For years I was using stand alone gear. Now the GB of rtx6001 offered the opportunity of using an excellent piece of sound card and with some sw + pc to do almost the same thing. I'm not too much familiar to this way of measurements so a wiki tutorial would be very helpful.
Last edited:
Ok guys, I started a WIKI page here on diyAudio. Please pick it up and put some life in it by adding text, photos, drawings, plots, etc. Show off how things work, and what does not work, and why. Use real-world examples. Don't replicate the user manual.
The page is here. I put it under Wiki --> Electronics --> Tools And Equipment
The Wiki service responds as "Closed" for now.
I'm all in favor. I would contribute but have too much paying works to even set foot in my shop for at least another month.
As a starting point Tek published a short tutorial on audio measurements (http://lcweb2.loc.gov/master/mbrs/r...ektronix Cookbook of Standard Audio Tests.pdf) complete with audio geek from the era. There are others from various sources including National Semi focused on measuring switching amps. I would start with one of those and run through how to do a typical battery of tests then some specifics like optimizing bias vs. distortion, debugging clipping behavior or finding noise sources.
If someone local was interested I could provide my lab and help walk a neophyte through these tests. Would make a good Youtube sequence.
I'm all in favor. I would contribute but have too much paying works to even set foot in my shop for at least another month.
As a starting point Tek published a short tutorial on audio measurements (http://lcweb2.loc.gov/master/mbrs/r...ektronix Cookbook of Standard Audio Tests.pdf) complete with audio geek from the era. There are others from various sources including National Semi focused on measuring switching amps. I would start with one of those and run through how to do a typical battery of tests then some specifics like optimizing bias vs. distortion, debugging clipping behavior or finding noise sources.
If someone local was interested I could provide my lab and help walk a neophyte through these tests. Would make a good Youtube sequence.
Strange. It works fine for me.
No hurry! The TEK-PDF ist great, thanks for the link! Looking forward to that PDF 🙂
One of the first things to build is the figure3 step attenuator.
I built a 12 switch precision version using 0.1% resistors.
Steps/switches for -20dB, -20dB, -10dB, -5dB, -2dB, -2dB, -1dB, -0.5dB, -0.2dB, -0.2dB, -0.1dB, -0.05dB.
This allows accurate attenuations from 0dB to -61dB in steps of 0.05dB
Pay attention to getting accuracy in the three highest attenuators -10dB & -20dB. It's no good if -20dB is actually -20.17dB
It should be assembled to -20dB +-0.02dB if you can manage that. That tolerance of +-0.02dB or better, applies to all steps.
Wiki is working now. May have been a timing issue.
.02 dB is a heroic accomplishment usually requiring .01% resistors and a perfect match on the input and output. Here is a detailed manual on a precision decade attenuator: http://www.ietlabs.com/pdf/Manuals/GR/1450 Decade Attenuators.pdf
Frex has a nice DIY variant here: AT2380v2 2Ch Digitally controlled Stepped Attenuator
Verifying .02 dB is also really challenging. It takes a pretty high grade voltmeter and very stable source. A Keithley 2015 is .05% midband but when you add all the uncertainties it will be higher than .02 dB. Even a Fluke 8506A is only .04%. The HP 3458 is good to .016% to 20 KHz for a price with sinewaves.
.02 dB is a heroic accomplishment usually requiring .01% resistors and a perfect match on the input and output. Here is a detailed manual on a precision decade attenuator: http://www.ietlabs.com/pdf/Manuals/GR/1450 Decade Attenuators.pdf
Frex has a nice DIY variant here: AT2380v2 2Ch Digitally controlled Stepped Attenuator
Verifying .02 dB is also really challenging. It takes a pretty high grade voltmeter and very stable source. A Keithley 2015 is .05% midband but when you add all the uncertainties it will be higher than .02 dB. Even a Fluke 8506A is only .04%. The HP 3458 is good to .016% to 20 KHz for a price with sinewaves.
Hi Demian,
You have such great resources! Thanks for the links.
I can second what you say about attenuator accuracy and resistors.
Hi Andrew,
It should be noted that even the selector switch contacts are a big issue at these levels. The other issue is resistor stability. You would need to keep them all at the same temperature on top of that. That's why most expensive meters have the divider resistors deposited on a ceramic substrate to both ensure temperature tracking and low capacitance.
The HP 3458A is "the meter" in the lab world, and there is a higher accuracy one available that uses a Fluke voltage reference. This meter is still the one to beat. It is often used as a voltage reference standard in most calibration labs.
I would love to have one, but I can't justify it for service use. It's ill suited to that job (the 34401A is about the best, or whatever Keysight has replaced it with) as a service meter. You would use it more for calibrating thermocouples and voltage references. Even the terminals (5 way binding posts) are more suited to lab use. You also would wait until the reading settles (all 8 1/2 digits), whereas the 34401A (6 1/2 digits) is normally used in the 5 1/2 digit mode for service bench work. The 34401A can be used for lab purposes at lower accuracy than the 3458A, and it has all the optimizations in menus for lab use. Many labs use these meters as they are an industry standard.
