That's entirely reasonable Jens. It's also standard practice in the test equipment world. It's the accessing of the internals that voids the calibration, not the added options.
-Chris
-Chris
I'm still contemplating of adding myself to the GB list...
One more question I can't find an answer for from spec or via search. Is the USB soundcard part of this unit basically USB 2.0 audio compliant - ie would it work without driver in OSX? Also the new Win10 creators update should support the USB 2.0 audio devices, so looking into future that would be a benefit.
One more question I can't find an answer for from spec or via search. Is the USB soundcard part of this unit basically USB 2.0 audio compliant - ie would it work without driver in OSX? Also the new Win10 creators update should support the USB 2.0 audio devices, so looking into future that would be a benefit.
Yes, it is USB Audio 2.0 compliant. The driver supplied with the unit is this:
USB Audio 2.0 Class Driver for Windows
And yes, it works without a driver with OSX.
For Win 10 the driver may also become unnecessary, when USB Audio 2.0 becomes part of Windows.
USB Audio 2.0 Class Driver for Windows
And yes, it works without a driver with OSX.
For Win 10 the driver may also become unnecessary, when USB Audio 2.0 becomes part of Windows.
What is the recommended cal interval?
Will the cal procedure be available to a third party cal lab or must the unit be returned to RTX?
Calibration of what, where?
Voltage levels?
I tend to calibrate these things myself if I really feel like it's needed, and apply my calibration results in the settings of the audio analysis software. I prefer this approach as I feel it is more reliable and more transparent to me than some external lab procedure that I don't really know. (I am not saying external lab data are not accurate.)
Hi mbrennwa,
If you give the calibration lab the specs and go for complete data, you should have 5 ~ 6 readings per function per range. That would be more than suitable.
-Chris
If you give the calibration lab the specs and go for complete data, you should have 5 ~ 6 readings per function per range. That would be more than suitable.
-Chris
Hi Jens,
Do you have a specification of the absolute amplitude accuracy at 1kHz and
the amplitude flatness in the audio band?
Thanks,
Udo
Do you have a specification of the absolute amplitude accuracy at 1kHz and
the amplitude flatness in the audio band?
Thanks,
Udo
This approach is acceptable for diy where you may not be concerned whether or not your test equipment meets a certain standard (the manufacturer's published spec) but is completely unacceptable in the commercial world.
I spend thousands a year maintaining calibration on the company test equipment. I have not had my home lab equipment calibrated in over a decade - but I can perform a verification if I choose to against the calibrated gear in the commercial lab
The advantage of RTX's involvement in this project, as I see it, is that
(a) the project gets over the line for which they should be applauded
(b) there is the possibility of future availability and support
Critical to the notion of support is providing a cal service and/or providing a cal procedure. The latter is more relevant to me because I am on the other side of this little planet we call home and freight is $diabolical.
That is what you get when you buy from the likes of Keysight (or whatever they are calling themselves this week) - I am interested in finding out where Jens/RTX plan to place themselves on the spectrum between "oneoff/diy kit" and Keysight/Audio Precision.
Or put it another way
- would I buy for the home lab?
- would I recommend a purchase for the commercial lab?
Jens gave me the OK to acknowledge that I have one of the preproduction units here and have been putting it through its paces. The physical execution from the outside is on a par with AP or the other premium test equipment vendors. I have not opened it up and its not really clear how the box opens (a good thing if you want to keep fingers out).
The only times I have had issues were all traceable to "cockpit error". This has the best performance from an AK5394A I have encountered, on a par with the tweaked AKD5394 demo board but much more flexible. His analog input circuit seems really good. Its mixed discrete and IC I believe.
I'll try to post a few measurements from this sample. I have used it with Praxis, ARTA, Audiotester so far. I have used it in both WMA and ASIO and as long as the host PC doesn't get involved the software all shows the same info.
Here are a few measurements from the RTX6001:
First is loopback at 1V
Second is Victors oscillator to RTX balanced with the oscillator floating (and some 60 Hz common mode)
Third is Victor oscillator with the ground tied to the ground reference at the XLR connector
Forth is an HP 467 power amp driving about 10V into 16 Ohms. This amp was made in the mid 1960's and has a power bandwidth of 1 MHz. Its an example of what you might actually be measuring in terms of less than perfect.
