That's simple, just accept that the RTX specification of 5nV/rtHz is real and you can measure most real life amplifier's noise without using a LNA.
I've built a 1000x LNA that can measure down to 390pV/rtHz, but the RTX has made it redundant except for measuring individual components (e.g. a 2SK170 is around 1.2nV/rtHz, a 1k resistor is 4nV/rtHz)
That's no surprise. I wrote that software, so I must like it
Take a look at the MATAA website. There is also an audiXpress article about MATAA.
My Mac statements are because I have an old semi-broken Mac in my workshop, and I use this to run MATAA. But MATAA runs just as well on Linux or Windows. MATAA is basically an audio analysis toolbox for use with Matlab or GNU Octave. If you're not afraid of Matlab or Octave, you may like MATAA (and it's free, as in free beer and in free speech.) Otherwise I, well, aehm, wish you good luck with it.
I think the question is, if the software does not report consistently real information, then how can one measure noise if the DUT? Given the fact that different settings will produce different results, there must be some standard FFT size, no averaging or some specific average number, etc as a universal standard to yield same results reference.
Assuming that the analyzer SW calculates it correctly that universal standard is the RMS value. The RMS value should be independent of the FFT size.
Getting a low noise floor in an FFT analysis is useful if you are looking for very low levels of harmonics or other discrete frequency signals, which would otherwise be hidden below the noise floor.
Getting a low noise floor in an FFT analysis is useful if you are looking for very low levels of harmonics or other discrete frequency signals, which would otherwise be hidden below the noise floor.
Jens,
Yes a low noise floor with large FFT window helps see low level harmonics.
But in general how to know if the noise levels of a design are dominating its harmonic distortions if changing the FFT size and window changes the apparent noise level.
Does the THD+N also depend on FFT window? Are you aware of SW which give accurate representation of noise as RMS values over some specified bandwidth?
Yes a low noise floor with large FFT window helps see low level harmonics.
But in general how to know if the noise levels of a design are dominating its harmonic distortions if changing the FFT size and window changes the apparent noise level.
Does the THD+N also depend on FFT window? Are you aware of SW which give accurate representation of noise as RMS values over some specified bandwidth?
Last edited:
The software does report consistent and real information, issue is how to interpret them correctly.
Typically noise of a device is specified as a RMS value over a certain bandwidth, or as noise spectral density in dbV/rtHz. Both won't change because of FFT size, you just need to understand how to compute them from the FFT results you've got.
Some software, like ARTA (REW cannot), can display the vertical scale in power spectral density (dBV/rtHz), which makes measuring noise more intuitive.
That's a useful feature, especially if calibrated. One could make a pretty simple test fixture for measuring transistor and IC noise with it.
Determining the input referenced self noise of the RTX is not that difficult. You need an indication of its noise (FFT etc. or similar) and a pot good for maybe 10K. Log taper probably better. Measure the noise with the input shorted. Then replace the short with the pot and raise the resistance until the noise increases by 3 dB. Measure the resistance of the pot and refer to this: Thermal Noise Calculator The pot resistance is the equivalent to the input noise of the DUT. This is essentially how the Quan-Tech semiconductor noise testers are calibrated.
Well, Matthias, this sounds good, for not to say promising, to me!
Then, is MATAA a stand-alone SW or does it depend on Matamp?
Best regards!
Matamp? Do you mean Matlab? Yes, MATAA needs either Matlab or GNU Octave. I personally prefer Octave, and I haven't used Matlab for years now. Octave is free software, and it has great support by the community.
Hello
I ve received it and couldn't be more happy
I confirm that with iMac and mac osx and latest REW 5.19 rev 8, the max rate is 96k and this has to be put also in the audio/midi preference. I could nt find solution to change the "16bit" info down the screen despite the changes made in audio/midi configuration, but 24bits creates a problem. I m now setup with 32 bits 96k in the audio config and 96k in new and all fine.
Some points/question about noise measurement,
to 1audio, I did your test with a 20k pot log. of course not easy nor accurate because this raise the 50/60hz. but let say that I got 3db above rms noise floor with 8k6 (ish) and 6 db above NF with 13k4 ish
thermal noise for 8k6 is 1.6uV and for 13k4 is 2uV.
any good conclusion ?
but ok, now I m trying to understand noise measurement and would need some help/advise.
assuming I m using 0dbv (1V) for output and input.
with -3dbfs on REW generator, I got 1V rms between xlr+ and xlr-, and with loopback this gives -3dbFS in RTA/FFT, fine.
when I stop the generator, RTA shows -100,8dbFS. so I assume the rms noise measured is 100.8-3 = 97.8db below 1v which is 13uVrms, is that right?
another question is about impact of the gain selector on the input. What is the gain which is the more transparent for the ADC, in term of noise. I guess 3.16 as this is closer to AK5394 input range ?
thx
I ve received it and couldn't be more happy
I confirm that with iMac and mac osx and latest REW 5.19 rev 8, the max rate is 96k and this has to be put also in the audio/midi preference. I could nt find solution to change the "16bit" info down the screen despite the changes made in audio/midi configuration, but 24bits creates a problem. I m now setup with 32 bits 96k in the audio config and 96k in new and all fine.
