TDM are the best plugs myth

[Andy_bt]

Originally Posted by ty604

Found an old thread talking about the myth that TDM plugs are the best.

Apparently it's a myth that TDM plugs are better than AU or RTAS and others.

Could someone clarify this?

Thanks

I'd say native versions are better as TDM is 24 or 48 bit fixed resolution at max but is then truncated and dithered systematically.
There are some real good TDM only plugs but that's a different matter.
A.

[ty604]

It's said that TDM costs much more not because of quality but because it's more difficult to develop for TDM.

[Andy_bt]

Originally Posted by ty604

It's said that TDM costs much more not because of quality but because it's more difficult to develop for TDM.

And smaller sales perspective, makes sense.
A.

[amost]

The myth about TDM in general...that is true

is about I/O & tracking. TDM works.





YMMV vary naturally

[deft_bonz]

TDM was for me low latency. Nevertheless I ditched my Mix Plus system 7 years ago.

[Jon Hodgson]

Originally Posted by ty604

Found an old thread talking about the myth that TDM plugs are the best.

Apparently it's a myth that TDM plugs are better than AU or RTAS and others.

Could someone clarify this?

Thanks

TDM and RTAS are capable of audibly equal results (differences less than you can hear), though it's probably often easier to get the better results in RTAS, simply because floating point takes care of precision and headroom issues for you, it's always working in its optimum range (but we're talking competent engineering, not rocket science to deal with that, so I would say it's usually an academic rather than practical detail).

The advantage of TDM (and now HDX, which being floating point is even more similar mathematically to Native), which a couple of people have mentioned, is that it works without needing large buffers (because the DSP does one job and one job only) and so can be used in low latency monitoring chains.

[ty604]

Originally Posted by Jon Hodgson

TDM and RTAS are capable of audibly equal results (differences less than you can hear), though it's probably often easier to get the better results in RTAS, simply because floating point takes care of precision and headroom issues for you, it's always working in its optimum range (but we're talking competent engineering, not rocket science to deal with that, so I would say it's usually an academic rather than practical detail).

The advantage of TDM (and now HDX, which being floating point is even more similar mathematically to Native), which a couple of people have mentioned, is that it works without needing large buffers (because the DSP does one job and one job only) and so can be used in low latency monitoring chains.

But that's because you're using DSP cards is it not? So really it's not an equal comparison of TDM to native. Incorrect?

[BasketCase Audio]

Originally Posted by Andy_bt

I'd say native versions are better as TDM is 24 or 48 bit fixed resolution at max but is then truncated and dithered systematically.
There are some real good TDM only plugs but that's a different matter.
A.

There are some really great TDM plugins in general.

Consider it nuts, but I don't think the difference in quality between 32bit floating point and 48bit fixed point (then dithered to 24bit) is capable of destroying a really ingenious and elegantly executed processing algorithm. A really great sounding plugin on any platform is all but indistinguishable.

As Jon Hodgson has stated, its more of an academic comparison on a computer science level than a practical comparison in audio engineering. It isn't ever that I have heard "This track sounds like shit! They must be using that rubbish old TDM system... not a nice 32bit floating point system."

[work2do]

Originally Posted by ty604

Found an old thread talking about the myth that TDM plugs are the best.

You said it. It was an old thread. There was a time when all the best plugs being made where TDM only. It was a long time ago. Computer CPU's were slow and there were not many native alternatives back then. Digidesign (Avid) invented the plug and had a big head start.

[BasketCase Audio]

Originally Posted by work2do

You said it. It was an old thread. There was a time when all the best plugs being made where TDM only. It was a long time ago. Computer CPU's were slow and there were not many native alternatives back then. Digidesign (Avid) invented the plug and had a big head start.

Yeah, its 100% technology related. When HD was released in 2002 it provided significantly more processing power compared to native desktop recording setups. More processing power equals more possibility to create complex algorithms. More complex algorithms equals potentially better sounding, more powerful plugins!

For a while during TDM's existence, there have been TDM only plugins simply because they were too demanding on computer resources to run natively.

Nowadays... not an issue. Multiple cores, multiple processors... They have taken over TDM in most departments in terms of processing power... except overall system latency.

[Andy_bt]

Originally Posted by BasketCase Audio

There are some really great TDM plugins in general.

Consider it nuts, but I don't think the difference in quality between 32bit floating point and 48bit fixed point (then dithered to 24bit) is capable of destroying a really ingenious and elegantly executed processing algorithm. A really great sounding plugin on any platform is all but indistinguishable.

As Jon Hodgson has stated, its more of an academic comparison on a computer science level than a practical comparison in audio engineering. It isn't ever that I have heard "This track sounds like shit! They must be using that rubbish old TDM system... not a nice 32bit floating point system."

I totally agree with Jon!
To the question "are they the best" my point is from a technical side native is more accurate, so they may be equivalent but certainly not better.
I feel like the question wasn't related to workflow, latency... thus my answer.
A.

