Close in jitter?

[DonH56]

Hmmm... I put in 90+ hours last week and this one is not looking any better. If and when I get my life back I could compare distortion measured with my CSL UMIK-1 compared to my Earthworks M30.

But wait -- Amir, don't you have an M40, you sly dog? And all that free time?

 

[amirm]

But wait -- Amir, don't you have an M40, you sly dog?

If you mean the M40 rifle, you have me confused with Sal.

If you mean microphone, I have the cheap USB mic plus the fancy one that came with my Audyssey Pro kit.

 

[Blumlein 88]

I did another run using the Avantone CK1 at half a meter from the speaker and used 32 averages to get a cleaner result. I haven't measured jitter on the DAC in use, but this is looking pretty reasonable. Guess I should measure the jitter and see what the result is for this DAC.

 

[DonH56]

If you mean the M40 rifle, you have me confused with Sal.

If you mean microphone, I have the cheap USB mic plus the fancy one that came with my Audyssey Pro kit.

Amir, you need an Earthworks M40, or maybe an M50, just for bragging rights to top my el cheapo M30. Take some of the funds you've been wasting on cheap DACs and get yourself a good mic and preamp. Heck doesn't even take a really fancy preamp, though for a guy with his own personal AP... Which reminds me, AP came out with their own line of calibrated mics recently that may be worth a look.

I did some quick comparisons between the two mics a year or two ago. IIRC the FR was very similar, almost identical, after applying the CSL cal file. The Earthworks had much greater dynamic range with lower noise and distortion but I didn't keep any plots, at least that I can find. At the time I just wanted to verify the FR was reasonable and get my new pre/pro set up and running.

 

[Blumlein 88]

Here is an electronic measure of the HK avr itself. It has a number of spikes spaced at 120 hz either side of the signal. So basically those you see from the speaker are those spikes that poke above the in room noise floor on the measures with the microphone. Appears speakers, analog preamps and my trusty Wyred4Sound class D amp aren't bottlenecks for jitter at these levels at least.

Moral of the story is don't use your lousy Umik1 for acoustical jitter measures. This also leans in the direction that doing jitter measures out of preamps or DACs is fine as the gear after it in the chain is not a limiting factor. In room noise of jitter side effects is the limiting factor if your DAC isn't Schiit.

 

[amirm]

Amir, you need an Earthworks M40, or maybe an M50, just for bragging rights to top my el cheapo M30. Take some of the funds you've been wasting on cheap DACs and get yourself a good mic and preamp. Heck doesn't even take a really fancy preamp, though for a guy with his own personal AP... Which reminds me, AP came out with their own line of calibrated mics recently that may be worth a look.

I hear you . It is just that for acoustic work I don't believe in anything fancy when it comes to mics. But your point about bragging rights makes sense.

Yes, I saw the microphone from AP at CES. It was pretty expensive from what I recall though.

BTW, I just ordered a $2,600 canon lens (11-24 mm zoom). So "funds" even for me are limited for the rest of the month.

 

[DonH56]

Off-topic: I can't bear to talk cameras. I finally broke down and bought a decent DSLR a few years ago (Canon 6D) with a couple of decent lenses (Canon 70-200 f4/L, 24-105 f4/L, and the Tamron 200-500 f5-6.3 'cuz I couldn't afford the big Canon). Not in your league but decent. It came the day we were packing everything up to evac due to wildfires so it went into storage. Have taken it out of the box a couple of times since but never really used it.

 

[amirm]

I hear you. I gave up on photography for some 20 years. "Life happened" with kids and such. Picked it back up when digital came to form but even now I don't use it nearly as much as I like.

Having bought my RV, I am getting more use out of it.

 

[March Audio]

I hadn't thought about that too hard, but, yeah...

For the MartinLogans the path after the preamp is all analog, for the JBL, I suppose there's a re-digitization before the Class-D amps.

Maybe it's an artifact of the microphone and its ADC. It doesn't show up on the preamp output, so it isn't a problem at the PC or at REW.

We need somebody else to output a 12kHz tone and show their 90dB speaker output capture not using a USB mic.

JBL
(reddish) vs MartinLogan (bluish)

View attachment 6938

Hi Ray

It may have to wait until tomorrow but I can repeat with my UMIK, a Rode NC5 and the output of a Bruel and Kjaer SLM.

 

[Jakob1863]

According to the data sheet the noise level of the Umik-1 is <=-74 dBFS (at max IPGA), sampling rate seems to be fixed at 48 kHz, the ADC is a 24 bit resolution device. So that information fits roughly the measured spectra, although i couldn´t find real values for the variables (like max. IPGA).

I always find it hard to understand that even specifications for measurement devices were today so ambiguous.......

