How to drive 20V digital signals?

[ebastler]

For an interface to a vintage computer, I need to drive eight digital signals, with signal levels 0V and 20V. (I also need to provide the complementary signals, so a driver that does both, like the venerable 4041, would be a plus.) Required output source/sink current is just a couple of mA per channel. The signals originate from a 3.3V microcontroller, so what I am looking for is essentially a level translator.

My prototype uses its own 20V supply, two ULN2803 darlington arrays to pull the outputs to ground, and 2k2 pullup resistors to bring them back up to 20V quickly enough. (Up to a microsecond of risetime is OK.) This works, but needs close to 80 mA for the pullups alone -- at any time there are eight positive or inverted signals pulled to ground against the pullup resistors' will.

I would like to power the final design from the vintage computer's built-in 20V supply, which is wimpy. So I am looking for a lower-power solution -- no more pullups, but proper push/pull output stages. Here's what I figured so far:

• Considered a two-chip solution, adding a UDN2981 (high-side switch) to the ULN2803 (low-side switch) I already use. But the voltage drop of the UDN2981 is uncomfortably high at nearly 2V.
  
• I had already mentioned the good old 4041 CMOS chip above. Two of these would conveniently drive the 8 channels and the inverted ones -- but operation at 20V would be marginal, and it lacks the level translation capability.
• I would have thought there are plenty of H-bridge or half-bridge arrays out there, to get eight push/pull output stages in one or two chips. But I have not found anything convincing so far. Again, low voltage drop is required for the output stages; I need to get "close" to 0V and 20V.  Say 1V drop at either end at most, less would be better.
  

Any ideas? Did I overlook some nice quad or octal drivers/bridges, preferably with low-drop MOS output stages? Other simple circuit ideas? Thank you very much in advance!

[RoGeorge]

What vintage interface/computer has had a 20V digital signals?

[ebastler]

What vintage interface/computer has had a 20V digital signals?

The LGP-30 from Librascope/General Precision. Tube flip-flops and inverters operating at a couple 100V, but the logic signals are clipped to -20V (logic low) and 0V (logic high). I have tied my microcontroller's ground to the -20V, to avoid the inconvenient negative voltage.

[amyk]

Again, low voltage drop is required for the output stages; I need to get "close" to 0V and 20V.  Say 1V drop at either end at most, less would be better.[/li]
[/list]
Any ideas? Did I overlook some nice quad or octal drivers/bridges, preferably with low-drop MOS output stages? Other simple circuit ideas? Thank you very much in advance!

Give it a 22-23V supply, and lower your MCU ground a few volts too if needed. Now you can drive to (or even slightly above) 20V.

Trying to save a few watts when you're interfacing with a tube computer that draws 1.5kW seems futile, I would just go with the pullups if they're fast enough, and make sure their power dissipation is within limits.

[DaJMasta]

I'd be tempted to try to increase the pullups to see how slow it could actually work at, but you could also try a high side equivalent of the darlington array, and it seems like there are fewer but something like the UDN2981 could work.  You could also do it with discrete p channel fets, if needed, or go the standard CMOS route and use pairs of fets for totem pole output.


If you have the ability, you can try to reduce capacitance on your drive lines - even small decreases would speed up your rise time significantly with your current pullup solution.




Though if you can tolerate 1V on either end and want to get crafty........ make yourself a virtual ground at 1V, then run your 4041 at 19V supply 

[RoGeorge]

The LGP-30 from Librascope/General Precision. Tube flip-flops and inverters operating at a couple 100V, but the logic signals are clipped to -20V (logic low) and 0V (logic high).

Never seen one, thanks!   

1488/1489 were RS232 buffer chips, but only +/-12V.
Maybe some generic quadruple op amps that can support more than 20V, used as comparators.

[Zero999]

There's the CD40106, which can be used to translate from 3.3V to 20V, although, as you've already said, operating at such a high voltage is marginal design, although you might be able to get away with 18V
http://www.ti.com/lit/ds/symlink/cd40106b.pdf

How about using a MOSFET driver?

Flicking through RS components, the IR4428 seems like it would be a close fit.
https://docs-emea.rs-online.com/webdocs/0791/0900766b807910c2.pdf

[ebastler]

Wow, thank you for many creative ideas in a short time! I'll comment on only a couple of them right now, will need a closer look...

Trying to save a few watts when you're interfacing with a tube computer that draws 1.5kW seems futile

I agree. It's not about saving power, but about keeping my modern-day addition to the LGP-30 minimally invasive. I would prefer not to hook up an extra power supply inside the machine, but just connect the interface board and power it from the voltages which are available inside the LGP-30. And its -20V line more like a voltage reference than a power supply -- used to clamp the logic levels inside the original computer. It's actually derived via a big series resistor in the return path of the -160V supply...

There's the CD40106, which can be used to translate from 3.3V to 20V,

Hmm, how would I do that? Doesn't its input threshold increase with the operating (and hence output) voltage, and always hovers around VDD/2 (+- hysteresis)?