[BasketCase Audio]

Originally Posted by Andy_bt

I totally agree with Jon!
To the question "are they the best" my point is from a technical side native is more accurate, so they may be equivalent but certainly not better.
I feel like the question wasn't related to workflow, latency... thus my answer.
A.

Roger that.

[Andy_bt]

Originally Posted by BasketCase Audio

Nowadays... not an issue. Multiple cores, multiple processors... They have taken over TDM in most departments in terms of processing power... except overall system latency.

The latency thing is a bit tricky IMO as native plugins often have less latency than their DSP based counterparts.
A.

[Andy_bt]

Originally Posted by BasketCase Audio

Roger that.

A.

[BasketCase Audio]

Originally Posted by Andy_bt

The latency thing is a bit tricky IMO as native plugins often have less latency than their DSP based counterparts.
A.

That is why I said "overall system latency".

As far as I am aware, all native plugins have lower latency then their TDM equivalents. TDM plugs run with a minimum of 3 or 4 samples of latency.

The advantage of TDM is, it is essentially designed as a hardware digital mixing console controlled by your mouse. Whether I use just 1 track or I max the TDM processing out with dozens and dozens of tracks, plugins, aux, heat, etc the latency remains constant and my buffer setting can stay happily at 64 samples and the overall latency of the system from input to output sits static at around 2ms.

[Andy_bt]

Originally Posted by BasketCase Audio

That is why I said "overall system latency".

Got it!

As far as I am aware, all native plugins have lower latency then their equivalents. TDM plugs run with a minimum of 3 or 4 samples of latency.
The advantage of TDM is, it is essentially designed as a hardware digital mixing console controlled by your mouse. Whether I use just 1 track or I max the TDM processing out with dozens and dozens of tracks, plugins, aux, heat, etc the latency remains constant and my buffer setting can stay happily at 64 samples and the overall latency of the system from input to output sits static at around 2ms.

The more power, whether native or DSP the less latency IMO.
The TDM design doesn't guarantee low latency, insert a LP EQ or something in the style as TDM and you're done with the 64 samples buffer
Though I get the point for most tracking plugs, etc...
A.

[Jon Hodgson]

I think there's a little confusion because we're talking about two different sorts of latency.

There is latency through the plugin, which is cummulative (put several plugins in series and the latencies will add up), and there is overall system latency caused by buffer size.

Native plugins work with zero through latency, unless they are doing something that requires a number of samples to actually work (e.g. you can't do a 1024 point FFT till you have 1024 samples, so if your process requires that, your plugin will have at least 1024 samples latency).

However, because you're running stuff on a machine which has a bunch of jobs to do, processing has to be done in blocks so as to "smooth out" CPU load (it's not that the processor isn't fast enough to deal with every sample in a single sample period, it's that it might be doing something else at the time, such as accessing a hard disk or reading a keypress or whatever), this results in an overall minimum system latency of one block, whatever that block size should be.

In DSP based systems because they're doing just the audio processing, and so performance is deterministic, stuff can be processed on a sample by sample basis, normally resulting in lower latency in a monitoring situation (it's not about how many samples the plugin code delays stuff by, it's about what block size your system needs to work with to get processing reliably finished in time.

Of course the faster your native system is, the smaller you can make that block size.

[Andy_bt]

Originally Posted by Jon Hodgson

In DSP based systems because they're doing just the audio processing, and so performance is deterministic, stuff can be processed on a sample by sample basis, normally resulting in lower latency in a monitoring situation (it's not about how many samples the plugin code delays stuff by, it's about what block size your system needs to work with to get processing reliably finished in time.

Of course the faster your native system is, the smaller you can make that block size.

Nice, thanks!
I hope I'm not completely wrong:
as I understand it DSP still needs to wait for the CPU to deal with a block and plugins which need to know what will be going on in x samples will output later anyway so there's a limit (which is also plug dependent) to how low you can get the latency even with DSP.
A.
PS: I guess in my previous post I should have said 2ms latency instead of 64 samples buffer

[stevens119]

Originally Posted by ty604

Found an old thread talking about the myth that TDM plugs are the best.

Apparently it's a myth that TDM plugs are better than AU or RTAS and others.

Could someone clarify this?

Thanks

Please clarify what you mean by "are better"? Like...sound better? Or like, better cause you used to could use more because computers were slower...Or like sounded better than the RTAS versions? No. I never believed that. Maybe they sounded better back in the day because of a delay compensation issue with LE systems, or converter reasons, or something to do with the old LE mixer.

There are some TDM plugins that still arent available for RTAS...hopefully soon AAX. Some great plugins

[BasketCase Audio]

Originally Posted by Andy_bt

Nice, thanks!
I hope I'm not completely wrong:
as I understand it DSP still needs to wait for the CPU to deal with a block and plugins which need to know what will be going on in x samples will output later anyway so there's a limit (which is also plug dependent) to how low you can get the latency even with DSP.
A.
PS: I guess in my previous post I should have said 2ms latency instead of 64 samples buffer

All digital processing has latency. Even the processing related to keyboard strokes.