 

[Blumlein 88]

According to the data sheet the noise level of the Umik-1 is <=-74 dBFS (at max IPGA), sampling rate seems to be fixed at 48 kHz, the ADC is a 24 bit resolution device. So that information fits roughly the measured spectra, although i couldn´t find real values for the variables (like max. IPGA).

I always find it hard to understand that even specifications for measurement devices were today so ambiguous.......

Reading some posts on the minidsp forum, the Programmable Gain amp has a max value of 36 db of gain. It is set with some internal DIP switches. Normally it is not set at this value. Doing so would raise the noise floor in use and reduce max input. I too wonder about what really that -74 dbfs spec means. I did find where someone with access to an anechoic chamber measured one. In room noise was said to be 9 db-spl in the room. This what they had to say and the chart.

The A-wheighted noise of the UMIK-1 (700-0756) was 29.1 dB(A) and 43.1dB(unwheighted), which is both absolutely OK (a lower noise floor would limit the max. SPL). Please find attached the measured noise-spectrum.

The spikes at 1 khz were a problem with earlier ones and said to be reduced in versions for the last 3 years (this chart was from about 4 years ago).

In general use it seems to do what one needs for measuring speakers. For the price and simplicity of use I am satisfied with mine. Self noise floors of small mics are worse than larger mikes. The Avantone CK1 pencil condenser I used earlier has an A weighted rating of 17 db. So this smaller Umik being something like 29 db A is not out of line. The Behringer ECM8000 has a noise floor of something like 23 db A. Some expensive mikes for measurement are also in the mid 20 dbA range. Plus as you see most of the noise is below 500 hz.

 

[Jakob1863]

Reading some posts on the minidsp forum, the Programmable Gain amp has a max value of 36 db of gain. It is set with some internal DIP switches. Normally it is not set at this value. Doing so would raise the noise floor in use and reduce max input. I too wonder about what really that -74 dbfs spec means. I did find where someone with access to an anechoic chamber measured one. In room noise was said to be 9 db-spl in the room. This what they had to say and the chart.

The A-wheighted noise of the UMIK-1 (700-0756) was 29.1 dB(A) and 43.1dB(unwheighted), which is both absolutely OK (a lower noise floor would limit the max. SPL). Please find attached the measured noise-spectrum.

The spikes at 1 khz were a problem with earlier ones and said to be reduced in versions for the last 3 years (this chart was from about 4 years ago).

In general use it seems to do what one needs for measuring speakers. For the price and simplicity of use I am satisfied with mine. Self noise floors of small mics are worse than larger mikes. The Avantone CK1 pencil condenser I used earlier has an A weighted rating of 17 db. So this smaller Umik being something like 29 db A is not out of line. The Behringer ECM8000 has a noise floor of something like 23 db A. Some expensive mikes for measurement are also in the mid 20 dbA range. Plus as you see most of the noise is below 500 hz.

Just in case, my critique wasn´t related to the performance of the device but to the specifications given. The mentioned 1 kHz spikes of the earlier versions could be related to the usb clock mode, while the spikes around the 12 kHz in RayDunzl´s graphs look more like mains related.

A comparable measurement by RayDunzl from the loudspeaker posts (comparable to the preamp out measurement) would be useful, but be careful with the much higher level, usually an appropriate voltage divider must be used.

 

[Cosmik]

The spikes at 1 khz were a problem with earlier ones

USB data packets are sent at 1ms intervals - this is a jittery transfer. In non-asynchronous DACs, they have receiver chips that use a PLL and buffer to adapt to the incoming packet rate to reduce jitter. On the other hand, what does a PC do when it receives data from a mic?

If it is simply recording and allowing the mic to provide the sample clock, it could just accumulate the packets into a file - no problem there. But if the app is setting the sample rate, some form of adaptation/stretching/interpolation is needed. In Linux, various layers in the audio hierarchy will do this without the user necessarily knowing about it. Various grades of resampling algorithm are available with the best using the most processing power; the poorest algorithms would exhibit what looked like aliasing spikes. In Windows, the commercial Virtual Audio Cable does resampling without the user needing to know about it.

In the case of REW which is usually used to correlate input with output, some jiggery pokery like this would be mandatory or else phase measurements and the like would be corrupted.

 

[Cosmik]

I think it is apparent that the topic of sample clocks is deeply misunderstood... I don't understand how it all works, but at least I know that I don't understand it!

The biggest misunderstanding is the one where people assume that if two devices are set to a sample rate of 48 kHz you can just mix them together. Of course you can't, because they're not locked together and are drifting, relative to each other. A triumph of software like JACK in Linux, or Creative Soundblaster cards (and others) is that they use resampling algorithms to allow you to mix digital streams with different sample clocks as though you are mixing analogue signals.

