Tube Cad / Spice Questions

[bob p]

I thought I'd start a thread for this since I wasn't getting any hits in my Regulated Power Supply thread, and the search function didn't turn up much helpful information.

I'd like to try out some of the tube amp design / modeling software. I'm hoping that I can find something that will help me get close to a useful design on paper (OK, on a PC) before I have to build the circuit for real.

I've been looking at TubeCad on their web site, and I get the impression that its primarily designed for triodes. At least I got that impression, as it appeared that triode data came with the software, but not pentode data. Please correct me if I'm wrong on this, or if there are other sources to get the necessary data files.

If anyone could give me a rundown on using something like Tube Cad or Spice for modeling a complete push-pull amplifier circuit, I'd really appreciate hearing about your experience.

TIA

[bob p]

Bump.

If my objectives listed in the first post were too optomistic, I'd be happy to model one stage of an amp at a time. If the software is capable of handling pentodes that would be great. UL hookups would be even better.

Is ANYONE here familiar with ANY tube modeling software? I'm not afraid to pay for something that does what I need it to do.

Thanks.

[Steve Conner]

It looks like Duncan Munro has made triode and pentode models for LTSpice:

http://www.duncanamps.com/technical/ltspice.html
http://www.duncanamps.com/spicevalves.html

LTSpice is free, and pretty good as SPICE simulators go. I've used it a lot for both business and hobby projects. I refuse to design tube circuits with it, though. It's like Hamlet wearing a Casio watch ;-)

[bob p]

Thanks for posting, Steve.

I understand completely about the aesthetic appeal of using CAD simulations for tube amps, but after spending some time trying to draw out push-pull composite load lines by hand, I figure there has got be a better way to do this!

I'm so new to the whole modeling concept that I don't know the difference between all of the different flavors of Spice. for example, I don't know what the difference is between the two spice variants listed on Duncan's page: OrCad's Pspice and Berkeley Spice 3f4.

Is LTSpice close to either one of these, and if so, which sets of files on Duncan's site are "compatible" with LTSpice?

FWIW, I've already downloaded LTSpice for Windows <grumble>, and I'm trying to figure out how it works. Is there online documentation somewhere? If so, I haven't been able to find it yet...

Thanks again!

[Steve Conner]

LTSpice is compatible with PSpice, as far as I know, so you can use the PSpice models that Duncan provides.

If you haven't used a PSpice style simulator before, there will be a learning curve. We were taught how to use PSpice at uni (a process that took several two-hour lab sessions with a tutor) and LTSpice is much the same. I seem to remember it has decent online help accessible through the Help menu, but it may not be much use if you're not familiar with simulation concepts in general. I think it will also run on Linux under WINE, if that helps :P

As far as composite load lines: If you have adjustable bias and/or screen voltage, and a generously sized OT with 4, 8 and 16 taps, then surely you don't need to choose your load line at the design stage. You can pick it with a scope for lowest THD once you're finished, or just choose the load line that sounds best to you. Or gives the best compromise between sound and room heating, in your case :-)

When you consider how much the impedance of a speaker varies over its frequency range, and how the little kinks and knees in the characteristic vary between different brands and batches of tubes, I think life's just too short for simulation. How's your load line going to look when you hit the bass resonance of the speaker and the impedance jumps up from 8 ohms to 50?

This is just my own personal opinion, as I program computers for a day job, and play with tubes to get away from them...

[bob p]

LTSpice is compatible with PSpice, as far as I know, so you can use the PSpice models that Duncan provides.

If you haven't used a PSpice style simulator before, there will be a learning curve. We were taught how to use PSpice at uni (a process that took several two-hour lab sessions with a tutor) and LTSpice is much the same. I seem to remember it has decent online help accessible through the Help menu, but it may not be much use if you're not familiar with simulation concepts in general. I think it will also run on Linux under WINE, if that helps :P

Yeah Steve, I certainly understand your point about using tubes to get away from the monotony that comes with being a codehead. And you're definitely right about the learning curve -- I opened up LTSpice and the first words out of my mouth were, "Duh, What in the hell am I supposed to do now?"

