Phase Correlation Meter

udalilprofile

@resonant Hello. Somewhere on the forum there is an algorithm you are interested in.)

resonant

@udalilprofile I think this is what you are referring to: https://forum.hise.audio/topic/10913/correlation/5?_=1765392189920

udalilprofile

@resonantNo) Someone from the user created a meter from the panel

dannytaurus

@resonant There's a phase correlation meter in the HISE master output panel.

The HISE source might give you some clues.

CleanShot 2025-12-29 at 10.44.03@2x.png

resonant

I'm trying to create a Correlation Meter using the method below. However, the meter gets stuck at +1 even if the signal is mono.

I'm getting the gain levels using getCurrentLevel Where do you think the mistake is?

L*R / sqrt(L² * R²)

HiseSnippet 1764.3oc6Y8zaaTDEe2DugD2+nVTQkCHzpJN3TUkl0IosnJjciSRSDNMV1gT3TY7tiiG55cbmc1TRPEwQ9JzaHwM9D.mnW4FRbfKbnHNgDH0KbFdyLq8NqyFW2PZgT08Pj22adu4898dy68lM0XTWbXHkYXleyc6hMLOoUicC3sqzFQBLVaICySasNJjiY1JRKtaWTXH1yvzb7aJHXNUNC4ySJsHxGE3hSHYXrEk3hqR5P3ITqU98I99qf7vaR5ns54KulKMnB0mFA1y3VyZzE4dWz13agDKaLKiUQgsMLunkq27ty6dkly6fPstpSyqfadMmq5ful6BduqqKxcgltNyMGxvbhk8HbJqAGwwgFl4Vj5sai1z6Gn1fsHgjl9XwKNFMfcVQdEpumvEE+1nRahuWsd.UnggoUsDXabErcNq0Idj9zSfuyHYXmHgN.ZNVZya7TlmyAYdYXRlZlTNkIcVqFtLRWdBGg8bBq0BfnYKDDmzME0ZMF6axYUgBqHfOSGzcwqvfW5KQghyN6krKVb1oudd34xW1tBkwv9HNgFXuNFVVdHDFxs2AAILLlujl86Y2Smai4Unc5RCfWJbg9q3BfBSDbiHd2HtH6BjTlNJja4VsvtfPIbERM39wE61xAaSBvy3xv.tJRxXaz7SDBKDHdYyDh4RVUP99MgDsBshBbENRgoy+YSkeJgJqBJKY+jVeD3vA7p3cv9E3rHLnR4RqO7k1B4GJWqZ0tfQrkOHR0KV29x1EVGANY38X79+pP0osunc+2pO8zJoAjLj5imoKi.PnROwb5glBWaKjeDNgc9GH9CDvfTMF2l2FayENu88I711e9bKr5dyjfLh0HwlBysfJT26IAsaP5z0GK7UmrhRZrSGba.FHldfoDJ1BQHA9PPztWTwlFn3IjiQ8K31SnKA5EbVHpkeJssUfB2fyYjlQbbAcF2l3waeIamYE4wO354GtkHzS7d1OUY.aQhQIt35zfC1AELy18DbNxbtg5ZJaHaGSyJTgdMK7VTNdC44CHe6A4sGjUqVYxKVc9P9TVrUYZCQvBAQcZhY5PgXgPktzkOmXzJe5pvDsERCVKfv2nKN98j5+NYTr0HFTgxtiGagvu3xhtmNtnaCehGlYP7DUb6evzP5.5sLMNHErXDmSCjJXBKQLomrVlRQGA4lzRkjlwt9AqsDhiD8Mh8Ev+5hYbh.FMWBuCz0V0EYJqkvg2kS6Jc13LHvlFI28SS1zaTdW8N8cRM4v22g344iqQCIhHu9.DFZap0SChR1uunh198nVkYHOB8lLZT2dQ0QQs8QvCRwn8q382HFFAf5EAMISOWf33ZLCgOq2LVzvM.vhc0CZCNrPtCNae1LmkIyb4QzbOqUMB2sc116XYXuPt0ya6MdzqSYo54jXr4rV4COryYM6nOm0j8mXte4W41eRqj4.Fx7xFFKASNkBxj0tSH7QkyXh50B1ANnVi5iXCB3ieTMMYtQNqPNj1wlr3W2RB4GmLXYJwwGC9LV0P+eHgnNMhSB1FFelQfh2V2JpSC3ZktXvTCBv9ghFoiIZunde1dMVafC7ju72vSLSGw6lwLc5wTGRNuBRNg0sQ6fkyBKQi2P9dKJqi8MwAXl.qbFRIgebTuBc2Q9Jza3xASXSFJHrKMD6nq4F3NjMglfgonNfDE038n9RTLccLdTPZUqHkZUBnXE.JzV2jk6QrnNQHEJkxf2Soo0IZCV73RKGffDjFXvy81HzE.FANqmiUKBt7k7jjitfIjKpSdUDyChgtCq1ZtQq1Z5d7ZqOihtS775qJLh15QUAf+sy0bzzbdpd1np47xPWSeXD29sAZgh748nl9LoXTxtsoAjTY.0wPojs2FmJwJSGBtWFbUpDJmqbcrOFEpcN8cJWEt7GhA3D9PhEO6ePnLiWuokxbsEERre4XdzwOtMO5y7IzLC4iu+z+SGm9WGCCM1L9.phjshV5T+5TZmFj85kS9we6a+mkVB0oKzJMlT44l62Jcar5qZES6qd3C+gRKw1c.Z+dI84YK+q68GkVggw6g0x5gk8ckNxFb8Eca+SnP4WSMVoDdmT86zvpjj3.VUbKst3uU49LpS1tMWujwJXrm3KxnKwO+0OtTO55B7n+pvuTpJ890fMDP360WOe4OUdUx9nZrT4A5ftIFtLb5VdO43aT4ftW1ozH37pKl8pKl8pKl8R+EyFNj.kAO9DAOuEb4lh1x4dr+uevmWDAxWD6QGjKidGW0mSWTicRIEvuCj+ammxZcw61NC9YzML5.2A6NttoU09Dr3gUv4NrBN+gUvENrBdkCqfW8vJ30d5BJth7Mh3zNpyIFFqWaY4+1CSS0mLPdjw3efQSF.Z