However, for the ultimate 'bit perfect' hi-fi, if you aren't mixing streams together, you want to bypass this resampling. How many systems are set up with resampling in operation where people don't realise it? And alternatively, how many are generating glitches occasionally because resampling should be being used but isn't?

 

[Blumlein 88]

Just in case, my critique wasn´t related to the performance of the device but to the specifications given. The mentioned 1 kHz spikes of the earlier versions could be related to the usb clock mode, while the spikes around the 12 kHz in RayDunzl´s graphs look more like mains related.

A comparable measurement by RayDunzl from the loudspeaker posts (comparable to the preamp out measurement) would be useful, but be careful with the much higher level, usually an appropriate voltage divider must be used.

I think the measures from my other microphone shows the Umik to have enough jitter it shouldn't be used for that purpose. So I don't know what measuring at the loudspeaker posts would accomplish. Ray's amp is known to be pretty clean itself and we already have the measures at the pre-amp output. You would of course be limited by in room noise for very low level jitter sidebands. Yet other than that it appears the microphones that don't have the issues of the UMIK1 are showing a very sharp quarter sample rate response. You also can see from my electronic measures of the AVR I used at its pre-amp outputs that the random jitter wasn't too bad (base of the tone was fairly narrow) while there were some considerable levels of jitter at 125 hz intervals. I could repeat that with a much cleaner DAC/preamp. What the microphone measures do show is what in theory I would have hypothesized which is sidebands of extremely low jitter or even moderately low jitter are buried underneath room noise.

 

[Jakob1863]

@ Blumlein88,

it´s the same reason why a measurement taken at the preamp out is useful....
Step by step, under the same conditions to gain further insight about what´s going on.

 

[RayDunzl]

A comparable measurement by RayDunzl from the loudspeaker posts (comparable to the preamp out measurement) would be useful

The ClassA amplifiers (and the speakers) were eliminated from the circuit, replaced with the ClassD amplifiers inside the JBL LSR 308, with similar results sent by the microphone.

Bluish - Krell and MartinLogan, Reddish - JBL LSR 308 (active, but adding another ADC) - both fed from the measurably clean preamp outputs.

 

[Jakob1863]

The ClassA amplifiers (and the speakers) were eliminated from the circuit, replaced with the ClassD amplifiers inside the JBL LSR 308, with similar results sent by the microphone.

Bluish - Krell and MartinLogan, Reddish - JBL LSR 308 (active, but adding another ADC) - both fed from the measurably clean preamp outputs.

View attachment 6958

Yeah, i´ve seen that, but nevertheless the additional measurement would make sense.....

 

[Brad]

Other than just sending a tone, I vaguely recall a DIYaudio thread about performing a J-test measurement using a j-test wave file and wavespectra. I think I grabbed the file, and brielfy played with wavespectra, but has anyone performed this measurement in loopback (or with speakers)?
If so, is the process as described?
I would like to characterise my Motu 1248

 

[Blumlein 88]

Other than just sending a tone, I vaguely recall a DIYaudio thread about performing a J-test measurement using a j-test wave file and wavespectra. I think I grabbed the file, and brielfy played with wavespectra, but has anyone performed this measurement in loopback (or with speakers)?
If so, is the process as described?
I would like to characterise my Motu 1248

Loopback won't tell you much of anything. Same clock will be used for clocking out and in so any jitter will be hidden. Your ADC and DAC clocks are effectively locked this way. Think about it, if a sample is clocked out early in loopback the sample of that is clocked in early too. Then if the next one is late, the one taken in loopback is also late by the same amount. It will give a result that might have noise, but appears to have zero jitter. Yes I tried it too.

The problem with using a separate DAC is unless something like a piece of measuring gear you can only say the two bits of gear had some combined amount of jitter. If the result is pretty good, then obviously both pieces have to be pretty good. If one is good and another is not you won't necessarily know without some outside reference. In my case, I have three DAC/ADC pairs that all give a pretty decent Jtest result though each differs slightly. I have also done this with smartphones or HDMI switchers or portable CD players and gotten rather poor Jtest results and I know it is these other devices that have the problem.

All you need do is play the Jtest file from one device and record it with another. Then use wavespectra or other software to examine the result in an FFT.

I wasn't familiar with exactly what the Motu 1248 is. I looked it up and now I am jealous. One would expect it to have very low jitter. So finding something else with low enough jitter to find out just how low is probably not an easy task. Benchmark's current DACs can put out an incredibly low jitter signal. If you know someone with one of those and can measure the Jtest with the Benchmark as a source you should learn something worthwhile.