Obviously, its going to take some time to figure out how to get up and running with Spice, and a bit more time to figure out how to model tube circuits.

Even though I'm in a pretty dumbfounded state right now, I'm still encouraged by the fact that some people have developed Spice models for vacuum tubes, and that they're sharing them. The missing link (at least for me right now) is how to use them. It would really help if I could find some simple tube circuits that go along with the program, so I can get an idea how this thing is supposed to work. Even a model of a single triode stage would be really helpful.

Regarding the whole Windows/Linux thing -- I swore off against using new software that tied me to the Windows platform several years ago -- primarily for economic reasons. I just don't like the world's richest man and his co-marketers unnecessarily reaching into my pocket every time that I need a software upgrade. I've been making a conscious effort to deploy open-source software solutions whenever I can. I've been successful enough that Linux is now my primary OS, and I only use Windows when my hand is forced in one way or another. Insofar as LTSpice is Windows only, and it seems to have a good support community, I guess that this means that I'll be using this app on Windows -- at least for now that seems to be the path of least resistance. Though I prefer linux I'm not really a rabid Anti-Windows zealot. Maybe once I'm familiar with Spice the move to linux will be easier.

[bob p]

As far as composite load lines: If you have adjustable bias and/or screen voltage, and a generously sized OT with 4, 8 and 16 taps, then surely you don't need to choose your load line at the design stage. You can pick it with a scope for lowest THD once you're finished, or just choose the load line that sounds best to you. Or gives the best compromise between sound and room heating, in your case :-)

Well, part of the reason that I want to use loadlines in the design stage is because I've never before used the rigorous approach of doing all of the composite loadline stuff in the design stage. In the past I've always stuck to the easy way out of following someone else's schematic.

In this case (the 100W Class A1 amp project), I don't really have someone else's schematic to follow, so the entire amp design process has been an exercise in number crunching. I know that you've been following the other threads, so it won't really surprise you to think about how much number crunching has been done. What continues to amaze me about this project is that a) there's been so much theoretical consideration in the design phase, b) I don't even have most of the parts yet, and c) after all this work, I still haven't even touched the soldering iron!

Part of the allure of the simulation idea for me is that it's supposed to eliminate a lot of the need for trial and error assembly and testing, and theoretically it will allow me to crunch numbers to the point that the numbers put me in the ballpark before I have to make the move to experimentation by soldering. At least that's how its supposed to work in theory. The whole modeling concept does seem like it could be a tremendous labor saving device once you're up and running. The whole process of drawing composite loadlines by hand is a royal PITA, and I'd like to eliminate that drudgery if its possible.

Naturally, the first time that you do this sort of thing its likely to take more time learning to use the computer than it would have taken to use a soldering iron and move ahead by trial and error. I guess that's always going to be true the first time that you try modeling. The real payoff comes along at some point in the future. I hope.

[bob p]

When you consider how much the impedance of a speaker varies over its frequency range, and how the little kinks and knees in the characteristic vary between different brands and batches of tubes, I think life's just too short for simulation. How's your load line going to look when you hit the bass resonance of the speaker and the impedance jumps up from 8 ohms to 50?

Good point. As far as variation from tube to tube is concerned, I'm hoping that when I'm using 3 push-pull pairs of tubes for each channel, the law of averages will work in my favor, so that the composite characteristics of the bunch of tubes will be closer to that of a "bogey" tube. We'll have to see how that works out.

One of the "problems" I'm dealing with now is one that's sort of related to feature creep. I've just acquired a pair of speakers that presents a rather odd load to the amp that's used to drive them -- although they're rated at 4 ohms, in reality they have an impedance of 3 ohms or less over most of the audio spectrum. To use them with off the shelf amps, they're pretty demanding in that they require an amp that can handle delivering lots of power into a low Z load. To build a tube amp that can drive them properly, I think that some intentional impedance "mismatching" might be required, so that the amp would have taps of 2/4/8 instead of 4/8/16.

