode logic page? —Mrand T-C 16:52, 23 September 2007 (UTC)

We need to base this on sources, not speculation. The current statement is sourced; if you have alternative sources that disagree, bring them up and let's see what we can do. As to your interpreation that the emitter followers are the key to ECL, I don't think so. I think it's the fact that there's a differential-pair type of logic structure (emitters of true and false branches are coupled to each other). The followers are just buffers to offset the voltage to what you need to go into the next logic stage, and to get more drive. It's possible that you're right, though, that the name changed based on those followers; let's look for sources. Dicklyon 16:59, 23 September 2007 (UTC)
Oops, I should read before I right. I see you did point out a source that supports your point. So now we should look for more and see if there's either a consensus or a historical change that we can mention to accomodate the difference. Dicklyon 17:02, 23 September 2007 (UTC)
Here is a book that implies the naming is based on the "source/emitter-coupled transistor pair" like I said, not on the followers. So if the emitter followers are omitted, CML is still ECL. Dicklyon 17:06, 23 September 2007 (UTC)
Your Ayers book also supports that as the basis for the name; it's not inconsistent with saying the ECL uses CML. Dicklyon 17:08, 23 September 2007 (UTC)
The alternative names were used way back in the heyday of ECL, in 1971. Dicklyon 17:11, 23 September 2007 (UTC)
This one suggests a simple name change, not depending on the emitter followers. Dicklyon 17:13, 23 September 2007 (UTC)
Very interesting - obviously not a straight-forward answer to this! Perhaps you (or anyone else?) could suggest a change to the intro to eliminate the confusion then, since as used in reference to real-world high speed logic signaling, CML is not the same thing as ECL?—Mrand T-C 18:27, 23 September 2007 (UTC)
The only "real world" I have access to is the space of references. See if you find a better explanation, or see what you can do with those I've found above. Dicklyon 18:56, 23 September 2007 (UTC)
I'm very sorry - I should have explained. What I meant by real-world was using IC's which have actual CML or ECL interfaces. As my link to Maxim's .pdf indicates (TI, National, and I'm sure others have the same type of app-notes as well) when engineers use the term CML or ECL, it is most certainly not used interchangeably with the other. Doing so would result in something that would not operate as intended since voltage levels, swings, biases, and terminations are different between the two. In short, I'm trying to say that while they share a transistor topology, they aren't the same thing and we need to find a concise way to explain that. Have fun, —Mrand T-C 19:08, 23 September 2007 (UTC)
Ah, right, I'm just being dense. I had noticed there is indeed a modern logic family called CML that's not the same as ECL. That's worth a mention in a footnote at least. Maybe there should also be an article on that CML. Dicklyon 19:16, 23 September 2007 (UTC)
The perspective of an encyclopedia is different than the perspective of a datasheet. While a datasheet may only be interested in what terminology engineers who order ICs built by someone else use, an encyclopedia must also consider what obsolete technologies were called, or what terminology might be used by the engineers who design the ICs, as distinguished from engineers who order fininshed ICs. It appears to me that historically, in the 1960s or 1970s, ECL and CML were used interchangeably, but today they are not. --Gerry Ashton 22:01, 23 September 2007 (UTC)

[edit] Expert editing needed

I was one of the few engineers in IBM that developed most of the discreet transistor switching circuits in the late 1950's and early 1960's. I believe I am about as much of an expert as there still is. Many Wickipedia articles on digital circuits appear to have been written by second hand observers and nearly every article I viewed on the subject have major and minor errors. I am forced to go by mostly memory and I am getting old. I do believe I know where my memory fails me. I fear much of the truth is about to be lost.

If there are others that lived this history or a historian who interviewed people like me twenty years ago, or more, then they may be able to help. If anyone has access to the actual documentation of that work that would be helpful?

I hope to write something but I am not sure how to present it. The subject seems to be scattered everywhere. I am thinking I should write one article in one place and link it to the many articles in Wikipedia.

