Talk:Resistor–transistor logic

first digital logic

Resistor-Transistor Logic may well have been the earliest form of transistorized digital logic, but most people tend to associate the term "RTL" with integrated circuits, and the first family of digital integrated circuits was Motorola's MECL family, introduced in 1962.

The fan-in limitation described is specific to RTL integrated circuits; the fan-in of discrete RTL circuits varies widely depending on the specific components used.

--Brouhaha 02:04, 14 Sep 2004 (UTC)

expert help requested: is it really RTL?

Is this really a RTL gate?

Or is this the real RTL gate?

• Some people claim a 3-input RTL gate has 3 transistors[1].
• Some people at first seem to claim a 3-input RTL gate has only one transistor[2].

I suspect that historically, the "3-input gate with 3 transistors" is the kind historically defined as "RTL", and the arrangement that was actually implemented. I suspect the "3-input gate with 1 transistor" view is a misunderstanding, was never actually implemented -- or if it was implemented, it wasn't called "RTL".

Could an expert clarify if a "RTL gate" always used one transistor per input, or if there were significant numbers of things built with only one transistor per gate, and that the builders actually called them "RTL gates"? --68.0.124.33 (talk) 13:47, 28 March 2008 (UTC)

Yes, all sources that I can find, going back to the 1958 3rd edition of the GE Transistor manual, show multiple-input gates using a single transistor, and a resistor per input, and call them RTL. The particular figure above, however, may not be exactly right, according to the editor who uploaded it, and he doesn't have a source for it. The base bias resistor is most often tied to a supply of the opposite polarity as the collector supply. These are usually shown with PNP and called NOR gates, but of course they can be made either way, and can be called NAND or NOR depending on how 0 and 1 are assigned to high and low voltages. Some sources: [3] [4] [5]. Dicklyon (talk) 05:52, 29 March 2008 (UTC)

Thank you very much for those links.

Oddly enough, all the sources I have handy make the opposite claim.

They say RTL always has one transistor per input.

Sources: "RTL Cookbook" by Don Lancaster (1969); [6]; [7]; [8]; [9]; [10]; [11]; [12]; [13].

The Apollo Guidance Computer article mentions "integrated circuits (ICs)... containing ... 3-input NOR gates ... made ... using resistor-transistor logic (RTL)."

Those gates had one transistor per input[14]. —Preceding unsigned comment added by 68.0.124.33 (talk) 17:58, 23 April 2008 (UTC)

While googling for more information, I came across "Q: When is RTL not really RTL? A: When it's on a chip, silly!", which seems to explain the conflict.

Dicklyon has convinced me that "one transistor per gate" was often (always?) used when building things out of discrete transistors.

However, as far as I can tell, all chips called RTL always use one transistor per input -- chips such as the MC789P (MC789F), µL914, MC724P, MC715P, MC785P, MC799B, µL900, MC790P, MC791P, MC776P, MC787P.

Perhaps we should give these 2 different logic family implementation styles distinct names.

What do you think about calling them "RTL" and "MRTL", at least in this article? Can you think of a better pair of names?

Is the difference between RTL and MRTL "notable enough" enough to warrant 2 different Wikipedia articles?

--68.0.124.33 (talk) 17:36, 23 April 2008 (UTC)

Since *both* of these circuits were actually used in real computers, I've restored 2 different schematics. --68.0.124.33 (talk) 02:16, 3 October 2008 (UTC)

I think that any 3-input RTL NOR gate that uses 3 transistors is actually 3 RTL inverters with their outputs connected as a wired-AND. Otherwise you would have to say that RTL allows infinite fan-in by just adding transistors. If you limit RTL gates to 1 transistor, then you must say that an RTL gate with multiple inputs connected via a resistor network has a very limited fan-in because each input has to compete with the others. Then you can see the advantage of replacing the resistors with diodes, and understand how this leads to DTL (which is slower but allows more fan-in).

I can easily imagine RTL designers both circuits. To minimise transistors they could use one-transistor-per-gate if there are only one or two input, and multiple-transistors-per-"gate" if they more inputs required. That would explain both circuits.

60.241.12.136 (talk) 09:59, 18 October 2009 (UTC) Computation structures

By Stephen A. Ward, Robert H. Halstead (Section 1.8.4) explains how fan-in limitations force RTL to shift from "input gating" (multiple inputs on 1 transistor) to "output gating" (1 transistor per input) "early in the history of RTL"

Google RTL "output gating" —Preceding unsigned comment added by 60.241.12.136 (talk) 10:44, 19 October 2009 (UTC)

... So then the problem of limited fan-in can be solved by using "output gating" instead of "input gating", but doing so means you must use at least as many transisitors as you have inputs, which would have made it it more expensive than DTL back then, which only require a separate diode for each input. 14.203.83.243 (talk) 08:00, 12 September 2016 (UTC)