-Chris
You have such great resources! Thanks for the links.
I can second what you say about attenuator accuracy and resistors.
Hi Andrew,
It should be noted that even the selector switch contacts are a big issue at these levels. The other issue is resistor stability. You would need to keep them all at the same temperature on top of that. That's why most expensive meters have the divider resistors deposited on a ceramic substrate to both ensure temperature tracking and low capacitance.
The HP 3458A is "the meter" in the lab world, and there is a higher accuracy one available that uses a Fluke voltage reference. This meter is still the one to beat. It is often used as a voltage reference standard in most calibration labs.
I would love to have one, but I can't justify it for service use. It's ill suited to that job (the 34401A is about the best, or whatever Keysight has replaced it with) as a service meter. You would use it more for calibrating thermocouples and voltage references. Even the terminals (5 way binding posts) are more suited to lab use. You also would wait until the reading settles (all 8 1/2 digits), whereas the 34401A (6 1/2 digits) is normally used in the 5 1/2 digit mode for service bench work. The 34401A can be used for lab purposes at lower accuracy than the 3458A, and it has all the optimizations in menus for lab use. Many labs use these meters as they are an industry standard.
-Chris
that's why I pointed to the tolerance of the three highest attenuators. Once the attenuator is @ -5dB or less, 0.1% and a little hand matching gets below my 0.02dB requirement.
Yes, I agree, but using comparison techniques comes to aid us. Compare two -5dB to one -10dB. Compare two -10dB to one -20dB and the second -20dB and to each other. With care and small adjustements I satisfied myself that I got close at DC. I have much less confidence on tolerances at AC.
Last edited:
One place to check for measurement guidelines is NTi audio. They use their FX100 analyzer in these videos, but there is also a lot of general know how and explanations of background theory.
NTi Audio > Support > Webinars on Demand
Amplifier Test Sequence
Dealing with Power [W]
Basics of FFT Analysis
Advanced Topics of FFT Analysis
Also the specific measurement videos give some useful guidelines. As the measurement software of the FX100 analyzer is proprietary that part of course does not map directly, but many of the setting and problems would be similar to REW or ARTA or HPW etc
Videos for FX100 Audio Analyzer
Also Audio Precision has some good general Whitepapers along with more specific app notes of their instruments
https://www.ap.com/technical-library/?search_type=tech-library&tech_library_category[]=white-papers-books&load_page=1
NTi Audio > Support > Webinars on Demand
Amplifier Test Sequence
Dealing with Power [W]
Basics of FFT Analysis
Advanced Topics of FFT Analysis
Also the specific measurement videos give some useful guidelines. As the measurement software of the FX100 analyzer is proprietary that part of course does not map directly, but many of the setting and problems would be similar to REW or ARTA or HPW etc
Videos for FX100 Audio Analyzer
Also Audio Precision has some good general Whitepapers along with more specific app notes of their instruments
https://www.ap.com/technical-library/?search_type=tech-library&tech_library_category[]=white-papers-books&load_page=1
I posted the excel sheet years ago.
I'll post again shortly.
Remove the .txt from the file name to leave .xls
It was created in a very old version of MS excel. It does open and operate in the free spreadsheet programs, but the graphics sometimes get corrupted.
Use a two pole double throw switch to connect the T into circuit, or bypass it.
Last edited:
I have described the use of this switched attenuator before:
to manually measure the frequency response of an amplifier using a DMM that cannot measure properly at most of the frequency range that the amplifier is capable of. It also accurately measures the "Gain+attenuation" of the Power amplifier at any sensible frequency.
Start with a power amp that has a gain of +27.8dB with a filter that attenuates the input signal by -0.4dB. The overall gain+attenuation should be 27.4dB
Set the switched attenuator to -27.4dB.
apply a 1.5Vac signal to the attenuator input. Feed the attenuated output to the amplifier, remembering to attach the 600ohms dummy load to the attenuator output (use two 300r in series). Measure the power amplifier output.
It should read 1.500Vac on the DMM.
Since the frequency at the source and the at the output are the same, any frequency response error in the DMM measurement is very effectively cancelled and thus ignored.
Similarly since the input voltage from the source is the same as the output voltage any scaling error within the DMM is cancelled and can be ignored.
If you measure 1.502Vac at the input and measure 1.502Vac at the output, you can assume that within 1 Least Significant Bit/Digit (LSB or LSD) that the two voltages are the same, irrespective of DMM errors due to scaling and/ response.
If the voltage is not quite equal, switch in/out the low dB T attenuators till it does match.
Whatever setting is left on the attenuator gives the "gain+attenuation" of the amplifier.
This can be repeated at many frequencies. I have gone past 50kHz and got good results for the comparison measurements. Yes, the DMM reading has gone down because it cannot give an accurate absolute value for voltage (it's frequency response is terrible).
But you can COMPARE two equal voltages.
