:-(

UAD plugins are killer sounding.


YouTube

My understanding is that all the vintage EQ plug-ins are essentially specific EQ curves. You buy those EQ curves. You can achieve the same curves with stock EQ plugins, but the more complex those curves get, the harder you can get the same results with stock plugins.

That's not correct. It's not just about having an EQ curve, it's about modeling the circuitry that the analog devices used to process the signal.

I am a fan of UA´s approach to anti-aliasing. Detected by trial and error that I need less time to dial in parameters with some of UA´s releases. Testing the 480L did reinforce my findings. Never ever before did I dial in a reverb as fast as UA´s 480L. Think that I fall back in old habits when the source is not alias free. Xfers Serum is known for being free of aliasing. Any other aliasing-free plugins you guys would recommend for testing?

Xfer Serum you say? Well, try UVI synths and their stuff. And some d16 Gruop stuff. Superb! And CPU intensive. And try u-he stuff. Also very very good. I find Arturia also good for sone specific stuff. Plenty to choose from.

We're living in an exciting times!

But, there's always a price to pay for quality.

You got exactly the expected response from your ridiculous post. You need to learn how to express your opinions and ask your questions in a less inflammatory way. Stop trying to turn it around on people who have rightfully pointed it out. You can start by changing the title of your post right now.

So much nonsense and false info in this thread. Mods better close it.

and how are their algorithms better than those made by others? Is there objective proof that they have it down to a science and others do not?

Reading the OP it was pretty clear to me that he was stating his opinion about his perception after having used UA plugins vs others. I really don't see why people are jumping on him for that. It's perfectly easy to simply write:

"Hey, I actually don't agree that UAD plugins sound better, but if there's a difference the reason is probably X, Y, Z."

That's a perfectly reasonable answer and is far from as combative and confrontational as some other responses here.

It's a pretty safe bet that the algorithms are different if the plugins process signals differently. How they are different specifically seems like something that Slate and UA (to name a couple) would keep secret since it's the core of their products. I have no problem with that.

Is there "objective proof"? Well, what do you mean by "objective proof"?

I have faith that Slate's products use algorithms that are carefully crafted and different from other manufacturers. I'm not going to say "better" or "better sounding", but "different". I think the same is true for UA. And I think it's then fair to say that if that's all correct then to some people UA plugins will sound better (assuming the difference is audible).

Not sure why people have a hard time believing that this is how these companies go about this. Heck, just run some well recorded content through some UA compressors that were updated and compare the results. It surely isn't only a simple filter curve applied to the output that's the difference.

PS: I love UA plugins, but I'm not using them and I probably won't buy their products again. So this isn't a fanboy reply.

Nah OP made the assumption that his perception was better than anyone else's. As for algorithms, I should hope that they would be different. If they were the same then there would be a lot of angry people wondering why they paid 2-3x price for a UAD plugin compared to native.

For objective proof you can't use your ears, or anybody else's ears, even "Golden ears". You would need to feed an identical signal in and record the signal out and see how well it compares to the original hardware in question. Whoever is the most accurate wins, if modelling accuracy is the goal.

Do UA engineers have better ears than anyone else since they implement anti-aliasing filters in their plugins at the price of additional latency and higher DSP usage? Anti-aliasing filters in a 480L? Seriously?

To quote the starter of the thread since he seems intimidated and doesn't appear anymore in his own thread.

If there is I'd be as interested as anyone to hear it. I never said they were better than all others, let alone that there is objective proof of such. My down-to-a-science remark was just an opinion on the consistently excellent sound of their mk2 hardware emulation plugins.

There are many tests that can be performed to evaluate/compare plugins from a technical standpoint. do some.

Yes. I think there are even plugins made that show the differences. plugins that show the differences of plugins - somehow strange…

I believe that it does not count a single thing if you can match some target curves of equalizers. Nobody who is making serious studio work does this - ok there may be some that do it - but it is not that important that you can find two plugins that can do similar things.