Simplified schematic

It should be emphisized that the diagram is a "Simplified schematic" and does not actually work. Someone should be able to find a "published" schematic that actually works. I can only provide a design similar to those used in early computer designs. I would hate to think that some future engineer would try to build the circuit as shown and find that he wasted his time. UPCMaker (talk) 23:02, 3 April 2008 (UTC)

Actually it's no simpler than a correct one, so I replaced it, using the NPN version of the circuits in the sources I mentioned above, which is exactly what's in the cited GE Transistor Manual. Dicklyon (talk) 03:46, 4 April 2008 (UTC)

If R1 returned to a negative supply it would then work. If R1 was removed and the transistor is silicon and the temperature is low enough the circuit might be workable but very slow turning off. If the transistor is germanium, R1 returning to a negative supply is needed to turn the transistor off. R3 and R4 can turn the transistor on without any help from R1 but R1 will make it difficult or impossible to turn the transistor off. R1 returning to a negative supply is needed to overcome base leakage and provide a low enough base voltage to assure the transistor is turned off especially with a germanium transistor which has high leakage and low base emitter voltage. UPCMaker (talk) 12:44, 4 April 2008 (UTC)

I had fixed that last night, but failed to complete my edit. Now it's done; the circuit is exactly as you say, as in all the refs. Dicklyon (talk) 15:53, 4 April 2008 (UTC)

It still looks like R1 returns to the plus supply not the negative supply??? UPCMaker (talk) 00:16, 5 April 2008 (UTC)

Are you still seeing the old picture, or is it not obvious that "V-" is intended to represent a negative supply? I didn't modify the image above, but made a new version, calling it a NOR instead of NAND, for the more usual polarity interpretation. Dicklyon (talk) 02:44, 5 April 2008 (UTC)

I was seeing the old picture but now I see the correct one. Thanks! UPCMaker (talk) 18:41, 5 April 2008 (UTC)

Some how the schematic got changed back. R1 is again returning to +V. UPCMaker —Preceding unsigned comment added by 206.53.104.196 (talk) 17:42, 24 October 2009 (UTC)

About the link to play-hooky.com

Dicklyon, I have no any connection with this source. I realize that it is not so reputable but it is the only source giving good intuitive RTL explanations, considerations and some calculations. This was the reason for including the link. Regards, Circuit dreamer (talk, contribs, email) 22:44, 12 October 2010 (UTC)

OK, no problem. I was afraid it was yours, after you had added it first as an EL. Dicklyon (talk) 04:49, 13 October 2010 (UTC)

Year of birth of RTL

When was RTL invented? Dicklyon on this page is mentioning that in 1958, RTL was already around. Was RTL born in that year or before?

ICE77 (talk) 23:59, 17 February 2011 (UTC)

It probably wasn't invented. There were logic circuits for a long time, and at some point people started trying to classify them. Dicklyon (talk) 00:22, 18 February 2011 (UTC)

I just ordered three cheap used books on switching circuits from 1958 and earlier; I'll let you know if I learn anything. Dicklyon (talk) 00:31, 18 February 2011 (UTC)

Please let me know if you found out anything.

ICE77 (talk) 16:33, 27 April 2011 (UTC)

I didn't find much. I think they didn't use that term at all. Some of the books concentrated more on vaccum tubes and relays and barely mentioned transistor circuits. I'll have to try to remember to look again and see if they had RTL-like circuits. Dicklyon (talk) 05:19, 28 April 2011 (UTC)

I just checked the Shea 1957 Transistor Circuit Engineering book that I got for a different purpose more recently. It has an AND gate and an OR gate made from two transistors each and some resistors, but they're completely lame and unworkable for gates as we know them (unworkable because inputs go to the bases of grounded-emitter transistors, making it very difficult to distribute a high logic level to multiple gates). This suggests that the concept of RTL wasn't quite there yet. Dicklyon (talk) 05:25, 28 April 2011 (UTC)

The oldest reference I have found is a paper by WD Rowe and GH Royer, "Transistor NOR circuit design", published in AIEE Transactions (Communications and Electronics) in July 1957. I have not read this paper but it seems to the common ancestral reference cited in several other texts that I have examined. I am half inclined to pop down the library and request a copy but inter-library loans tend to take months to be fulfilled. Crispmuncher (talk) 10:41, 2 May 2011 (UTC).

