SMX3 & FKG3 Offset Differences

While passing a decoded source through SMX3 I noticed an offset. I removed the source and compared the output of two SMX3s in the configurations pictured below. (The patch and output of each is organized by columns.)

Adjustment of C1/B1 fully clockwise seems to help the issue.

Is this the intended behavior? Should the output between two different SMX3s in similar configuration be more similar than the output pictured?

The SMX3 on the left appears to be a little out of spec. Could you send an email to support@lzxindustries.net to arrange an RMA?

Here is the same type of comparison between two FKG3s. Is it possible a similar issue is occurring on the left?

The FKG3 example appears to be within spec.

Here is the output from the SMX3 on the right with per channel adjustments. Are these results a little out of spec?

I appreciate that the fidelity is much higher on FKG3, but the offset when compared to a similar patch completed with previous generation modules is significant. If this characteristic was made known to me prior to purchasing I would have avoided this release, which has been referred to as a much needed financial stimulus for the company.

Below is a step-wise construction of a keyed foreground with background layer. The photos present the configuration with output on the right. All elements in blue involve FKG3, the final row of images in purple is an approximation of the previous configuration but achieved with Marble Index.

That all looks correct, as you are applying offset with the knobs.

Isn’t SMX-3 an attenuverting matrix mixer with no DC bias normalled to the inputs? I know the R->G->B inputs are normalled but how could you apply a bias if nothing is plugged into the attenuverters? Zero volts multiplied by 1 or -1 should still be 0V. Am I misunderstanding how SMX-3 works and it actually has a bias/offset function built in?

Also, if a bias was applied to any of those attenuverters, it looks like we’re only seeing an output when an inversion is applied (CCW) and not on ‘full open’ (CW). How would that happen unless a negative (rather than a positive) offset/bias is present? Is there a normalled negative bias somewhere in the signal flow of SMX-3? I’m not sure I understand the utility of that signal flow but would be glad to have it clarified!

1 Like

correct. there is no bias voltage normalled to input jacks.

smx adds and subtracts. there is no 1-N functionality

also very curious what this signal flow is intending? the mixer needs inputs.

2 Likes

Just chiming in to add a few notes:

– There will be some slight tints and offset differences throughout your patch if you’re just viewing blank signals. Our signal standard isn’t based on precision color balancing of DC offsets – it is optimized more for transmission speed (bandwidth) and signal integrity (noise floor) since those are the feature points more relevant to video artists (than DC precision, which is more relevant to test equipment or video capture gear.)

– So just like any analog video broadcast gear, you are expected to color balance your IO points for any given signal path. The slight bias shifts over gain ranges are part of the way the system breathes, and what makes it analog. The output encoder modules (ESG, Cortex) have pro camps/ RGB controls for exactly this reason. You are intended to adjust your output image’s colors there to give you the color range you want to see.

– Knobs with outlined arrows are level knobs, not offset knobs. When there is no input patched, leave them centered to minimize offset.

– With the knob centered, the DC offset should be around +/-2% tolerance around black level. So for 1.00V signals, that’s from -20mV to +20mV. If this were a slider in photoshop that goes 0 to 255, that means your blue value might sit anywhere from 0 to 5. So that’s definitely enough to create a noticeable tint.

– I can’t tell if the images in the OP deviate from this spec or not. We need to take measurements of the resting DC level on the outputs. The red looks a little hot, but remember you’re supposed to be centering knobs for inputs not in use, so maybe it’s a non issue.

2 Likes

The photos in series and the resulting observations about the offsets arose from attempting to find a configuration that would allow for pass through of a decoded source, as below:

To your eye, are the differences in tint for A+1, B+1, and C+1 approximately 2%?

This is the solution. Thank you for your practical reply. It is wonderful to learn that the expectation is to color balance at IO points! As pictured, I work out of a relatively small system and am often colorizing a DVE-based layer with the encoder. I can see how its more professional to manage colorization of such a layer prior to the encoder and maintain the proc amp function for another layer that requires approximation of pass through. In generating the following images it was quite handy to be able to toggle the mix to assist in color balancing with Visual Cortex.

1 Like

I’d have to take a measurement – but it’s nothing I would blink over personally – I’d just be grabbing the encoder/output knobs to balance it out before I hit record, and after doing whatever else is going to happen. You can use the eye dropper and histogram features in Photoshop to do some image analysis. If you’ve got a zip of the raw captures somewhere I could do the same.

The new encoder has 6x trimmers accessible right from the front! So if I were doing studio work I’d be trimming those after getting a patch fully set up: send in color bars to the input, calibrate color, then hit record. I will also enjoy having full color proc amps on my inputs with TBC2. That way you can balance all your input sources to match, if you have more than one.

4 Likes