Any thoughts? Any offers of help?UPCMaker 22:14, 25 August 2007 (UTC)

Documents from that time would certainly be helpful, but only if they are reliably published or otherwise publicly available. Copies of internal memos or invention notebooks "published" on a personal web page, or that sort of thing, would not be suitable for use in Wikipedia. A newly written history of 1950s and 1960s technology does not belong in Wikipedia, it should be published by a reliable publisher. For example, MIT Press has published books on the history of computing. --Gerry Ashton 16:15, 26 August 2007 (UTC)

So, an account of the history of transistor logic circuits is not welcome if it is just provided by someone who lived it. It is only valid if it is written somewhere else? Just let it get lost? Any other opinions???UPCMaker 19:12, 26 August 2007 (UTC)

You are correct, an account provided directly to Wikipedia by a person who lived it is not welcome. This is because Wikipedia does not have the ability, as a traditional publisher would, of establishing that the person who claims to have lived it really did. Of course, the material should not be lost; instead, it should be submitted to a traditional publisher who is in a position to verifiy the expertise of the author. --Gerry Ashton 21:25, 26 August 2007 (UTC)

The new ECL article is a big improvement! If only paper referenced articles are justified how did the Old ECL, TTL, and DTL articles get there? The circuits provided can not work and much of the text was in error. They had no meaningful paper references. I still think something should be said about the first Current Mode circuits used in IBM. The TTL was not the first logic circuits.UPCMaker 00:41, 27 August 2007 (UTC)

Wikipedia editors are not always good at following the rules and guidelines; your expertise can help a lot in finding errors, in finding references, and in marking where references are needed. On the early IBM logic, the old manuals needed as refs are probably accessible at the Computer History Museum, where an IBM 1620 has been restored and other old IBM restoration projects are ongoing. Dicklyon 06:39, 28 August 2007 (UTC)

In the schematic "Image:ECL.svg", I believe the base of Q5 should be connected the the collector of Q3. As shown the output from Q5 would have no signal.UPCMaker 11:52, 28 August 2007 (UTC)

Thank you for finding the error. I've fixed it. I also labeled the 907 ohm resistor that feeds the two diodes. --Gerry Ashton 16:56, 28 August 2007 (UTC)

In your Image:CurrentSwitchLogic.svg, I believe the output F=A+B should be F=(NOT A)x(NOT B), or F=NOT (A+B). The other (non-inverting) output of the first gate would be F=A+B. It is getting better!UPCMaker 19:50, 28 August 2007 (UTC)

You're quite right. I fixed it, but the Wikipedia SVG viewing software seems to be hit-or-miss about showing the fix. --Gerry Ashton 20:14, 28 August 2007 (UTC)

Looks good on the small display but when I click on it it still is the old way?? Hope you do not mind my changes I made in the text. Let me know if you have a problem with them. I guess they can be changed back but I think they make the discription more accurate.UPCMaker 21:25, 28 August 2007 (UTC)

I have the same problem as UPCMaker. I intend to leave it alone for 24 hours to give any Wikipedia mechanisms I don't understand to update, and then I'll ask for expert advice at the Village Pump or someplace like that. --Gerry Ashton 21:38, 28 August 2007 (UTC)

As long as you're making diagram fixes, the "differential amplifier section" should include Q3. Dicklyon 01:27, 29 August 2007 (UTC)

Thanks, Dicklyon, I made the change.
I noticed this has been added to the article: "IBM's Current Mode family in 1958 had 20 nano seconds delay as opposed to 500 nano seconds for DTL and 3000 nano seconds for RTL." Unfortunately, I only have one public source giving delay times for IBM technology, which is the Rymaszewski article cited in the Wikipedia article. Page 608 says Solid Logic Technology (SLT) offered four DTL families in the 7 to 700 nanosecond range, while the Advanced Solid Logic Technology (ASLT) offered "CSEF (current switch emitter follower)" in the 1.5 to 3.0 nanosecond range. A paper in the same issue, "Electronic Packaging Evolution in IBM" by D. P. Seraphim and I. Feinberg indicates SLT was introduced in 1964 (page 620). No date is given for ASLT, but it must have been before 1969, when Monolitic Systems Technology was introduced.
It would be great if someone could provide a source for the delay times. --Gerry Ashton 01:47, 29 August 2007 (UTC)

The delays I defined where for SMS and 608 families using germanium transistors and prior to SLT. IBM's current mode was introduced in SMS and the delays I specified where what it compeated with at the time. Faster germanium transistors came out shortly there after. I doubt if there is any documented information remaining for these delays. If there is I would like to find it. IBM did not make such information available to the public but kept its own records. I clearly remember the 20 nsec. current mode as I designed delay line circuits with this family and the 3 usec. was a spec for the basic RTL family. The 500 nsec. is from the 608 circuts used on a Sled II project that I worked on in 1957-1958. I later replaced one gate that had to handle a 0.5 usec. pulse but could not. I replaced it with my new design using a planar transistor that was much faster. The origial family was specified to handle 1 usec. pulses or longer. Maybe if it was presented as my memory it would make some feel better.UPCMaker 12:07, 29 August 2007 (UTC)

