< Back to IRCAM Forum

Help with CPU usage using SimpleFreeFieldHRIR in SPAT

Spat
#1

Hello,

I have a personal HRTF from Genelec Aural ID in SimpleFreeFieldHRIR format (48kHz sample rate, 836 positions, 1024 data samples per FIR). Whenever I load that HRTF into spat5.spat~ my CPU spikes to 100% and my audio drops out.

I’m trying to use SPAT to simulate a 7.1.4 monitoring setup. I’ve setup 12 inputs (one for each speaker), set the room and input properties, and I’d like to render that in binaural using my personal HRTF. However, I cannot find a way to do this without overloading the CPU (even trying just 1 input causes CPU spikes and audio drop outs).

My system: Windows 10, 64bit, Intel i7-5820 6-core 3.3GHz CPU (so, a little old, but I’d hope it could handle this). RME audio card, ASIO drivers, 1024 samples (tried with larger buffers and it made no difference). Max/MSP “Signal Vector Size” is also set to 1024 samples.

I tried using spat5.binaural~ (in case the source/room properties of spat5.spat~ were causing the CPU usage), but I experience the exact same problem with that object, also.

I read on this forum that SimpleFreeFieldSOS is less CPU intensive. Testing HRTFs from the IRCAM database, it appears this is true - I can use SimpleFreeFieldSOS (about 20%-30% CPU usage), but I cannot use any SimpleFreeFieldHRIR without 100% CPU usage.

I’ve spent a few days searching online for ways to convert SimpleFreeFieldHRIR into SimpleFreeFieldSOS SOFA format, and I’ve not been able to find any tools or even programming APIs that appear to do this. (The closest I found was this matlab script, but it doesn’t support converting to SOS, only from SOS: SOFAtoolbox/SOFAconvertConventions.m at master · sofacoustics/SOFAtoolbox · GitHub)

Does anyone know of a way to convert SimpleFreeFieldHRIR into SimpleFreeFieldSOS?

Or, perhaps, something I might be doing wrong with how I’m using spat/Max?

Is it possible to render 7.1.4 to binaural in realtime? Am I being too ambitious?

Any advice or guidance is most appreciated. Thank you.

Steve

#2

Can you post a screenshot of your patch?

#3

Sure!

I’ve attached a screenshot (only using 2 inputs in the screenshot, but I’ve tested w/ various combinations: 1 input, 2 inputs, 12 inputs etc - all result in audio dropouts)

Here’s the patch data itself:

<pre><code>
----------begin_max5_patcher----------
4624.3ocyct0biairG+4IUkuCpzySrw8KySa1jM09vt0VUlbpyCYN0TzRzVL
iDoBIk8LYqc+re.uXIJIhtaJKO65Tisr5l.na.7CPN+A4+7a+l2L+thOmVMe
16l8qydya9mg24MsuWy67lmei2LeSxmWrNop0w4aRqpRdHc9a6MVm945VC2t
pr99YeJcSR4diE6pWmVW+ksocUx74y9+d1V1x1Kq3te66T6ufsI0KVkk+vGK
SWT2cMFk4F1amIctleXZ+Eg3F1gRJe2lr7P8z1.ECd2tpu8s4su6+5a+lleF
9waIGv4oOEZimGu0qJ2EOPuKI+gwCVKTvp8Mgm21FjdjPk+0JTqBsT8MEaSg
5Ye6f+MZfK4fcyxlnk6ZCci9xBc0yu8xj5j9H84P8MyuspXW4hzaCWycgPoo
XddzxAi7aeLqJ6t0oCqlglSWlUm.XuJccePwFw5prkoOsJMeyt5zX9jjtrKm
7cgg.FsxKELgf4UVg8syXs4oaTNYvfRYkLo0J1mmFVPKJxqpKSxxqucQV4hc
qSJiUmKJJJWlkmTmVgk.JJyRyqSpyJxucSwxV+luL89jcqqmC6eWX0FAG9Fe
Pubr5AoIstn3SI0294u7GiTCiV3aKK1VT9bahyZbdrrx5hxmKRkmw8FlM7Cs
UJZa3HgP1lFTYS9YrDy5j6RW2ZkG0LZj25zQsSRI2tq621UUmce1h8cNyWDR
4kgweiMuX2lz5zxgsHFfaqSerK59Ny3I2j.OodWYaA4fcYX.x6CqCeaz.rLY
Y1tJrF6prp5hxuP0spr+Hse7BfWuzQMOWN0qxV7o7zN3LejDj.dzg.gUIFvp
FyJNqRbfUoB7atUpMNqPqcZu5.pR5UdMyYLFuWLFpRPGUInipDSDUItPTkfN
pRLYTkfDpR7BGzI.QUhAnJQTynQ9EgpDSGUIngpDTQUBbTk3xQUBRnJAMTkf
DpRbkPUB5nJI7nCIBpRBtsJIATk7.pZ3PN8HgkjNMRRmFImHMRdgzHIcZjbx
zHIIZj7ENtRBRijCnQxnlQi7KhFImNMRRiFIoRij3zH4kSijjnQRZzHIIZj7
JQijzoQJ3QGJDZjBjFoHPiTC9PdbGFPRQGHonCjTSDHotPfjhNPRMYfjhDPR
8BGZo.ARpA.IUTynQ9EAjTSGHonAjTTARJbfj5xARJR.IEMfjhDPRck.RJ5.
IM7nCMBPRCBjzD.R5g.IEFPRSGHooCjzSDHouPfjlNPROYfjlDPR+BGZoAAR
5A.IcTynQ9EAjzSGHooAjzTARZbfj9xARZR.IMMfjlDPRek.RZ5.IC7nCCBP
x.BjLD.RlC.IbdjgNOxPmGYlHOxbg7HCcdjYx7HCIdj4ENxx.xiLC3QlnlQi
7KhGYlNOxPiGYnxiL37HykyiLj3QFZ7HCIdj4JwiLz4QV3QGVDdjEjGYIvir
C1fjGiGYoyirz4Q1IxirWHOxRmGYmLOxRhGYegirrf7H6.djMpYzH+h3Q1oy
irz3QVp7HKNOxd47HKIdjkFOxRhGYuR7HKcdjCdzgCgG4.4QNB7H2AdDJNxQ
GG4niibSDG4tPbjiNNxMYbjiDNx8BGX4.wQtA3HWTynQ9EgibSGG4ngibTwQ
Nbbj6xwQNR3HGMbjiDNxckvQN53HO7nCOBNxChi7DvQ9AaOR0pdqtuONPxSG
H4oCj7SDH4uPfjmNPxOYfjmDPx+BGZ4AAR9A.IeTynQ9EAj7SGH4oAj7TARd
bfj+xARdR.IOMfjmDPxek.Rd5.INCQ4ZLLgTxfURIihTJYCzmDJTJ3McASxl
fhIYSUxjrKUyjrIHZR1zUMIilrIYuTEvwfENIanxIYwc.OAbYhmjcApmjQT9
jLx5mjQP.krWfBJYzjPIinFJYzDQI6ZohR1DXUXZ9FUz2Hp9ljruGp6atDmV
ME4cOE8cOYAdewJ7dJR79Bz3MQQd+hU4MhLuORm273Nfm.tLZ0kn0aph8ltZ
uoH26WhduIJ3apJ9lnjuuZZ9dBh9lin5aNlru4BXZEEgeyGn7aJvpIHva9DT
3MepR7leoZ7lOAQdymtJu4zj4M+kpyaNrPu4CU5MWD2A7DvkAqt.0dyIJ2aN
Y8dyIH3a9KPw2bZR9lSTy2bZh9lesT8Mmjru2ll7oPp9vYsanwPmPZQDa6t6
ohh6icoqC0bZ9w807QLODAN1kOJAbu0w9SqexD289hC1F3JFWauqDwZi4Ok4
aiVOiOJauqixz3PoFPj1frRLHQrxcUZxxsqJxSiMCZSHKkM9.nMYKBsplKdb
6oYODV3K6SQLeeQ4ljtSEaxQzn6RV7oGJK1ku73.xx7NlzKcNi1XDuk3acR+
0gRuEee68YqOigepOiOB6TuJ1lrHq9KmMA9T+pVjrNcjtwS8KI+g9tCHuJt+
9pzlwRG2uCEx+9tj08Mz4aBSu2sYPruLqZ65jub6eTTrYjF4yl2WsXKKOr7B
C7W7oSFFj74rXSgaMAsHYqCg4cOjkC4A4UUZ89HRrXXr+PY1xH0Sqo8zk8zq
SbnpYDRyoY+jQHsFWmczDI4o1CCGZJykYMsf.obfuNLWAZzKSpV0wnOsBim1
X2naWoywkdumaMCShsWa7kyZMuKOqtIIsNEpoMvqi.AmT4uk7aIjdAWablva
HY5yZ0CpvtsUbVqecZfoEaPPuQxC258uoV6mKFVMuZvflzMaWkTEFS1segSp
tCVa2DPTqc6e3T37ylOPwOwixcqitQvNajCzN2iGmqJRFeMhj7EqJhr+EfUd
RJSiTfaSpWEY2RcooU6VuN1lDG3A8siO3hBK0rFnLaLCtDtnckMgvZEJl83O
cXZccfrDcevOaNxmmM.nK10eGpHqcYmQOn+c2gSbtaziX9ojxMcfqQO.52Ul
sFvbnvaCakaTykoOts8pUiZNM+wVZ83Vaw5aSJS1TEayQg9Ma68KCsQaibXx
aKFv61DMNb2W11eS.YLWJ1jmcdS46j6aFgWZ8seoERtQ3TTZZsEKTSq0A3l1
xhsaWmF81ZQRVYr6bBAS2deY5u2ucgH2AHVFvLMdXhZ8vBnqKdXrOx7QT4wF