The only times I have had issues were all traceable to "cockpit error". This has the best performance from an AK5394A I have encountered, on a par with the tweaked AKD5394 demo board but much more flexible. His analog input circuit seems really good. Its mixed discrete and IC I believe.
I'll try to post a few measurements from this sample. I have used it with Praxis, ARTA, Audiotester so far. I have used it in both WMA and ASIO and as long as the host PC doesn't get involved the software all shows the same info.
Here are a few measurements from the RTX6001:
First is loopback at 1V
Second is Victors oscillator to RTX balanced with the oscillator floating (and some 60 Hz common mode)
Third is Victor oscillator with the ground tied to the ground reference at the XLR connector
Forth is an HP 467 power amp driving about 10V into 16 Ohms. This amp was made in the mid 1960's and has a power bandwidth of 1 MHz. Its an example of what you might actually be measuring in terms of less than perfect.
Attachments
Both are more difficult to spec than you may think, especially when the software is third party. There is a lot of confusion about what 0 dBFS means even.
Also there is source accuracy vs. receive voltage accuracy. I'll try to get some idea but realistically this is not a substitute for a Fluke 5790 or a Fluke 792. I think I have transfer accuracy of about .02% with what I have here. What is your goal?
I ran into the same question when measuring flatness in my autoranger. If you have a reasonable wideband, but not hi-end, RMS multimeter you can do as follows:
- measure input and output voltages at say 10kHz, call it A and B;
- measure input and output voltage at say 50kHz, call it C and D;
- the deviation X of the 50kHz level wrt 10kHz is: X = B/D * C/D.
You can convert to dB if you want. It sounds more difficult than it really is with a spreadsheet doing the legwork. The autoranger will have a spreadsheet for user calibration.
Jan
I wholeheartedly disagree. It's not about being commercial or not. It's about being able to do the calibration yourself (availability of suitable reference instruments, ability to use them). And it's about taking responsibility. Either you rely on your own abilities, or you delegate the responsibility to someone else. If I do the calibration myself and something is wrong with my measurements because my calibration was not right, I have to blame myself. If I rely on some external lab calibration, it's easy to argue that they screwed up.
Demian thanks this is very useful. Don't you find it odd that there is a significant difference in H3 between case 2 and case 3 (with the same osc )
Demian,
Thanks for doing the measurement with the Victor Generator.
The THD+N is about 102 dB typical with -3 dBFS input at 1 kHz?
Could you repeat the measurement with 0 dBFS and with higher frequencies (20 kHz) ?
Thx,
Udo
Last edited:
I only want to know what the measurement uncertainty is.
This is important for measuring amplifier transfer function.
For about $1000 i expect reliable specifications (they must not be great).
AP specifies 0.35% absolute accuracy at 1kHz and 0.1% flatness from 5Hz - 20 kHz (0.05% typical).
Prismsound DScope II specifies 0.7% absolute accuracy at 1kHz and
1.1 % flatness from DC - 20kHz.
Case 2 is balanced: THD+N = 102 dB
Case 3 is unbalanced: THD+N = 104 dB
Both are at 1kHz and -3.4 dBFS.
Maybe this explains the difference in H2 and H3?
This is specified in the (preliminary) specification provided in the first post of the GB section.
The measurement accuracy is specified as < +/- 0.05dB (+/- 0.57%).
The analyzer flatness is specified as +/- 0.01dB (+/- 0.11%) from 20 Hz to 20 kHz.
Of course this assumes that the analysis SW doesn't introduce errors.
@VivaVee
Calibration is fairly simple and requires only a good voltmeter and a screwdriver. It might make sense to offer both options, calibration by RTX or user calibration (following a calibration procedure).
Regarding recommendation about how often it should be calibrated, this is still open.
Considering the chain of calibration equipment that also need periodic calibration. I never could figure out whether there is an ultimate source that does not need calibration by nature, or whether just the statistics of this chain of calibration reduces the error to a certain level.
Sent from my iPhone using Tapatalk
Sent from my iPhone using Tapatalk