How about using a MOSFET driver?
Flicking through RS components, the IR4428 seems like it would be a close fit.
https://docs-emea.rs-online.com/webdocs/0791/0900766b807910c2.pdf

That's a great idea! While I had hoped for a slightly higher level of integration, the IR4428 is quite convenient with its pair of normal and inverting drivers. Unfortunately it is obsolete and seems quite hard to find. [See edit below!] But you have given me a new search term; hadn't thought of looking for low-side MOSFET drivers before. Thank you!

Edit: The IRS4428 is alive and well! I will take a closer look, but it looks promising indeed.

[FenTiger]

VFD driver?

https://www.maximintegrated.com/en/products/power/display-power-control/MAX6921.html

It'll take some software work on the non-legacy side.

[Zero999]

There's the CD40106, which can be used to translate from 3.3V to 20V,

Hmm, how would I do that? Doesn't its input threshold increase with the operating (and hence output) voltage, and always hovers around VDD/2 (+- hysteresis)?

It doesn't. I got the wrong part number. I should have said the CD4504, but it will require a 5V supply.
https://www.ti.com/lit/ds/symlink/cd4504b-ep.pdf

How about using a MOSFET driver?
Flicking through RS components, the IR4428 seems like it would be a close fit.
https://docs-emea.rs-online.com/webdocs/0791/0900766b807910c2.pdf

That's a great idea! While I had hoped for a slightly higher level of integration, the IR4428 is quite convenient with its pair of normal and inverting drivers. Unfortunately it is obsolete and seems quite hard to find. [See edit below!] But you have given me a new search term; hadn't thought of looking for low-side MOSFET drivers before. Thank you!

Edit: The IRS4428 is alive and well! I will take a closer look, but it looks promising indeed.

I haven't read the IRS4428 data sheet in derail, but at first glance it seems to be an updated IR442, which they're phasing out. No doubt there are other parts. I did look at the Microchip range but they're limited to 18V.

[ebastler]

VFD driver?
https://www.maximintegrated.com/en/products/power/display-power-control/MAX6921.html
It'll take some software work on the non-legacy side.

Hmm, that's an interesting thought actually. I had come across VFD drivers in my search, but always skipped them when I saw the SPI interface. But come to think of it: I need to update my outputs at approx. 130 kHz, so I have a few µs to send the contents. There is only one fast bit which I need to set more quickly in response to new inputs from the LGP-30, but I could deal with that one separately.

The rise times and propagation delays of the MAX6921 are still marginal, but I will have a close look at that chip and other VFD drivers. Thanks!

[SiliconWizard]

What vintage interface/computer has had a 20V digital signals?

The LGP-30 from Librascope/General Precision.

https://en.wikipedia.org/wiki/LGP-30

Damn. Even @legacy will appear to work with cutting-edge technology now...

120 kHz, 1500 W. Hehe.

Reading the "Starting the machine" was entertaining too.

[ebastler]

https://en.wikipedia.org/wiki/LGP-30
Damn. Even @legacy will appear to work with cutting-edge technology now...
120 kHz, 1500 W. Hehe.
Reading the "Starting the machine" was entertaining too.

Yes, great fun! 

The LGP-30 has an amazingly tight architecture. The actual CPU only uses 24 tubes! Bit-serial  operation, tightly integrated with a magnetic drum, which serves as the working memory and provides some clock and timing signals. All designed by one man, Stan Frenkel, and it really shows in how close all the components are intermeshed.

I originally dug into this building an FPGA replica: http://e-basteln.de/lgp30/.
But now I get to play with and restore a real one here: https://www.technikum29.de/en/computer/lgp30

[SiliconWizard]

The LGP-30 has an amazingly tight architecture. The actual CPU only uses 24 tubes!

Yes, I was amazed to see the specs and then look at the size of the machine. Even the power draw was still not that bad for such a machine...

[ebastler]

Yes, I was amazed to see the specs and then look at the size of the machine. Even the power draw was still not that bad for such a machine...

It sits on casters, and just needs a regular wall outlet for power. The first "mobile" computer which you could roll over to your desk! 

[ebastler]

Quick update on my search, after having done a bit more homework:

VFD drivers seem to come with a significant output voltage drop as well, it seems. E.g. the MAX6921 outputs stay at least 2V below the supply voltage. Not critical for driving VFDs, but unfortunately a problem for me. Too bad, as the form factor would be nice, and SPI control should work for my application.

Low-side drivers for MOSFETs seem like the most promising route. The IRS4428 may not be the ideal candidate, since it also has a 1.4V output drop, and suspiciously its current consumption is not specified. But the NCP81071C (ON Semi) is very similar, with a manageable supply current below 2 mA, and the 2EDN7524/7523 (Infineon) offers two either direct or inverting drivers, with approx. 1mA supply current. All of these are specified up to 20V, matching my needs nicely.

Time to order some samples, I guess... Thank you all for your input, and special thanks to Zero999 for the low-side driver idea!