More important is that you find some plugins that can do the things that are going on in your head while mixing.

And that's hard enough sometimes.

Well Sticky,

you just nailed it! Hehehe...

Now seriously, this thread has really gone to sh***!

There's so much tension on several occasions which makes me sad.

Krešo

I must have missed that then. Several of the responses here seem unnecessarily confrontational and hostile.

The problem with the above is that these companies, if they do things "right" in relation to what I wrote, will use an actual hardware unit to model after. So that means an individual unit which other companies may or may not have access to... probably don't have access to actually.

Since we're dealing with analog devices, often vintage units, they're not going to sound the same even if they're the same make/model. I think a lot of people don't "get" this. One vintage 1176 that sounds great isn't necessarily going to sound the same as another 1176. So this means that one could potentially get UAD-2 plugins and like the sound of some of them yet not like the original units one comes across in real life. And by extension it's possible that the actual hardware unit that UA modeled a plugin after might not sound good to you. So the term "better" is difficult because of this.

So it's pretty much impossible to actually prove your (or my) point the way you suggest, because we/they don't have access to the same units.

I read often the thing with the different units with plugins that model analog gear.
I mostly use hardware while tracking these days. And sometimes i help out in other studios.
I own a 1176 hardware. And I have been in studios where there are 1176´s in the racks. And they all do sound very much like an FET Compressor.

And if this hardware 1176 does not fit the sound while tracking, the other one that sounds different to a small degree will also not sound good. Cause it is not the right tool for the source.

While mixing I would never use my hardware 1176. Cause the UA one sounds very similar, as similar as two different hardware boxes. At least to my ears.
And if it fits the sound not while mixing the hardware 1176 would also not sound good. Cause an FET style compressor is not the right tool at this moment.
When I try out plugins from other companies, I try not to match the sound with an UA Plugin. I also try not to match the sound of my hardware 1176.
And i think that some of the plugin stuff out there gets one in a really good ballpark.
But the difference for me - and I have the feeling that I have to point out that this is what I see when I use UA and other companies - is that the UA stuff very much behaves as you would the hardware expect to behave.
For me it is not just about matching the sound of the hardware, even if the sound is the most important in the end. Such things are very hard to describe cause they are very subjective, but increasing the attack on some of the UA compressors feels somehow really natural to me. And very often I get this feeling not when I use some other plugins.

85mat,

I don't disagree. My point was only that it's very problematic to conduct the test the other person was suggesting.

I agree with what you wrote.

Of course my post was not ment to disagree with you. Hope you did not get this impression.

Specifically, equalization is just a Fourier transform, which is a linear function. Your ability to reproduce equalization is only limited by the resolution of your capture device and the precision of your math. In the case of the audio spectrum, our analytic and computational ability goes well below the noise floor. Moreover, the computational requirements of doing this are lower than one might expect, thanks to Nyquist.

But actual devices aren't perfectly linear, and we can certainly hear differences between them that aren't accounted for by any spectrum analysis. A non-linear equation will typically have more than one solution, which isn't uniquely divisible into linear and non-linear parts. Simply put, an IR (i.e. equalization) can never model non-linear behavior, because it's a linear function.

The problem then becomes a question of how you analyze the device, and how you choose to make your model. Those choices make all the difference in the world. They amount to a little spit here and there, a little dynamic spectrum limited noise, and a little non-harmonic distortion. Add them up, and it becomes audible character. That's what people pay UAudio (and everyone else) for - those choices. And that's why there's room to prefer one model over another.

I agree that in the early days, there were a lot of coders, including UAudio, who didn't go to these lengths. But it's been a while since anyone was really limiting themselves that way.

Haha, and I'm going to agree with you too. Just like going to a guitar store and playing 10 American strats. They will sound different. I'm ok with any of the top modellers as I feel that if they get 95% of the way there that's good enough for me.