If you believe this information isn't available anywhere, you might consider writing an article for the IEEE Annals of the History of Computing --Gerry Ashton 17:10, 29 August 2007 (UTC)

I am thinking about it. I would really like to make it available to any one that is interested in it especially engineers who might like to know what it was like in the early development of solid state computers. Are you going to allow what I have offered here? Also, is a history from real life not wanted here even?UPCMaker 19:15, 29 August 2007 (UTC)

I'm going to leave it for now, in the hopes someone will find some sources. When we have a back and forth discussion with an editor, we may become convinced he is an expert, but if he stops using the Internet, future editors may doubt that the person really is an expert. Also, most users just read the articles, they are not editors and they don't read the talk pages. These readers have no way to know that unsourced facts were contributed by an expert. Finally, Wikipedia content is copied to other sites where there is on opportunity to interact with the editors. --Gerry Ashton 19:39, 29 August 2007 (UTC)

Last comment! I was thee expert in circuit design in IBM Research Triangle Park, NC from its start in 1965 till 1990. Should I remove my unverified material?????UPCMaker 19:57, 29 August 2007 (UTC)

You don't have to remove it, but some of us might. Or we might tag it with a "citation needed" tag, and if no citation is forthcoming it may eventually get deleted. Unfortunately, that's the way it works. Wikipedia is NOT a place for personal recollections; those should be published some place, and then can be incorporated and cited here. Talk to the computer history museum people about getting your recollections into their collection, and then maybe we can access that and cite it here. Dicklyon 20:20, 29 August 2007 (UTC)

The link in IBM Stretch references below page 369 verifies my 20 nsec. I think I will just give up and let you all have your fun! You do not need another amature writting in your electronics articles.UPCMaker 20:36, 29 August 2007 (UTC)

[edit] IBM Stretch references

Looks like what we need for refs is easily found here; see p.52 and search for other uses of "current" or "emitter". Dicklyon 01:36, 29 August 2007 (UTC)

Unfortunately, only selected pages are available, and I think what we want is on the pages that are not provided. But there's always the library. --Gerry Ashton 01:55, 29 August 2007 (UTC)
I just ordered a copy off abebooks.com, so in a few days I can answer some relevant questions about what's in it. Dicklyon 02:33, 29 August 2007 (UTC)

[edit] Expert verification?

I have 30 patents with patent numbers. Thirteen patent publications. Most of the early patents are in circuit design. One is for the coding method used in the UPC Label. All are in the public domain. I have three IBM corporate technical achievement awards, certificates and jewelry but not published. I have a number of informal awards. I invented the IBM Capacitive keyboard. I designed a number of circuits deemed impossible including, an A/D converter, a D/A converter, several telephone line modem functions, a modem for a police radio, a sense amplifier for my capacitive keyboard, and the sense amplifier for the IBM 650 double density drum. I have many other technical achievements but all but the patents could only be verified by witness testimony. I worked for IBM for 33 years and since retirement worked as a consultant on three project. I have very little achievements in hobbies or literature research.UPCMaker 22:06, 29 August 2007 (UTC)

That's all very cool. I hope you'll learn about how wikipedia works, and make it a new hobby. Dicklyon 00:03, 30 August 2007 (UTC)

Not likely. I'll leave the hobbies to hobbiests. I prefere reality.UPCMaker 00:56, 30 August 2007 (UTC)

[edit] Second Paragraph

In the second paragraph it still states that the transistors are always in the active region. The current switching, logic, transistors switch between active region and cutoff. They do not, however, saturate which is where they get their speed advantage. UPCMaker (talk) 22:50, 3 April 2008 (UTC)

Please concider my request for a correction that the current switching transistors do not stay in the "active" region. If it did it would be very noise sensitive and the outputs would be very poorly defined. UPCMaker (talk) 01:21, 11 May 2008 (UTC)

Done. --Gerry Ashton (talk) 01:06, 11 May 2008 (UTC)

Great! Thank you. UPCMaker (talk) 01:21, 11 May 2008 (UTC)