At the time I started doing this I only had a 3.5digit DMM (2000count) with a Vdc accuracy of 0.1%, but uncalibrated.
After I bought a bench top 4.75digit DMM (50000count) with a Vdc accuracy of 0.01% and measures rms for AC out to 5MHz, I rechecked my method by compared the 2000count results to the 50000count results and I was satisified that the comparison method is accurate if you take care to match frequency and voltage at the IN/OUT.
to manually measure the frequency response of an amplifier using a DMM that cannot measure properly at most of the frequency range that the amplifier is capable of. It also accurately measures the "Gain+attenuation" of the Power amplifier at any sensible frequency.
Start with a power amp that has a gain of +27.8dB with a filter that attenuates the input signal by -0.4dB. The overall gain+attenuation should be 27.4dB
Set the switched attenuator to -27.4dB.
apply a 1.5Vac signal to the attenuator input. Feed the attenuated output to the amplifier, remembering to attach the 600ohms dummy load to the attenuator output (use two 300r in series). Measure the power amplifier output.
It should read 1.500Vac on the DMM.
Since the frequency at the source and the at the output are the same, any frequency response error in the DMM measurement is very effectively cancelled and thus ignored.
Similarly since the input voltage from the source is the same as the output voltage any scaling error within the DMM is cancelled and can be ignored.
If you measure 1.502Vac at the input and measure 1.502Vac at the output, you can assume that within 1 Least Significant Bit/Digit (LSB or LSD) that the two voltages are the same, irrespective of DMM errors due to scaling and/ response.
If the voltage is not quite equal, switch in/out the low dB T attenuators till it does match.
Whatever setting is left on the attenuator gives the "gain+attenuation" of the amplifier.
This can be repeated at many frequencies. I have gone past 50kHz and got good results for the comparison measurements. Yes, the DMM reading has gone down because it cannot give an accurate absolute value for voltage (it's frequency response is terrible).
But you can COMPARE two equal voltages.
At the time I started doing this I only had a 3.5digit DMM (2000count) with a Vdc accuracy of 0.1%, but uncalibrated.
After I bought a bench top 4.75digit DMM (50000count) with a Vdc accuracy of 0.01% and measures rms for AC out to 5MHz, I rechecked my method by compared the 2000count results to the 50000count results and I was satisified that the comparison method is accurate if you take care to match frequency and voltage at the IN/OUT.
Last edited:
I use a similar procedure to compensate the autoranger for flat response. I routinely get 0.1dB out till 100kHz.
That doesn't mean that the absolute attenuation is that accurate, but the response is flat.
Jan
That doesn't mean that the absolute attenuation is that accurate, but the response is flat.
Jan
This DMM+attenuator method is nice and straightforward -- thanks for sharing this! But I don't understand how this relates to the RTX (or similar soundcard based setup). Can you please enlighten me?
Last edited:
This started with post1585 where 1audio posted a link to Tek testing procedures.
In there is an attenuator and I suggested that anyone/everyone could/should build an accurate stepped attenuator.
Last edited:
Yes, I understand this -- but how is the attenuator relevant for use with the RTX6001? I am just trying to understand if this is something that might be useful for the WIKI, and how. I am confused because the RTX already has built-in level switches both for the outputs and for the inputs. Are they not good enough?
There was a standard test set that incorporated an attenuator and a meter to streamline gain testing. The technique is independent of the source or meter flatness as long as the attenuator test set is accurate at the frequencies needed. The technique is still used for microwave calibration.
Within the audio band most premium ADC's and DAC's are very flat and the digital attenuation can be very precise. The limitation of the RTX is the 10 dB gain/atten. steps. Because the optimum distortion is found across a limited dynamic range finer analog gain/attenuation control would keep the distortion more optimized. It also considerably increases the complication of the instrument.
Absolute level and flatness for AC becomes really involved with thermal converters precision voltage references, etc. and is a low painstaking process. Fluke 540B Thermal Transfer Standard - Page 1 I have the necessary stuff but I rarely need it. A good oscillator with be really flat measured that way (.05 dB or better). And you can't get better than that without a physics lab and Josephson Junctions.
We have drifted way off base here. Need some real "how to" test normal audio equipment.
Within the audio band most premium ADC's and DAC's are very flat and the digital attenuation can be very precise. The limitation of the RTX is the 10 dB gain/atten. steps. Because the optimum distortion is found across a limited dynamic range finer analog gain/attenuation control would keep the distortion more optimized. It also considerably increases the complication of the instrument.
Absolute level and flatness for AC becomes really involved with thermal converters precision voltage references, etc. and is a low painstaking process. Fluke 540B Thermal Transfer Standard - Page 1 I have the necessary stuff but I rarely need it. A good oscillator with be really flat measured that way (.05 dB or better). And you can't get better than that without a physics lab and Josephson Junctions.
We have drifted way off base here. Need some real "how to" test normal audio equipment.
AndrewT, thanks!
It was OK the file, I have an old version of spreadsheet.
I must find time to look it, with attention.