I found a paper referencing that one also references this 1956 paper: [15]; so I got a copy. Email me if you'd like me to send you a copy. It has an RTL NOR gate with speedup capacitors (a "two-input resistor gate"), which it touts as a major advance via the concept of a standard interconnectable gate. Dicklyon (talk) 15:49, 2 May 2011 (UTC)

I found Lloyd P. Hunter, Handbook of Semiconductor Electronics, Mc Graw Hill , 1956. Figures 15.40 and 15.41 show "or" and "and" gates, with one transistor per input and a speed-up capacitor around the input resistor. Logic levels in these circuits are 0 V is logic 1, and -5 V is logic 0. They use PNP transistors of a fairly unusual type, a "thyratron-like" transistor with a point contact collector junction and an alloy emitter junction. One of the design problems was aging of transistors - input currents changed with time. No indication of what computers might have used this circuit or which device number or manufacturers made this transistor type. --Wtshymanski (talk) 14:52, 7 May 2011 (UTC)

I've just got a copy of the Rowe/Royer paper (actually it came through a month a so ago, but I've only just started looking at it, since it was immediately obvious it didn't cover my primary reason for obtaining it). Interestingly it does not really focus on the design of an XOR circuit per se, but on with well characterised performance - fan in/fan out, voltage tolerances, speed and so on. In short, creating an off-the-shelf circuit that may be used under specified conditions, that can be integrated in a very free form manner without needing to tweak the configuration of one gate for the requirements of the suceeding gates.

This therefore appears to be an early example of a "standard" logic family, as opposed to documenting a gate that happens to perform a NOR function in one specific context. That seems notable enough to me but this is probably not the place for it. I may take Dicklyon up on that offer of that paper - I reciprocate if you want. I'll mail you in the next couple of days. Crispmuncher (talk) 20:13, 16 July 2011 (UTC).

Alright folks, thanks a lot. I guess we can say RTL has been around since at least 1956. That's all I wanted to know.

ICE77 (talk) 05:17, 15 May 2011 (UTC)

The whole idea of "logic families" wasn't really necessary back in the days when you had to design the whole circuit from scratch using wires, resistors, diodes and transistor (or valves/tubes). It was only when they started manufacturing ICs with pre-assembled gates that it became necessary to classify them, and by that time TTL and DTL would have been the main families. Besides, in a circuit built out of discrete components, it's arbitrary whether a resistor is part of the output of one gate or the input of the following gate, so there's no hard distinction between "output-gating" or "input-gating". 14.203.83.243 (talk) 08:10, 12 September 2016 (UTC)

RCTL

I propose creating an article for RCTL for reasons of consistency. Information about RCTL is already part of this article under the section called "Speeding up RTL". I looked at http://en.wikipedia.org/wiki/Logic_family and I noticed that most subclasses of circuits are linked to specific articles (aside from RCTL which points to RTL and therefore it's inconsistent).

ICE77 (talk) 02:27, 18 February 2011 (UTC)

Threshold logic gates

I've snipped the assertion regarding threshold logic gates since it misinterprets the source. I'm familar with the Hurst book, and yes it does give a very similar circuit to that described here, but the emphasis and intent are different. The circuit described therein is an RTL example from the DONUT computer and itself is a reimplementation of a threshold gate first advanced by SC Chao in a 1959 paper in IRE Transactions on Electronic Computers. This is not even the only form of threshold gate - the Hurst book outlines a number of different approaches in the chapter referenced.

Although any Boolean gate can by definition be considered a threshold gate (since threshold gates are a superset of Boolean gates) it misses the key elements that make threshold logic distinct from Boolean logic, namely thresholds that lie conceptually between the Boolean AND and OR operations (e.g. output true if at least three of five inputs are true, instread of simply one or more [OR], or all [AND] of them), and optionally the weighting of inputs so that some inputs count for more than others. Threshold gates therefore bear more in common with artificial neurons than conventional Boolean logic gates. Crispmuncher (talk) 15:44, 21 April 2011 (UTC)

Obviously, Boolean logic gates are special case of threshold gates but yet they are threshold gates. I intend to develop this idea in the introductory part of Logic gate to show what the fundamental idea behind logic gates is and how it is implemented by various electronic components. Circuit dreamer (talk, contribs, email) 21:32, 21 April 2011 (UTC)

I see that threshold gate is a redlink in the TC0 article, and threshold logic is a redlink in the artificial neural network article.

Perhaps we should make an article at one, and make the other redirect to it. Or is there maybe already an article on this topic, and both of those terms should redirect to it? --DavidCary (talk) 23:50, 12 July 2015 (UTC)

Use the ASCII hyphen "-" instead of the Em Dash "– " (U+2013 ie. 0xe28093) for the page title ?

I think it would be better to use the ASCII hyphen "-" (0x2d) instead of the Em Dash "–" (U+2013 ie. 0xe28093) in the page title.

This would make links to this page look cleaner, ie. https://en.wikipedia.org/wiki/Resistor-transistor_logic instead of the current https://en.wikipedia.org/wiki/Resistor%E2%80%93transistor_logic

Currently, this page is at https://en.wikipedia.org/wiki/Resistor%E2%80%93transistor_logic , and https://en.wikipedia.org/wiki/Resistor-transistor_logic redirects to https://en.wikipedia.org/wiki/Resistor%E2%80%93transistor_logic .

— Preceding unsigned comment added by PeterBehnam (talk • contribs) 08:52, 4 February 2022 (UTC)

Don't, per MOS:DASH. • Sbmeirow • Talk • 10:24, 5 February 2022 (UTC)