dannytaurus

@resonant Looks like the correct formula for correlation is:

1.0 - (abs(left - right) / max(abs(left), abs(right)))

From the HISE source StereoInfo::calculate function:

void MainTopBar::ClickablePeakMeter::PopupComponent::StereoInfo::calculate(float left, float right)
{
	auto normGain = hmath::max(hmath::abs(left), hmath::abs(right));

	if(normGain > smoothedGain)
		smoothedGain = normGain;
	else
		smoothedGain *= coeff;

	if(normGain > 0.0f)
	{
		auto thisPan = hmath::abs(left) * -1.0f + hmath::abs(right);
		thisPan /= normGain;
		auto thisCorellation = 1.0 - (hmath::abs(left - right) / normGain);

		auto idx = jlimit(0, 100, roundToInt(101.0f * (thisPan * 0.5f + 0.5f)));

		panPos[0][idx] += normGain;

		smoothedPan = coeff * smoothedPan + (1.0f - coeff) * thisPan;

		smoothedCorrellation += thisCorellation;

		min_c = jmin<float>(min_c, smoothedCorrellation);
		max_c = jmax<float>(max_c, smoothedCorrellation);
	}
}

You would also need to implement some smoothing too.

dannytaurus

@resonant FYI +1 is the correct value for a mono signal.

It means there is 100% correlation between the left and right signals, which is the same as saying mono.

resonant

@dannytaurus Thank you for the idea.

Even if I don't use the Smoothed function for now and only use the formula, there's still an issue.

While HISE's own master analyzer correlation meter occasionally shows values less than 0, this correlation meter never shows values less than 0.

Also even when the signal ends, it remains stuck at positive values again.