CV1rIikogOwa99MvOVKFJUUElAFMGk0baVA7xSSJW+EPOVrdWCwFzmv38zx6
HTQOksrdEv7pNupVkzc6kQOtWUgoeIKOaulGlcljGZQwsOx95Fbn66wFpQ5x
DvXCAL1P.iMDvXCAH1PbcvFBLrg.GaHdcvFBLrg.GaHfwFh3XCAIrg.DaHHg
MDHXCAQrg.LUEGaHvwFBTrgf.1PfiMDjvFBRXCAB1PffMD.XC4drg0ORWlDF
aHgwFRXrgDFaHAwFxqC1PhgMj3XC4qC1PhgMj3XCIL1PFGaHIgMjfXCIIrgD
AaHIhMjfop3XCIN1PhhMjDvFRbrgjD1PRBaHQvFRDrgD.anfwFJXrgBFanfw
FJXrgBDantNXCEF1PgiMTQvF7arr1uTViSZ4lW.CQgwPT3LDELCQEmgnHwPT
fLDEIFhBggnHxPTfop3LDENCQgxPTDXHJbFhhDCQQhgnPXHJDFhBfgngYHZX
FhFlgngYHZXFhFjgnuNLDMFCQiyPzuNa8PigMz3XCML1PGGanIgMzfXCMIrg
FAanIhMzfop3XCMN1PihMzDvFZbrglD1PSBanQvFZDrgF.aXfwFFXrgAFaXf
wFFXrgADaXtNXCCF1vfiMLuNXCCF1vfiMLvXCSbrggD1v.hMLjvFFDrggH1v
.lphiML3XCCJ1vP.aXvwFFRXCCIrgAAaXPvFF.rgEFaXgwFVXrgEFaXgwFVP
rg85fMrXXCKN1v95fMrXXCKN1vBiMrwwFVRXCKH1vRBaXQvFVhXCKXpJN1vh
iMrnXCKArgEGaXIgMrjvFVDrgEAaXAvFNXrgCFa3fwFNXrgCFa3.wFtqC1vg
gMb3XC2qC1vggMb3XCGL1vEGa3HgMbfXCGIrgCAa3HhMbfop3XCGN1vghMbD
vFNbrgiD1vQBa3PvFNDrgC.a3gwFdXrgGFa3gwFdXrgGDa3uNXCOF1viiM7u
NXCOF1viiM7vXCebrgmD1vChM7jvFdDrgmH1vClphiM73XCOJ1vS.a3wwFdR
XCOIrgGAa3QvFd.rQ6wUFfaDrCq9KFh7uXH5+hgH.LFrBvXWIIfwP0.FifHv
XuRp.igJCLFAcfwPDBFCPIXLZRAiAqELFMwfwvTCFipbvXv4L.AgwHnHLFtj
vXTzDFifnvXzTEFilrvXX5BigILLFDTgi.UvjTJllRwDUJlpRQjU50RWo3BK
khxResjVJt1RoHtTL0kBIuTh5KEQfoDUXJpDSIqwTNbNC.pPQloDzYJIglRQ
ooDkZJQslhJ1TT0lBI2Tt.ApfH3TNhhS4HRNkin4TNrnS4WIUmxQkcJmftS4
uRBOkip7TNAomxQzdJGP7obZpOkCK+TNM8mxwDfJmpBT4B3bF.TgfHT43pPk
SQFpbB5PkSSHpbZJQkiIEUNlVT4iJF01dkAOiSOwxyvyzxj7tSmk7nCKVuSq
RSdLa+YgycioUEURGWY7MOGRO0+0YOlt+7GZPcuOWeeKzRbb94DWV0cH7Y.9
b9TlyrGeZywtlkeeVd1Yi.N1olea+Ysm6Gq35GTl0NoPvtQHZwMJgPZjBWL+
S9b+vJSq+bCyZLFlHh+Kypp6HI2nTs.Mmq4n7aOOUreJPWSRJtw1v.cguK7F
mUG+J5ZTF+M9FwzwEBlkI4h3WwvlkwOrcEKi12pLxazsWfwqbBkf6icAKSyq
5ONvgJg2UGFmKLbyA0s0cDfOaF0wNFYZ2XNM9iUtiOHgcW1Hbkr7P1JOYc0I
mG1GyReJ583fdiE4q+xY2WE9R+wB+9jE0EGcZE2Usqb6I9+TV9xhmtsNqt+x
d+1j5Y+iAOmmO3z3GQudiaJdLcjCnWu0vVax9i3lSyia6tly+c096yLgkokF
WHAyzM2wYTGkle9RhbP7cBqRy7dlwFFl7VRuyIcj80Pw1jeeWjFbat7tjxws
d+t0qqVTlllGIOFPOaxphjkicD72mlyWlVlk+PHi9PV9Y2e.Nys62VA2sd+5
hj59i984VatKAUUDJrvXsrGSpiTH2sqtt4zbuXcQUrd4dWFF9.dE1B34dkFV
ua4gG+4sO3y6eXfWk7X5xOVb2uE1IwGSpqKyBEU21sO54Cd6NFaNkue7vPR1
9R6U5opdy2+2y9SY4a2UWMSL6OUrq94W1tLyLdi0r5mxpWM6CMGF377ldulm
85ytKryqckIq+v73OY1qxdHfXZdtre3Ui+DZW.9romaL2vcC+OaXNnQ0dyep
8yM.8DaWL9Src4+kzIsIr0kl6KRm0Kc6px56m8Cu6Cgu9epRKqZdQXQtGSad
w2uc6OlTmz7xetHIDtOz7x22X9WdJrHXyu82yVl8iIO8CE40kEqadm+5O+K+
z49c329npc6NUE2mDuaMRWHTOnp6IMXnqq4GFt4x5y3WbVdQwlMo40mmk+9c
KyJl0MzeVcwr+596vKyGIH4ZnnTHZiKMqcHI2z9w9DrnQIe7njc0mvuLYw+d
F2KC+SMZXwAm9I597or13wItr9tqeTEViX4cIgEIhNRs05niVOb2aczHVJ5F
t1F2Nj.lesGrFEITUmTVO0IldBQpt4vFOSJ9ulok1a32nlkzN4rcEpY2WVrY
1O98+uu6C4B86l82Ruud1HeILua1Om8vpwLJrua1Oj1ry2Qr4Bk4O8WBEtuu
ve+tx1My0XUxdtTeewyusj+tYuOrSjwZJRQus1K5C4R46l8KEam8SkAV79KP
pF9t8sZot6M+ygcSdTQKMCLz579zVy8BmEgFUatSMxX.PrkpabtucVtR0BuL
h+ySsRVFnVBcnKMzyE5flI7gNhPdOjdmIkgzWHYERKSaCHu5uZzIgRvIgV9g
N.Q2M0xKf2HNoGn6xaGbz+AY12iz1HV2uU8A8KGtw47q6WUe1f1wI+I290C6
dqwq3i.nWeMEEGqBaVupwqqUExHUgWmHTRI.eNMbkpuul8fR9W4JT8Ut97Tp
uSx5EkK6+6t+0pAv4QZA7qPKfzbF40aNSyt0vqP+qXNmVKvRLm2SlS1t8wvm
qquL5p7vRi+Vwf6RcyCf+teu6O547xzt6RcGVpedRYXEk5vxI822Xm+YS+17
m27+qkx7cYO2J5+iSLuc027vmms4drW2E0rJ829MAG9+AI1wOS.
-----------end_max5_patcher-----------
</code></pre>

