STS latency test, worse than Sofware sampler.

djmicron · Post by **djmicron** » Sat Apr 08, 2006 8:35 am

On 2006-04-08 05:36, lagoausente wrote:

As I explained here, I route STS to analog output, then just take a jack-rca cable, and conect Creamware analog output to Terratec analog input and record it. As well as the original mic signal. So I can compare the time between the mic signal, and the final Creamware analog output.
Don´t need Asio at all, just route from STS to Analog output.

in this way,
the sts/scope latency is added to the terratec latency.
i really don't understand this kind of test and why do you record the sts output from the terratec when you can use the asio destination....

lagoausente · Post by **lagoausente** » Sat Apr 08, 2006 9:04 am

When using the software-sampler, do you route out of the terratec and then into it again?

like:

trigger -> terratec MIDI -> Cubase w/sampler -> terratec D/A -> terratec A/D -> Cool Edit
+
Mic -> terratec A/D -> Cool Edit

Exactly!

djmicron · Post by **djmicron** » Sat Apr 08, 2006 9:07 am

how do you check the midi to cv latency of the roland hardware?

lagoausente · Post by **lagoausente** » Sat Apr 08, 2006 9:09 am

On 2006-04-08 09:35, djmicron wrote:
On 2006-04-08 05:36, lagoausente wrote:

As I explained here, I route STS to analog output, then just take a jack-rca cable, and conect Creamware analog output to Terratec analog input and record it. As well as the original mic signal. So I can compare the time between the mic signal, and the final Creamware analog output.
Don´t need Asio at all, just route from STS to Analog output.
in this way,
the sts/scope latency is added to the terratec latency.
i really don't understand this kind of test and why do you record the sts output from the terratec when you can use the asio destination....

It´s as simple as recording the real latency, that´s the one you hear, the analog one. The recording is just to compare on a stereo file, the time between the mic, and the final sound you hear, so is impartial really.
latency has variations, is not 4.5 just.
from 4.5 to 6ms on Terratec.
from 7-8 on STS.

lagoausente · Post by **lagoausente** » Sat Apr 08, 2006 9:14 am

That is still low latency. Isn't it ?
Particularly since it claims to be a 'total cycle measurement'

Yes is total cycle measurement. Is low, but no better than sofware while it´s selled as that.
I measured the Roland latency, recording in audio the midi signal. When send the first bit, you can see on the waveform, is about 1 ms from audio to midi.
But in this test is no important, since is a A-B comparison where the Roland is present just the same on both.

djmicron · Post by **djmicron** » Sat Apr 08, 2006 9:24 am

the real bad thing in this test is the audio to cv conversion of the roland machine,
it can not be reliable.

[ This Message was edited by: djmicron on 2006-04-08 10:25 ]

lagoausente · Post by **lagoausente** » Sat Apr 08, 2006 8:05 pm

On 2006-04-08 10:24, djmicron wrote:
the real bad thing in this test is the audio to cv conversion of the roland machine,
it can not be reliable.

[ This Message was edited by: djmicron on 2006-04-08 10:25 ]

I thing the really bad thing is you don´t like what I say.
First, I have several tests, with similar results, (perhaps Roland has an internal chip that detects when I´m testing creamware one to sabotage it).
Second, on both sampler measurement are present the same, Roland, and D/A.
Third, I have found really no constructive discuss here, just "NO", "It´s your are wrong" "Creamware has Zero latency", just repeat the same, without thinking first about it.
If you have no interest, while reply??
If you have interest, test it yourself.

[ This Message was edited by: lagoausente on 2006-04-08 21:06 ]

djmicron · Post by **djmicron** » Sun Apr 09, 2006 7:08 am

i've not said that "creamware has zero latency", but i'd like to understand your test.

You talk about this test, but you don't talk about your system configuration (this could be very important for a test).
I still don't understand this test and don't agree the way you've done it, but it could be true that the sts has 8 ms latency.

lagoausente · Post by **lagoausente** » Sun Apr 09, 2006 7:31 am

8 ms is not STS latency, is my setup latency, were is inserted the Roland, so STS latency would be less. The test is really very simple, results are just an A-B comparison usin the same things.
Is as simple as measure the time between the trigger and the final sound you hear from the analog output. For do this, just need to record, on another soundcard, both signal, the trigger one and the final sound.
Just this:
1)Trigger-creamware midi in-sts-creamware analog out.
2)Trigger-Terratec midi in- cubase sx (vsti)- Terratec analog out.