HiseSnippet 1752.3oc6Y8zaaTDEe23rkD2+nVTQkCbXUEGbPkTuNIsEUgrabRZiHowxNjBmZGu633gt6LlcmMsInh3BRbiy8FeMfKTgDG3FH9B.BIjPvg9Mn7lYV6cVGaWSnjlhpOT048l2adueu27duYSsPlKNJhEZXleyc6fMLOgUico71UaiHTiUWxv7TVqih33PaEoE2sCJJB6YXZl65BBlSOog72iKuHxGQcwojLL1hQbwqQBH7Tp0p7dDe+UPd3MIAZ6d9Jq5xnUY9rXvdxYUznCx8tnsw2DI11DVF2.E01v7srb8l2cd2K0bdGDp0kcZdIbyq3bYG7UbWv6cbcQtKzz0Yt4PFlGaYOBmE1fi33HCyIWj4sai1r6QUGvVjHRSerXgiQC3jUjWg46IbQAUipsI9d05BTQFFlV0RgsbJX6rVqS7H8nmBemVxvNUBc.zbhrlWtLlmit4UTy7FfIYpYRSpLoyX0vMjzgmxQXOG2ZUJDMagf3jton1qwDeyjVUYvNn7YCP2EuRHrnmDEJUr3ErKUp3LWMO76hWztJKLD6i3DF0dcLrs7PHLhauCBRXBC8kzreW6t5baLuJKnCiBKJb9d637fBSEbiXdmXtH6BjTlNJja4VsvtfPobER0+4wEm1xzsIT7rtgX.WEIYgaz7iDBKDHYayFg4RVUQ99MgDsBshotBGovL4+joyOsPkqAJK87jVeL3vT9Z3cv9E3gwXPkxsVezasExORtW0tcAiXKePDmYKZ+11EVGANIpYTg0fU0mw9h1RJAn6qwZlKX2aQ8YlQoM.YiX93Y6DR.HUo2DNcQWgqtExOFmxN+CD+CD.gTuPtMuM1lK.C66Q3ss+z4V3F6MaJRI1iDqJL2BpPe2eoneCRPGerv2cFTTSic1fcCv.wrglhnXKDgP8gfpc2njMip3IjKj4WvsqPW.zK3rPTL+zZGq.EtFmGRZFywEzYbKhGuMjWWTjW+fqlezVhPOImYuTm9rEIFk5hqynC2AELGr6I37Ly4FoqorgA6XZVgJzqYg2jwwaHuu.4aOHuc+rZ0Zf7RTmOjOMH1pLsQHXAZbPSbnNTH1HT4Ka4TqgWNUuZuqBSz1HitJkv2nClNrd.FI.IT5MWhUA+Otrv6oRJ71vm3gCMHdhpt8tLZHMZs1lOwXXJXwXNmQkJ3XVh3PWYsLkhNFxMkkJwbemZYi2e0kPbjn2Qhu.9WGbHmHfNykv6.ctUcRl1ZIbzc4rNRmMIqArowxcue5gdsJ6ltXgJAYld3aCHdd93ZrHhHZqODgg1gZ8zfnzy6yppcdOpUkPjGgc8PVbmL3vXo7d33vTOZXpe+MlgQBXdwPSyryIHttlvP3+5MmEMfo.trqtlelM7v3ZhmwpFg61dv13DCvFgbq+KrwjQtNokp2RpANo0Jevgv7US0aR4dkYkG+IrR6+Oh4jMLVBlXJCLIqQmR3CqLfIoWktCb4rFyGE1OHmar8RmQ6kSN1YBxgyNRms9pVRX9ntQJC8GsMxSaUC87JXWmEyIzsgwdCIPoWqaFGz.ddnKFLOJE6K5xaNgnEgZcwtMGafodxEOA9kvzQr1LgoSWl5vv4TvvwstEZGrbFVIB7Zx0sXgA1WGSwgB7wYDWw+ww8ovcF6mBugKGLgMCQznNrHritlafCHaBsvhxPsOIJow6Q8jnT15R7XZVUqHkYWBnXE.Jz12TU5RrjNQHsIixf0Yzz5DsgC9kxKSQPRQCL34daD4B.i.m0yqpECOhRdiwQWvTxkzIeCTnGDCcGUsxIGuZkOkOHvwNx8AAdAXriL8TmtqMp5otLzryGlFsWk7VnXedWpYu5Il5qSaFkjIPWGCUL1dabl7mA5PvylfW5jR4rUpi8wnHsqiuYk0f2lgBAbBe.wBm+we+lAFudcKk4ZKpWX+h63h49+83h41ep8oRRsqig43Zlb4SQxVQKaZccFKnAYut4a2oUqeu7RnfNP2vDRUlateq7svpOvTBsu5gO7GJuT3t8Q6OJqOhYkecu+p7JgX7dXsLZXaec4mYyRdX2493JT9UTS8Ig2oT++rvpjj3xyZ3VZMheiJ8XTmrcatd4fUvXOwGCQSB6O+O+oxcoqKvc9xu66KuF6d0fCDP3Otmd9hetxMH6ipwRU5qI3lX30nY6Z83WbiJC6oRmTify+p2JQd4akd4akd4akdNb2dzv.TV6ncj5bVvaLJYKmQw94yPJGFArCiyH.4Fxtsq5qQKpMNkjB32T4e41osVWr11o+uBsgQ.7Vna65lUU6SvRGTAm6fJ37GTAW3fJ3kNnBd4CpfW4oKn3opWKlyBT2MfmrWaY4e0.SS0q0kWSL9afCFdsF

dannytaurus

@resonant Here's a solution using ScriptFX processBlock() (no nodes) and a global cable. It's a mashup of Claude code and my own, and it works quite nicely.

