More On Vacuum Tube Models & SPICE
I spent a good part of the weekend trying to sort out the various worksheets, software, Spice models used for vacuum tubes. It was pretty messy… As there are numerous ways to do the same thing and I have already collect a lot of material over the past year or so. The goal was to distill the process down to a few simple steps and to nail down the simulation platform.
Let’s get the simple thing out of the way – picking the simulation platform. While most of the online Spice models and discussions on simulations tend to gravitate to the older PSpice or the newer LTSpice, neither were to my liking after spending time with SIMetrix, so there is just no going back to those platforms. Now the task before me, was to gather tube models specifically for SIMetrix, which wasn’t really that difficult, since it can already import various models written in several popular Spice languages. Not only that, I also discovered a complete SIMetrix-ready tube library at the great Ayumi’s Lab, thanks Nakabayashi-Son!
Of course, there are always going to be tubes that are not included in these all-in-one libraries, especially when I tend to use a lot of Chinese and Russian tubes, so the need to generate individual tube models based on datasheets is a dreaded but necessary task.
As discussed in an earlier blog post, I have developed the process for this task, but it is worth repeating here, as there have been some updates:
1) data capture – I use Andrei Frolov’s CurveCaptor to quickly capture the basic curve data from scanned datasheets, there are many other programs that can do the same thing, but this one is preferred because it can also generate triode models directly as we will see shortly;
2) modeling the triode – CurveCaptor is specifically designed for this task, so this is what I use to get the basic Koren or Rydel parameters, but these models do not include grid current parameters, so they will need to be added in the step below;
3) triode grid current parameters – since guitar amps often push the input signal of the tubes beyond typical “hi-fi” operating limits, the grid current needs to be modeled, as it affects the clipping and compression characteristics of the tube thus the amplifier. There are a couple of ways to do this, such as a simple diode model or using the more complex models proposed by Hutt and Sjursens, a good discussion is available at AX84.
4) modeling the pentode – this is a tough one, but I managed to get a copy of Robert McLean’s worksheet which makes the task easier, I think Robert’s model is a definite improvement over Koren’s version. An alternative is to use Eureqa Formulize to automatically generate the parameters based on the pentode characteristic curves.
During my research, I also discovered quite a few good sources of information that I have not seen before, these fine gentlemen have clearly gone through the modeling process in great detail as reflected in their write-ups, programs and worksheets. I also had some short email exchanges with them, they are both great and I want to give my thanks to them again here.
David Heiser’s site contains a treasure trove of information, he does a thorough review of the methodologies used by the past practitioners, and provides great insights to the short-comings as well as suggestions for improvement on those methodologies. This is a must read for anyone that’s interested in the subject.
Bill Elliott’s site contains a very detail discussion on the Koren’s models with a heavy mathematics bend. Bill also provides a novel methodology to extract key tube parameters from the datasheet with as little as 5 data points. For those that have the Derive 6 software, they can also use Bill’s program to solve the Koren parameters automatically.
As usual, AX84 proved to be a great source of information, as many experienced amp designers/DIYers frequently use Spice in their work, and searching the site always turns up some useful information on the subject.