To me it's like choosing what is the best hammer to hit a nail. It's not the tool anymore, as all the tool makers are pretty capable... It's the carpenter that matters.

I think moderators close it when the threshold of 10 posts is reached revealing to us pions that the UAD accelerator is just an expensive dongle, plus 20 additional posts proclaiming how expensive everything UAD is.

Then you need 5 more from "insiders" who proclaim UAD is close to financial ruin, and 3 of those 5 predict a total recovery...but only if they go native by the end of the year.

Ecclesiastes 1:9

This is one good example of what I am talking about.

What EQ plug do you think sounds better than a GLM 8200?
What Reverb plug do you think sounds better than a Lexicon 480?

Some UAD plugins sound better than competitors but actually most UAD plugins sound worse

1176, la2a, SSL e Channel and pultec, are Uad winners.

32 bit floating point calculations offers over 1000dB of dynamic range. If someone compares this with a 40 bit integer calculation of hardware, you diminish the sound difference more or less to rounding differences way beyond by what one can hear. Which means that I highly doubt that the actual different sound between the hardware and the software (of the quoted Weiss products) is due to the bit depth.
I was not talking about SPL, but about dynamic range.

I would point out, though, that at DAW values (we're typically working on sampled values between -1.0 and 1.0, not +76236742374 as a sample value) the rounding of the mantissa is constantly happening at what would be around the 25 bit level for EVERY stored number and calculation made, only scaled down into the noisefloor by the exponent. Samples of 0.5 to 1.0 and -0.5 to -1.0 are getting quantized as if they're roughly 25 bit fixed point every time a new float is stored.

That's just how floating point works.

So you're correct that the representable sample values (dynamic range) go from infinitesimal to impossibly huge, but when it's impossibly huge the quantization values are themselves so loud you couldn't stand to hear them at even one-bit level. It's just a mantissa being slid up and down in volume all the time. When it's huge, it's kind of coarse (not for output audio, but if you were to do thousands of calculations it'd degrade on you from quantization). When it's impossibly tiny, that's when you get the high quality audio resolution.

Also, when you do math on floats the math has to settle on an exponent that'll cover both values, so the order of accumulating audio data can matter: this is why multiplying by 1.0 doesn't always equal 'no change'. If your tiny value can't be represented at the same exponent as the 1, it can get rounded to zero (assuming you're not using a different word length as a buffer). If you were going to amplify it up again like in a compressor, that's just too bad

I stopped using 32 bit floats for calculations when I realized that. It's odd: the quiet noises remain as good as you could want, but when you overprocess, the loud parts go flat on you and it's because of extensive quantization rounding. I recommend longer data types for this reason (and minimalist algorithms) and when I get a 'house sound', that's why. You can notice a difference in sound texture from doing this, though it's very fugitive… and sometimes the more accurate choice doesn't always sound better (I did a retro echo-diffusion effect where using floats sounded better for that effect).

All my stuff is in floating point formats, but I actually don't think floating point is optimal for processing audio data. I'd prefer a really long wordlength fixed point format, not all that much headroom, and just be careful about producing overs.

This whole thread is so subjective. There is no way to convince the opposing opinion.

Let’s get scientific. The real question we should be asking is:

What can a Sharc DSP do that would be more difficult or not possible on an Intel CPU?

Perhaps there are processes that are unique to Sharc chips... or nothing at all and it’s based on pure coding experience.

I would imagine one advantage of DSP chips is that you are writing code for the same architecture for over a decade—regardless of Mac, PC, brand of CPU, many different models of CPU, etc...

Can anyone chime in who has work with Sharc DSP coding?

I bet it's coding and has nothing to do with the SHARC chips, otherwise the native Softube and BX plug ins would sound better in UAD. Supposedly they don't....... Also if UAD is paying for authorized emulations, they probably have access to resources the others don't, from the manufacturers of the gear they are modelling. No real mystery.......Classy company and never invade threads defending themselves. They let their product speak for itself.....