There's another reason that I'd really like to do modeling -- Although I guess its easy enough to choose the operating conditions based on the values in the 6550A / KT88 tube charts, then there's the problem of driving them.

I've been thinking that 3 pair of 6550 would present a fairly low Z load for the driver circuit, and I'd probably need to use a CF driver after the PI to drive them. I have to admit that I've never seen any schems where anyone has actually done this to drive 3 pair of tubes rather than just 1 pair. Essentially, I'm designing the amp stage by stage from the power section back toward the input. Its a big project, and I can use all of the help I can get.

[Steve Conner]

Hi Bob,

Duncan has some example models of whole amplifiers on his site. I think they were on one of the pages I linked. You should be able to load these directly in LTSpice, since that's what he used to create them.

Ken Gilbert's BAGA used cathode followers to drive three pairs of KT90s. He has full schematics online. The Genelex 400 watt KT88 application circuit is also relevant: http://www.freeinfosociety.com/elect...ew.php?id=1152

The requirement for grid drive really depends on whether you intend to wander into Class AB2 or not. If you're going to stay in AB1, tube grids are very easy to drive, they look like an almost open circuit with negligible capacitance. For pentodes that is: power triodes have large Miller capacitance.

But with multiple paralleled tubes, the grid leak resistance needed to stop thermal runaway is pretty low. You more or less need those cathode followers to drive the grid leak resistors, and the tube grids just come along for the ride.

Class AB2 is a whole different ballgame. IMO, the interstage transformer takes a lot of beating for this, until you try to put feedback around it

FWIW: I'm split about 50/50 between Ubuntu and XP at the moment. I find Linux way better for l33t hacker toolz, but Windows has stuff like Ableton, Sound Forge, Pro Tools, and Photoshop CS2... I just got a new job where I'll be developing on Windows CE using Visual Studio.

[bob p]

Thanks Steve. It sounds like there's more info on Duncan's site than I was aware of.

How did I manage to miss the fact that KG's BAGA used 6BQ5 as CF drivers??? How did I manage to forget about thsoe big 400W amps in the appendix of the GEC book??? Thanks for the reminders.

For my purposes, I prefer the "1" subclasses of amplifier biasing, not the "2" subclasses. We'll see how it goes...

[Earl]

I use an 'older' version of PSpice...It's the MicroSim Eval Version 8. It used to be given as part of a college class, and is still out there on the net for free.

This version is much more user friendly than the, say, OrCad versions...and works perfectly on my W98 machine.

It's also much easier to import a library, than a individual model, and http://www.normankoren.com/Audio/Tubemodelinstr.HTM has a couple of libraries with various tubes in them for free.

I've modeled a full BF Champ (only with a 6L6) and it works pretty dang good.

I've run Plate Characteristic curves of typical triodes, and get almost identical resulsts as the published data. With that capability, it's relatively easy to do Load Line graphs. Here's an example, not too clear - http://home.texoma.net/~flhh/x7pccs.jpg

As mentioned, there is a learning curve, and the Eval versions aren't loaded with full blown help files.

Regarding speakers - there are models of spekers on the net...I've got a good one for the V30. You can just build it from the schematical speaker models out there if you want to.

[Glasgow]

Bob,

I downloaded OrCad PSpice Demo Version 9.1 (Student PSpice) from:
http://www.electronics-lab.com/downloads/schematic/013/

I also downloaded the PSPICE.PDF from the same page and have been going through the tutorial.

This is cool. The software is totally late 90's, so you expirience the whole MDI phenomenon, but it works really well. The DOC seems to be for the full version, so you'll see a few missing libraries, but nothing of concern.

Even if I didn't end up using this much for building amps -- though I probably would -- it is very interesting being able to simulate circuits and I think this will prove to be a valuable learning tool.

BrianB