Welcome to od|forum

Register now to gain access to all of our features. Once registered and logged in, you will be able to contribute to this site by submitting your own content or replying to existing content. You'll be able to customize your profile, receive reputation points as a reward for submitting content, while also communicating with other members via your own private inbox, plus much more! This message will be removed once you have signed in.


  • Content count

  • Joined

  • Last visited

Community Reputation

1 Neutral

About jimmay

  • Rank

Personal Information

  • Name Jim May
  • Location Los Angeles
  1. The first thing to do is make sure that the Texture Color Space is set to Linear, for images where "flat" or unperturbed is defined as 0.5 float (or 127/128 in 8-bit channels). I've had PNG's bite me that way before, where the file had a nonlinear gamma or something, and you just get "wrongness". I'm fairly certain that Houdini uses +X +Y (on my Displacement Texture VOP node, both "Flip" checkboxes are off.) This is, in terms of the appearance of the channels in Photoshop, "lit from the right" in X (red) and "lit from the top" in Y (green). I can't speak to the Z blue channel, as its effects are really subtle, so if the blue channel were "wrong" in the textures I've been using, I wouldn't necessarily know by looking. My current normal maps use 1.0 (255) for unperturbed, and range down to 128 for maximum perturbation, though mine are derived mine from the X and Y channel. I would expect it to be +Z if for no other reason that it would be arbitrary to make that one -Z just to be contrary I'm also assuming tangent space normals here.
  2. The one conclusion that jumps out at me is Conclusion 3 - RAM can age on you and get less reliable. I had a set of DDR3 RAM check out fine with memtest when I upgraded a system in 2012, which went bad about a year later. I didn't isolate it until I noticed the occasional corrupt PNG file in renders and about monthly bluescreens, and after eliminating the HD as the cause, ran MEMTEST again and found a bad DIMM. Might be a good idea to do that again.
  3. I'm running an Asus X99-A board with 5960X OC to 4GHz, with 96GB of RAM, 4x8 and 4x16. It's been super-stable so far; I've not had issues even though I've had several multiday renders over the past three months. I haven't done a whole lot of RAM-filling work, but so far so good. Lousy software remains my primary source of crashes I made sure to pick the first 32GB RAM from the motherboard manufacturer's compatibility list - in this case, Corsair Vengeance 2400 cas 14. When I picked up the 64GB later, it wasn't on that list yet, but I chose the 16GB version Corsair memory with the exact same timing specs. I also stick to stock timings for stability; 3D rendering work is overwhelmingly CPU limited, AFAIK.
  4. It's a shadowing issue. The cube has a solid shadow, despite its transparent material. If you stretch your sphere on Z so it sticks out far enough beyond the cube, you'll see it. The cheap solution is to shut off shadow on the light (Shadow -> Shadow Type: "No Shadow"), or from the cube (note sure if that's doable in PBR Mantra). The more expensive one would be to use caustics, but sloooow. I'm still a noob on shading, so I don't know how you'd do caustics using Classic Shaders, but the Principled Shader has a "fake caustics" option.
  5. Solved. This one is actually due to what seems like a minor UI bug in the "Collapse Selected into Material" menu option. I post the solution here for other newbies who might run into it. Production build 600. SESI created the "Collapse Selected into Material" menu option to automate the process of setting up a material from a root level or "loose" shader network. In particular, it sets up several nodes (the yellow ones in my fig.3 above) for you. CSiM breaks in two places, at least with my particular shader net. Here are the fixes: 1. The Parameter node has "uv2" in its Name parameter instead of the default "uv" that's likely on your geometry. Rename to "uv". 2. The "properties" node creates a Displacement Bound parameter for export. It has a long expression that evaluates to "0", which will create holes all over your displaced geometry. Delete that and set it manually if yo uknow your maximum expected displacement, or you can copy the channel reference from Principled Shader / Displacement / Displacement Bound (vm_displacebound).
  6. Hi all, I've got a principled shader network working as I want, so I'd like to collapse it into a material builder (fig1). When I do this, via selecting nodes and Edit-->Collapse Selected Into Material, I get fig 2. Any idea what's happening? It looks like the UV's are getting nuked, but I don't know how or why. I've tried removing the "uv2" parameter node that gets added inside the Material Builder, but that changes nothing. Network view shown is immediately after Collapse operation. White nodes are my original shader network; second pane shows SOP net applying material and being rendered.
  7. I had the idea to look up "deformation" motion blur, as I wondered whether SOP-level animation is actually "deformational"... That was it; found the answer here: https://www.sidefx.com/forum/topic/20849/
