Close

MPSHxx Teaser: would you believe 500ps?

A project log for CBJT Logic

Complementary bipolar junction transistor logic - like CMOS, but with NPNs and PNPs

Ted YapoTed Yapo 01/20/2017 at 18:1817 Comments

I shouldn't post this yet; there's surely some catch that will prevent you from going this fast. But, I'm doing it anyway. One of my favorite NPN/PNP pairs are the MPSH10 / MPSH81 RF transistors. Obsolete in TO92 now (although I have a thousand or so kicking around), they're still available in SOT23 as MMBTH10 / MMBTH81. With an ft of 600+ MHz, this pair is twice as fast as my other "fast" favorites, the 2N3904/2N3906 (300 MHz ft). Anyway, I couldn't help but simulate some CBJT inverters with the RF pair. Here's the input to, and output from a chain of 10 inverters. One edge is delayed by 3.6ns, the other by 5.0 ns. That equates to a propagation delay of under 500ps per inverter. Pretty fast.

Here's the circuit. It might be difficult to find in the parameters, but the input resistor is 5.1k, the speedup capacitor 47p, and the feedback resistor 15k. The power supply is 1.2V.

If you want to play around in LTspice with this, I've uploaded the simulation file. I haven't even simulated a 2-input gate yet, so that may or may not work. I also haven't built one of these yet. Even testing it may provide some unique challenges.

I am still working on choosing the R's and C's to optimize the 2N3904/2N3906 pair. I started with a 1k input resistor because I have a giant reel of them from a late-night ebay purchase years ago. This is perhaps not the best objective function for optimizing the design :-) So, I'm having a close look at the R's and C's, considering speed, power, fan-out, etc, and then testing on hardware to make sure the simulations didn't lie. It's going to take a little while longer.

UPDATE

I ran a 5-inverter ring oscillator simulation with these transistors. This makes it look like actual propagation delays are closer to 1 ns. I still have no idea if these speeds will be achievable in reality.

Discussions

David H Haffner Sr wrote 01/21/2017 at 06:47 point

Hey Ted, you may not be far off the mark, there is a family of ultra-high speed devices that has been designed to maximize the performance of digital systems. there is a family based on MSI parts developed on the high-Plessey Process III(V); these circuits offer local gate delays of 500 picoseconds and flip-flop... more

  Are you sure? yes | no

Yann Guidon / YGDES wrote 01/21/2017 at 04:59 point

My latest thought about CBJT power :

Most of the voltage references work at 1.25V (LM317 for example) but CBJT works best around 1.1V.

Why not "drop" the remaining 150mV with a resistor ?

The complementary transistor pair would be powered by a RC network (R around 10 Ohms, C around 1 to 10nF ?)

Damn, I wish I can finish renovating my workshop really soon because I'm so frustrated of being unable to test my hypotheses :-/

  Are you sure? yes | no

Ted Yapo wrote 01/21/2017 at 05:07 point

I have bumped the latest ones (with feedback) to 1.2V, so 1.25 might be OK - writing up now.  I have also simulated a power supply with an LT3080 regulator (which uses a current reference so can go down to 0V) and an PNP and NPN diode-connected as a reference.  This way, the supply voltage tracks the Vf change with ambient temperature, which is probably wise, and may be a necessity.

I like the 10nF/pair idea.  I have used 470nF/pair on prototypes, but that's mostly because I have an ebay'ed reel of them :-)  Dropping with local resistors would also provide noise immunity.

But with symmetrical logic, which rail is the reference?  It could be either one!  Maybe a 5 ohm resistor on each side :-P

I have put off renovating/de-cluttering/cleaning my workshop for over a year now because once I start, it's going to take months...

  Are you sure? yes | no

Yann Guidon / YGDES wrote 01/21/2017 at 05:17 point

More voltage might also decrease the speed, yet consume much more power...  but working at 1.25V would be awesome.

What's your NPN/PNP reference circuit/schematic ?

