xformer object

Hey, sorry for the delay on responding to this, and thanks for asking about xformer. xformer is a strange module, and I agree it’s not well documented – neither in the official documentation, nor within our internal notes. We are actively doing R&D for a newer, more suitable module to accomplish transformer-style saturation with clearer documentation/controls/sonic output, though I can’t comment on when that might be released to the public. There aren’t any sigfiles you can view which use the xformer module.

For the time being, I recommend you keep area set constant to 1 and saturation set to -1. Then n1 becomes the number of primary turns (lower turns = the system saturates quicker, and the volume increases) and ratio can be used to match the output level (higher ratio = louder output).

A quick technical note: xformer is only valid for a sample rate of 48k. Weird things can happen when you move off of 48k. We are actively working to fix this.

sadguitarius wrote:

Ok just so it doesn’t look like I didn’t do any searching on my own, I did find the Godlike Productions transformer saturator algorithm as well as the documentation in the unofficial module reference. His implementation is more or less the same as what I was doing in testing out the parameters of the object, but I still feel like it’s missing some guardrails in order to make it musically useful. I’m wondering if there is a sweet spot of parameter ranges or an intended use for the object. Are there any official Eventide algorithms that use it?

I tend to leave the guardrails off unless I’m chasing a specific feature set for 2 main reasons.  Firstly my own music (techo/trance/industrial) often gets tones from pushing things to the edge, or sometimes over the edge, and to this end, I wanted the ability to use the algorithm to distort.  Pretty and musical is very much in the ears of the artist or audience, secondly I don’t want to make those decisions for other artists, though having said that, I do understand the value of something that always sounds good, and constraining the UI around that.  I’m happy to put some guardrails in place for a version of the transformer saturator if you can provide some guidance of where you’d like it constrained, though as you are delving into VSIG, you could probably customize it for your own particular needs.

My own observations in working with this block are that area tends to impact the low frequency part of the signal.  Larger areas tended to give better low frequency handling.  Also, it depends a little bit on the application.  Like physical circuits, transformers have a number of uses, from pure amplification (voltage step up and step down), to non-linear response (pushing into saturation), phase rotation (power factor correction for capacitative loads, though normally it’s inductive loads that need correcting), filtering (LC, LR and LCR filters), and probably not relevant to VSIG, isolation.  Increasing turns (with a fixed ratio), tends to increase “power” handling, though you can drive it pretty hard before you get audible distortion.  My algorithm needs input gain up around 40dB to distort (though if you reduce area and turn to very low values, it can distort at much lower levels).

Looking a the documentation in VSIG, the xformer block at its heart seems to consist of an integrator with flux passed to a differentiator (which makes sense). One approach to modelling transformers requires consideration of the voltage and amperage in the 2 coils, the core leakage and an assumption about the output load.  I’m not sure what assumptions they have made on core losses or output load. If you want to reduce audible distortion, don’t put too much gain into the input, keep the area well above 1 (probably 5+) and turns above 1000.  Saturation level will depend on if you want a clean amp (then set this to 1), or if you want colour (then set this lower).

Hope this provides some insight.

ndeshpande wrote:

Hey, sorry for the delay on responding to this, and thanks for asking about xformer. xformer is a strange module, and I agree it’s not well documented – neither in the official documentation, nor within our internal notes. We are actively doing R&D for a newer, more suitable module to accomplish transformer-style saturation with clearer documentation/controls/sonic output, though I can’t comment on when that might be released to the public. There aren’t any sigfiles you can view which use the xformer module. For the time being, I recommend you keep area set constant to 1 and saturation set to -1. Then n1 becomes the number of primary turns (lower turns = the system saturates quicker, and the volume increases) and ratio can be used to match the output level (higher ratio = louder output). A quick technical note: xformer is only valid for a sample rate of 48k. Weird things can happen when you move off of 48k. We are actively working to fix this.

There is an interesting and computationally efficient approach to transformer saturation modelling here – https://www.ntnu.edu/documents/1001201110/1266017954/DAFx-15_submission_68.pdf.  If I have some free time I might try to implement it in VSIG or RNBO

Puppeteer wrote:

My own observations in working with this block are that area tends to impact the low frequency part of the signal. Larger areas tended to give better low frequency handling. Also, it depends a little bit on the application. Like physical circuits, transformers have a number of uses, from pure amplification (voltage step up and step down), to non-linear response (pushing into saturation), phase rotation (power factor correction for capacitative loads, though normally it’s inductive loads that need correcting), filtering (LC, LR and LCR filters), and probably not relevant to VSIG, isolation. Increasing turns (with a fixed ratio), tends to increase “power” handling, though you can drive it pretty hard before you get audible distortion. My algorithm needs input gain up around 40dB to distort (though if you reduce area and turn to very low values, it can distort at much lower levels).

Looking a the documentation in VSIG, the xformer block at its heart seems to consist of an integrator with flux passed to a differentiator (which makes sense). One approach to modelling transformers requires consideration of the voltage and amperage in the 2 coils, the core leakage and an assumption about the output load. I’m not sure what assumptions they have made on core losses or output load. If you want to reduce audible distortion, don’t put too much gain into the input, keep the area well above 1 (probably 5+) and turns above 1000. Saturation level will depend on if you want a clean amp (then set this to 1), or if you want colour (then set this lower).

Your guesses are good guesses – but they don’t accurately reflect what’s happening under the hood. The documentation is obtuse, xformer’s lineage is actually quite different from most of our other modules. What’s really frustrating is that the controls don’t always do what they say they do in Vsig’s documentation.

I heavily encourage y’all to sit tight, we are actively looking at the xformer module, its control structure, potential uses, alternative saturation models, and overhauling xformer’s documentation to better reflect what’s happening. In the meantime, I encourage users to follow the advice in my last post here.

sadguitarius wrote:

Can you check the link again? I’m getting a 404.

I’ll check it when I’m back home. Worst case I have a copy of the paper that I’ll upload.

Here’s the file. Not sure why the link isn’t working.