  8. Hi all, I have a helicopter rig from Steve Knipping's Rigids III which uses an animated transform node to set up rotation on helicopter rotors prior to going into a rigid body sim. For reasons I can't figure out, this type of animation doesn't show the correct rounding in motion blur. As seen in attached pic, the geometry animated at scene level (in light red) has the correct rounding, but the white one animated at SOP level with Transform does not. What am I missing? rotationalmoblur.hiplc
  9. Hi Atom, Yes, that was the first context in which this problem cropped up, as a matter of fact. Using the ROP Alembic node was an attempted workaround at the suggestion of an experienced colleague who was trying to help me out when this issue was threatening to blow a deadline. I'm pretty sure it's the same underlying code anyway. If I have time this week, I might try one of the daily builds, and perhaps ask support if they'd like to take a look.
  10. Hi Atom, That's what I did with the helicopter scene. The original scene is in Apprentice on my personal laptop, so I exported .bgeo.sc to try exporting Alembic from the commercial install on my desktop at work. The spaceship shot was done and tried in Commercial from the beginning. I tried a direct export File --> ROP Alembic node, and also tried inserting an Attribute Delete in between to wipe out all but the most basic attributes prior to export, and a Facet node to see if the normals had anything to do with it - no difference. It blows up after writing out 7 megabytes.
  11. Hi all, I'd like to pick the brains of the TD's out there on exporting RBD dynamic sims out to Alembic, for their experience with the best settings and practices to use. In my limited experience so far, I've found H16's Alembic exporter to be EXTREMELY crashy (build 557). Googling around, however, I don't see hardly any similar complaints, so it means it's probably something I'm doing. The one common thread I can identify is that of the two sims I've had this issue with, both are based on Steven Knipping's Rigids III tutorial, in which there's an extra stage using Point Deform to simulate dented metal. One of them is the actual helicopter crash from that tutorial that I built in Apprentice, exported as .bgeo to a commercial Houdini license for testing - the other is an all-commercial scene of a spaceship exploding. So what suggestions would you have to try and get around this? Scrub attributes? Clean/normal/facet right before the export? The only thing that I've seen work so far was to export an unpacked version of the spaceship to Maya - huge, but it got the shot done. I'm really concerned that something as simple as that helicopter tutorial should be so reliably crashy, though.
  12. ARGH, that's it. Thank you Malf. For reference should any other noobs run into this: there's two places to switch on rest fields: on the container (smoke object) "Add Rest Field" under "Initial data", and on the pyro solver under the "Advanced" tab, "Enable Rest" - this is also where the other controls for rest are located.
  13. Hi guys, Attached is a scene that does something nobody here can explain. It's a fairly simple pyro sim intended to look like a plasma trail from a damaged spaceship. The thing we can't explain, is the fact that it pulses out every 50 frames. Load up this scene, and render frames 51, 101, 151 and the frame renders black. Render the frames around it (or render the whole thing) and you'll see the element simply drops to black for several frames, then recovers for several frames after. The only thing we do that affects it is to shut off the Noises on the fields in the Pyro shader. No amount of messing with the parameters therein has affected it; the 50 frame period is constant. There are no animated properties in the scene that have it either; the spinning emitter runs from 1 to 120. The issue is the same with H16 style Pyro shader, and with the earlier H15.5 pyro in shopnet when it's set up with similar settings (both are present in the attached scene). I'm using those noises b/c for the life of me, I can't find any way to do detailing displacements of volumes at the shader level after the fact - googling such just gets me hits on how to do it for Arnold, or in the older shopnet. I'm very much a noob, and mostly have done shader work in H16's new context. My supe likes the results we have here, but if I shut off the Pyro shader field noises I lose the look. Any help appreciated - including workaround pointers on how to do my own volume displacements in shaders, post-pyro sim. Trail_v04.hip
  14. thankyouthankyouthankyou woodenduck. I really love the power and capability of Houdini.... but sometimes it reminds me of why I got out of programming 25 years ago
  15. Hi all, I'm setting up a basic missile trails test, where the missiles are emitted by one popnet "missiles", and the trails from a second popnet "trails" I'd like a consistent amount of particles per missile as more are fired, so after some reading around I set up npoints("obj/geo1/xxxxx") in the Impulse Count for the trails popnet. However, npoints consistently evaluates to 0. I tried pointing it to a poly sphere instead of the "missiles" popnet, and it still evaluates to 0. Any suggestions? I found this on H16 build 504, updated to 557, no change. npoints.hipnc