Just put a knob in the UI called knbCorrelation with range -1.0 to +1.0, or use the snippet below.

// Interface inInit()

const rm = Engine.getGlobalRoutingManager();
const corrCable = rm.getCable("correlation");
const knbCorrelation = Content.getComponent("knbCorrelation");

corrCable.registerCallback(function(vNorm)
{
    Console.print(vNorm);
    knbCorrelation.setValueNormalized(vNorm);

}, AsyncNotification);

// ScriptFX 1 onInit()
const var rm = Engine.getGlobalRoutingManager();
const var corrCable = rm.getCable("correlation");
corrCable.setRange(-1.0, 1.0);

reg correlationSum = 0.0;
reg sampleCount = 0;
reg smoothedCorrelation = 1;
reg prevSmoothedCorrelation = smoothedCorrelation;

// ScriptFX processBlock()
function processBlock(channels)
{
	var left = channels[0];
    var right = channels[1];
    var numSamples = left.length;
    
    correlationSum = 0.0;
    sampleCount = 0;
    
    // Process each sample pair
    for (var i = 0; i < numSamples; i++)
    {
        var leftSample = left[i];
        var rightSample = right[i];
        
        var normGain = Math.max(Math.abs(leftSample), Math.abs(rightSample));
        
        if (normGain > 0.0)
        {
            var thisCorrelation = 1.0 - (Math.abs(leftSample - rightSample) / normGain);
            correlationSum += thisCorrelation;
            sampleCount++;
        }
    }
    
    // Average correlation for this block
    if (sampleCount > 0)
    {
        var avgCorrelation = correlationSum / sampleCount;
        
        // Apply smoothing (same as backend: 0.8 * previous + 0.2 * current)
        smoothedCorrelation = 0.8 * prevSmoothedCorrelation + 0.2 * avgCorrelation;
        prevSmoothedCorrelation = smoothedCorrelation;
        