470nF is probably too high, 10nF is used/preferred as local decoupling for very high speed circuits. 100nF is welcome every few gates too :-)

The reference could be either one but I'd say 0V :-D And yes I've thought about a symmetrical drop but 1) that doubles the parts count 2) you'll need a capacitor on the low side as well 3) we can choose to "drop" the side that has the highest slew rate to make a more symmetrical waveform and reduce the EMI...

I've waited for 11 years to renovate my dear, life-saving workshop so I'll be patient and dedicated. Next month, it will be such a pleasure to use it !

  Are you sure? yes | no

Ted Yapo wrote 01/21/2017 at 05:32 point

Q1 and Q2 are the reference/sensor pair.  R1 can be adjusted to raise/lower the supply below the conduction point of the powered circuit pairs.  The plot shows the regulator output (yellow) over -20 to 100C.  Q4 and Q3 are a test here - the cyan line is the voltage they need to self-conduct 10 mA. 

Oh, I forgot the required minimum load on the output - the original simulation was on my laptop. Add  a resistor across the output to draw at least 10 mA - like 100 ohms.

The default PNP/NPN in SPICE are silly devices; 2N390x in there give around 1.2V.

  Are you sure? yes | no

K.C. Lee wrote 01/21/2017 at 14:21 point

For the purist that wants a discrete solution:

You can use the discrete regulator with the reference pair + a diode out of NPN junction or CR1.  

Q1 is wired as a voltage follower, so it'll need an additional VBE to raise the output back.  You can get a better match if you biased this diode separately as Q1 B-E draw a lower current.

  Are you sure? yes | no

Ted Yapo wrote 01/21/2017 at 16:30 point

@K.C. Lee I like the discrete/purist solution!  I tend to forget purism with power supplies - the built-in current limiting and temperature cutouts are nice to have, and difficult to reproduce economically - but you could say similar things about commercial logic gates, I guess :-)

  Are you sure? yes | no

K.C. Lee wrote 01/21/2017 at 22:18 point

I don't recommend doing a supply that would be subjected to external faults with discrete.  It can be done, but really not worth the effort.  My bench supply entry uses the LM317 for that reason. 

The purist part was poking fun at @Yann Guidon.  :)

  Are you sure? yes | no

Yann Guidon / YGDES wrote 01/22/2017 at 02:04 point

@K.C. Lee hmmm was that a poke ? It looked like candy to me ^_^

  Are you sure? yes | no

Ted Yapo wrote 01/21/2017 at 18:30 point

What about using LTspice during workshop down-time?

  Are you sure? yes | no

Yann Guidon / YGDES wrote 01/21/2017 at 19:56 point

"I love the smell of solder in the morning" ;-)

you can't beat actually rumaging through the parts bins !

  Are you sure? yes | no

Yann Guidon / YGDES wrote 01/20/2017 at 23:50 point

I want to believe...

https://hackaday.io/project/19304-ygrec-si/log/51479-inventory : 1500 MPSH10 but I didn't know the PNP equivalent, time for ebaying...

  Are you sure? yes | no

Ted Yapo wrote 01/21/2017 at 02:22 point

1500 MPSH81's will be very expensive :-(

  Are you sure? yes | no

Yann Guidon / YGDES wrote 01/21/2017 at 02:40 point

Maybe.

  Are you sure? yes | no

Yann Guidon / YGDES wrote 01/21/2017 at 03:56 point

OOooops they're all gone...
http://www.ebay.com/itm/272261944982
sorry not sorry.

  Are you sure? yes | no

Ted Yapo wrote 01/21/2017 at 04:14 point

How did I not see those??  Nice score!

  Are you sure? yes | no

Yann Guidon / YGDES wrote 01/21/2017 at 04:29 point

these would not "match" the TO92 MPSH10 I already have but 9K transistors would make a nice little SOT23-based ECL computer ;-)
Unless I find a similar stock of MMBTH10 at the same price. I've seen some reels but not enough or too expensive...

I'll have to compare practical implementations of logic gates, ECL vs CBJT: speed, number of parts for each basic gate... I'm mostly concerned by the cost of a MUX2, MUX4 and Latch. Actually, everything else can be made with MUX2.

  Are you sure? yes | no