Some points/question about noise measurement,
to 1audio, I did your test with a 20k pot log. of course not easy nor accurate because this raise the 50/60hz. but let say that I got 3db above rms noise floor with 8k6 (ish) and 6 db above NF with 13k4 ish
thermal noise for 8k6 is 1.6uV and for 13k4 is 2uV.
any good conclusion ?
but ok, now I m trying to understand noise measurement and would need some help/advise.
assuming I m using 0dbv (1V) for output and input.
with -3dbfs on REW generator, I got 1V rms between xlr+ and xlr-, and with loopback this gives -3dbFS in RTA/FFT, fine.
when I stop the generator, RTA shows -100,8dbFS. so I assume the rms noise measured is 100.8-3 = 97.8db below 1v which is 13uVrms, is that right?
another question is about impact of the gain selector on the input. What is the gain which is the more transparent for the ADC, in term of noise. I guess 3.16 as this is closer to AK5394 input range ?
thx
mlloyd1,
It is ready for sale. Please send a mail to sales@rtx.dk
Quantities are limited at this point, but we do have a few available.
Jens
Glad to hear that you are happy with the RTX6001.Hello
I ve received it and couldn't be more happy
I confirm that with iMac and mac osx and latest REW 5.19 rev 8, the max rate is 96k and this has to be put also in the audio/midi preference. I could nt find solution to change the "16bit" info down the screen despite the changes made in audio/midi configuration, but 24bits creates a problem. I m now setup with 32 bits 96k in the audio config and 96k in new and all fine.
Some points/question about noise measurement,
to 1audio, I did your test with a 20k pot log. of course not easy nor accurate because this raise the 50/60hz. but let say that I got 3db above rms noise floor with 8k6 (ish) and 6 db above NF with 13k4 ish
thermal noise for 8k6 is 1.6uV and for 13k4 is 2uV.
any good conclusion ?
but ok, now I m trying to understand noise measurement and would need some help/advise.
assuming I m using 0dbv (1V) for output and input.
with -3dbfs on REW generator, I got 1V rms between xlr+ and xlr-, and with loopback this gives -3dbFS in RTA/FFT, fine.
when I stop the generator, RTA shows -100,8dbFS. so I assume the rms noise measured is 100.8-3 = 97.8db below 1v which is 13uVrms, is that right?
another question is about impact of the gain selector on the input. What is the gain which is the more transparent for the ADC, in term of noise. I guess 3.16 as this is closer to AK5394 input range ?
thx
To get the lowest input referred noise with the RTX6001, you should select the -20 dBV input range. With this setting there is a relatively high gain in the input amplifier, so the noise is dominated by the amplifier, not the ADC.
The 10 dBV input setting gives you the most "direct" connection to the ADC, with the lowest amplifier gain and no attenuator in front of it. With this setting the noise is a combination of the noise from the input amplifier and the noise from the ADC. Actually it is dominated by the noise of the ADC.
When selecting +20 dBV input level and higher there will be some thermal noise from the input attenuator.
The 10 dBV input setting will generally also give you the lowest distortion.
Setting the output to 0 dBV should give you a 707 mVrms signal at -3 dBFS, not 1 V rms.
The -100.8dBFS you have measured indicates that your interface is running at 16 bit. I just tried it with audioTester and got -101.25 dBFS with the 0 dBV/0 dBV setting using a 16 bit interface. Using a 24 bit interface gave me a noise level of -115.6 dBFS.
This is the combined noise of the output and the input, which in my case is around 1.6 uV.
With a short circuit on the input, and 0 dBV input range, I measured -119.26 dBFS, which is equivalent to around 1.1 uV (20 kHz BW).
With a short circuit, and -20 dBV input range, I measured -105.9 dBFS, which is equal to 0.51 uV (20 kHz BW). This is equivalent to around 3.6 nV/rtHz.
thank you for your very clear and precise answer.
in REW, by default when you put -3dbFS on the generator, it produce the maximum sine wave, so 1Vrms for sure but you can change this setting in REW preference so that the max sine wave is generated for a value of 0db.
then I tried to use the RTX and your usb driver in a VM W7-64 (virtual box) and it says that the device driver signature is not OK and I should install MS KB3033929...but of course this doesn't work. Any feedback or similar experience welcome . anyway I ll try on my W7 laptop later.
EDIT: work fine on my laptop and I have the same -115.6 at the 0dbv
EDIT2: and all this story about 16/24bit is somehow described in the other technical thread ! sorry for redundancy
in REW, by default when you put -3dbFS on the generator, it produce the maximum sine wave, so 1Vrms for sure
but you can change this setting in REW preference so that the max sine wave is generated for a value of 0db.then I tried to use the RTX and your usb driver in a VM W7-64 (virtual box) and it says that the device driver signature is not OK and I should install MS KB3033929...but of course this doesn't work. Any feedback or similar experience welcome . anyway I ll try on my W7 laptop later.
EDIT: work fine on my laptop and I have the same -115.6 at the 0dbv
EDIT2: and all this story about 16/24bit is somehow described in the other technical thread ! sorry for redundancy
Last edited:
- Status
- Not open for further replies.