        corrCable.setValue(smoothedCorrelation);
    }
}

HiseSnippet 1650.3oc2X0sTaaDEVBPYJNoYZ5jNs8NML8BQf.VDBgVZZLXfDlFCdvz7yjoSl0qVasCR6pQZsab6vyPeE5aRyiPeDx08p9Fzd1URVq.kTvSaZR8EF1yO64aO+sm0si4XRRBO1vr1QihHFlWwpyHlvuoOhxL1aaCyqZ0BkHHw1oj1ZTDJIg3YXZN88kDLmcFC0m+3dagBPLLofjgwi3TL4gzPpnfZ6FeKMHXWjG4HZnlzq1XOLm0jGvG.3YZq5FQH7wn9j8QRwlxx3AnDeCyaXg8VEuJdstq5hP8tia20HcW28Ntj0w216KwXD91cwt25VHCyKsiGUvi6HPBRhg4Law8F0wm+CrTC7HZBsa.Qtv0nCX4Tx6xC7jGQIUil9z.u14NpDCCSq1EtsoScaW2pE0iNldg66iTLrKzP2AZNUY3McI34pCu5ZvqBHYpAoYRgz0r5fioQhBNR7bYq8XPzrGBhS5PIUVio9cSqlbPBlXoPzwjcigEi0vYs50WzF9Z9MpUChUIB63P66ZuCqOkQVpOQb+.dWTvg7ABJqeKDChdwNfzoBi4wwMQvIBzINTJuZkybRFj.jfxYyMV5iYcaVPGTIGXR83gQbFrvYtxhMWJzxryRwj9TY1aSTPPWHaxo2.FVJmyv84wgyW6mpYCefcNgChGESgsLk0FJNk28kRHhGgBFPjRfBn+Hwarz0NYQ6MSFwv6yEzdTrRAfdtEs4LfC4.lizpyV6jZ1mlUudUxSdti4AARWYErkUQwuIEcXCB6RhWzdnD5iEDxTJm9Y85S+zqNvogAMA4r8XTwAQD1qqlwHK1ISUyPEHpPknd0rD0NATORrA0SV0T1uanPtdiEiuausQBT99AaMXtHRrfJOIlaSFBMdRKDl0ZaRxwBdjR1r7Fn0v425uXrk+kdMFMdwq50HrTGveMj54EPZySnJ8zw6YKPgVCbuAfIJ2uP1UMigbOzKRkEhLXuGo208erlHmWHdMq1TA1uZLNUEXDBR+afwrVuen0N85QvhB.Ni0tO4sWe1ZYIPfMkF+JYKs28ItU1g8msrR6uMDEewZeJU372BMuEHzx5PDqOw4ltKAcugujMqfFi1ZpzYfDH0Wp9FJNInvn.RS9.lPRNiXHmK7IdkaJ6lxLJlLrSkBTgZaLtGEnVDJlbDuc.ZjSpYODBRKZ2MfiOtCzg8rM1hRiAaIkvA6iXLRPhRLoGJfzSB5b5Oq92m1IW4to88KwzUiIzkriB.IfDxcYo.BquvOUB0WU6wjbNiGarNKub9s+1DD1OSR6HDMVwuGO11QZepRQ3OesFTf0Krv7JASupJGsR.lJSFZeFM6rT5vNVD0pRxTRXFbIlryCHYKjvGt7+ENp+A0MwovVyun8XpZ6+7yWw1R6Y6Lda+FoqZ9w7JNK4.P3SSNUl0R0suocUn.Hqab6kGCeMbTQ7Zg6dZyTVbsf3BKTv5jZEemGR2bHIFpN0sfJRJ2+zb2Z49.8TCvMTUvDMre4y9o.9x5PqBWsDPQQAixJ1flGJyRrQ.XfwdHLuuBB.qaeCUgJkOHwdAfvJ.A7.vTLQQro557BsqpLOeyJePJP5Eq8vYOfkZmol.yoBEyh9mbAGCx5be6f6a9xowMtEwTXlAq8gxX3wLXRyr9M.MyojSCktttbs79wNP.Rs3OgOYLckqMyX5lyrXPpIZBH86O+7TlW15wngD0q9TWf8Ip0Ptbn88ILHKGbMtug288am228Ecte22AXA.gihQrjHdBwUameYGRH8HXDtDcpmViUpTCcpFaSDCX5a8pMRIURJoqXWvU3pOwUNwUzI1FwzD5ouDVqw+oMZQKlhz3U2aGlLatCAN4dGjfAGizOWZCGDjPdL0S36pqXA4UzI+.TrGDCw5fGRniEs8Qk8fZjKcTMm9M8VY2I7sxW5ct2J+dvj3kFyb1bL1gJuBXG1PR.Tmqv3GCE18PCBD4TKWn1hy3Q9bFsTZwgDn8TeXxRcrW4AZSg.t7nfx0abHIf.INEj9hFODlcEEC9IxD5KbuvibWY75yrRgqsr6h86uufZ5+W+Bpye0Gzv7c6n3mZAMtWwVEJs+uIV91Xfi2F1HDgi4OGmNmlrn+CTTfyMS8a+NqUK4Za2y9CAEBWY7bLt7VcFEWYRU7VSphqNoJd6IUw0lTEuyjp35+8JJuQeyABdXZsggQq16nFa0zLcDHUYhweQsTysb

dannytaurus

Note that the above code spams the UI quite a lot with new correlation values.

You would probably add some throttling with a UI timer and/or on the processBlock end.

resonant

@dannytaurus Thank you su much for this useful tip, it is so sweet! That is exactly what a correlation meter do.

Do you think using Claude with Hise is useful? Are you using the paid plan?

dannytaurus

@resonant I find Claude very useful when coding. I use it every day in the Cursor IDE when I'm building web applications in Ruby on Rails. I'm about 5-10x more productive with Claude.

I'm on the $20/mo plan with additional OnDemand pricing for extra requests. Given that I use it every single day, often for hours at a time, $20/mo is a bargain.

If you open the HISE source in Cursor, you can ask Claude questions about code locations/definitions, workflows, how to write scripts in HISE, etc.

Phase Correlation Meter

30

2.1k

13.1k

113.9k