Just record on another soundcard the trigger signal and the analog output for creamware and measure the time diference.
Then record the trigger signal and the terratec analog out, and measure.
Results are near the same, only 2 ms of diference o less.
STS is better, since is more estable, and have no drops, even latency variations are lower.
On vSTi one, varies from 4.5 to 6.
On STS varies less. Most times 7 ms, and worst case 8ms.
I have found STS quite good because has no CPU involved, and so don´t must take care about Asio or any drops when adding more tracks of effects like in Cubase.
Anyway, I think STS could have better latency. Is extrange what I ask in the other thread about the midi port.
Usin STS, Routing midi to sequencer first, and then to STS, gives the same latency result, as routing directly form mini in to STS, what looks like if really, the Midi in A module, is routed from windows, instead directly in the card, from the midi-in directly to dsp, what most of us here have been thinking.

lagoausente · Post by **lagoausente** » Sun Apr 09, 2006 7:34 am

Have to add that for having no drops on vsti on real working, probably would must increase the Asio buffer a little, so the real performance of both would be near similar, or perhaps worse on VSTi, depending on the Cpu use.

[ This Message was edited by: lagoausente on 2006-04-09 08:35 ]

Shroomz~> · Post by **Shroomz~>** » Sun Apr 09, 2006 8:57 am

"STS latency test, worse than Sofware sampler."
(that's the title of your thread)

Personally, I think comparing the latency of STS to a software sampler would be more relevant if the terratec card wasn't involved.

I also think that the test could benefit significantly from the inclusion of results using various HARDWARE samplers (Akai, Emu etc)

I wouldn't use the microphone triggering the Roland either in this test, but whatever floats you boat.

symbiote · Post by **symbiote** » Sun Apr 09, 2006 12:31 pm

Seriously lagoausente, your signal chain is way too long and way too broken to measure anything useful. All it does, like you say, is test the STS latency with your current setup, which is hopefulyl fairly unique, and of no use to anyone else.

Sorry if this seems a bit harsh, but your thread's title is totally misleading and false.

What you are measuring, from you description of your signal chain in your second post, is:

SPD11 Audio to Midi latency +
SPD11 Midi out to CW Midi in latency +
STS Latency +
CW DAC Latency

And your comparison with the software sampler is again meaningless, as Shroomz points out, since you use different kinds of DAC (Terratec vs CW) and a different software setup (SFP vs Cubase SX.)

If you want to make a precise measurement of an element of a signal chain, like STS latency, you need to setup a chain, lock all the elements into place, and ONLY vary the element you want to measure.

Also, you need a pretty clean and rigorous description of the setup and testing procedures, so that other people can test it out also, since tests like these are worthless without peer review.

Mr Arkadin · Post by **Mr Arkadin** » Sun Apr 09, 2006 2:21 pm

i'm still unsure as to what you're actually trying to achieve.

clops · Post by **clops** » Mon Apr 10, 2006 12:32 am

I´ve been watching this discussion over the last few days and I´m wondering what good it is.
Why did you start this test. Did you feel the latency when playing the STS? To me when triggering it with my DDrum it feels good in terms of timing, much better then for example a Yamaha Dtxtreme with its internal sounds.
On the other hand we are talking here about a delay between the two samplers which is equal to being away about a meter from a sound source.
But If your intention is to show that creamware isn´t right about the specifications the claim their products have, then I agree. Not only I´m still waiting for the directX path in the waveeditor the promised with the first release of the STS4000, some of the effects add quite a bit of latency, what can sound quite strange when mixing down.

lagoausente · Post by **lagoausente** » Mon Apr 10, 2006 10:46 am

On 2006-04-10 01:32, clops wrote:
IBut If your intention is to show that creamware isn´t right about the specifications the claim their products have, then I agree.

It was just that. As I have told, it feels good playing STS, but I think it could be better. Still extrange for me that routing midi to sequencer first give the same latency as routing directly to STS, what makes me think if would really go directly, would have a few ms less, since everybody knows that when going midi to windows, a driver latency occurs.
Regargind for the test, I hope anyone more make it´s own and tell us about, perhaps more accuracy test than mine. I still don´t understand the bug. Even they use diferent D/A, I really don´t understand what importance it has. The STS is done for be listened from his D/A, and the audio-midi of the Roland is the same on both tests. The test may be no useful for you, but I just have to say, that as I understand, it measures what you really hear, the time between you play and hear, what is the really important, isn´t?