The patch is quite simple. A previously-configured spat5.oper (with 12 inputs, 1 room) that has ‘embed’ enabled, basic audio I/O, a spat5.spat~ doing the hard work, and then two messages to set the HRTF.

Thanks
Steve

spat_binauralMonitoring_patch
spat_binauralMonitoring_patch1045×708 118 KB

#4

Hello Steve,

Rendering 7.1.4 to binaural in realtime should work, even with relatively short FIR.
Could you please send me your HRTF file ? (via private message if you prefer)
I’ll have a look (at the CPU usage) on my PC, and I will try to convert it to SimpleFreeFieldSOS.

Best,
T.

1 Like
#5

Hello tcarpent,

Thanks so much! I’ve sent you a private message.

Thanks again,
Steve

#6

Have you tried to place your HTRF in another folder, like in Documents?
It may be that in the AppData folder you need administrator rights and that could slow down everything?
I’m completely guessing here :slight_smile:

#7

Thanks. Yeah, I’ve tried a few different folders/drives, but the performance was unfortunately the same. I suspect it’s just that the SimpleFreeFieldHRIR is too CPU intensive for my computer (which is getting a bit old now, heh).

tcarpent was kind enough to convert my HRTF to SimpleFreeFieldSOS and I’m getting much better results. I’m able to play back 7.1.4 in realtime, with reasonable CPU usage and no audio drop outs. (thanks again, @tcarpent!)

1 Like