Wow, that was even easier to fix than it was to miss, thank you! I assumed L and R channels were asymmetric by design (one for measurement and one for self-calibration).
Now I'm getting THD numbers close to what RMAA shows (still worse, but maybe not because of REW itself). I wonder, are there different standards for calculating THD that might explain different software showing different % figures when measuring the same thing?
Last edited:
What an amazing thread this is. I have played around myself with recycled parts using my wife's old MacBook (still has the CD-ROM!), an old Echo Audiofire 4 (firewire sound card) that has collected dust for 10 years or so and a case I had around where I placed four 8 Ohm / 100 W resistors (unknown chinese stuff), few switches and the Aktika-style attenuator. BNC goes to the oscilloscope, TRS to the soundcard.
I know that the resistors are not ideal and will be replaced at some point. However, I have been able to chase and learn more about hum and buzz for instance. Quite revealing has been the measurements of the Aleph P1.7 where the BAL connection has almost eliminated the 50Hz peak.
Nice work! That is a very useful rig. You will find yourself using it a lot I think. I agree that it is very powerful for seeking to make an amp truly quiet. That’s an impressively clean noise floor for a Class A amp you have there. Congrats!
Check your interface settings. If you use basic Java vs ASIO interface that can affect the noise and other measurement data. Also make sure you have similar sample length, %averaging, number of averaging, etc in the “gear” icon for settings.
Does anyone know historically how the process for taking THD measurements for a product specification goes? And before you answer, what I mean is that when taking my own measurements, I noticed I was somewhat able to "coax" out the best THD+N numbers through adjusting the various gain stages to get the best dynamic range. Then I cherry picked those numbers. Do manufacturers do this too?
Obviously when you're listening to an amp, at least in a casual context, you're not going to be fiddling with the "optimal gain settings" that you might have previously figured out, but it was a good learning exercise for me on the topic of dynamic range.
Obviously when you're listening to an amp, at least in a casual context, you're not going to be fiddling with the "optimal gain settings" that you might have previously figured out, but it was a good learning exercise for me on the topic of dynamic range.
In the post FTC world, you want to publish measurement values that you are certain will be matched or exceeded by skeptical reviewers and even more skeptical regulators. Otherwise you will be branded a liar and a cheat. So don't do anything that an overworked technician at a government lab wouldn't do, and also measure several different amplifier units and publish the worst measured number for each of the specifications.
I’m using REW and Focusrite Scarlett 2i2 to measure frequency response and distortion of my audio gear. A slight inconvience with Scarlett 2i2 is that it’s maximum unbalanced output level is only 2.34V RMS.
Can anybody advise how to get full balanced output level (4.67V RMS) into unbalanced amplifier input? Any high quality and cost effective solution available?
I’d need the higher level to drive my F6 amp into full power. F6 gain is only 14dB (5x) so I need at least 3.0V RMS. Now I need to use an external sine wave generator and it makes THD vs. power measurement inconvenient.
Can anybody advise how to get full balanced output level (4.67V RMS) into unbalanced amplifier input? Any high quality and cost effective solution available?
I’d need the higher level to drive my F6 amp into full power. F6 gain is only 14dB (5x) so I need at least 3.0V RMS. Now I need to use an external sine wave generator and it makes THD vs. power measurement inconvenient.
A good preamp with +/-15v rails will do what you need. You can look for commercial preamps or buy DIY kits. If you get one, you might as well figure proof it and make sure it has balanced capability as well.
I generally use the one designed by Jhofland called the Best Thing since Sliced Bread (BTSB) buffer. It has balanced/SE in and Bal and SE out and DIP switches to select gain of 0/6/14/20dB and has built in +/-15v PSU and only needs 12v wall wart. There are through hole and SMT version. Noise floor is exceedingly low around -130dB and distortion is also low (based on state of the art OPA1656 and LME49724 for v1.3 and OPA1656 and OPA1637 for v1.2p).
OTOH, you could round yourself up a clean +/-15v supply and use this $10 board as a preamp. You will need to figure out the resistors for setting the gain though:
GHXAMP NE5532 Balanced XLR to Single ended RCA Output Dual op amp circuit board small size low distortion low noise|Amplifier| - AliExpress
I generally use the one designed by Jhofland called the Best Thing since Sliced Bread (BTSB) buffer. It has balanced/SE in and Bal and SE out and DIP switches to select gain of 0/6/14/20dB and has built in +/-15v PSU and only needs 12v wall wart. There are through hole and SMT version. Noise floor is exceedingly low around -130dB and distortion is also low (based on state of the art OPA1656 and LME49724 for v1.3 and OPA1656 and OPA1637 for v1.2p).
OTOH, you could round yourself up a clean +/-15v supply and use this $10 board as a preamp. You will need to figure out the resistors for setting the gain though:
GHXAMP NE5532 Balanced XLR to Single ended RCA Output Dual op amp circuit board small size low distortion low noise|Amplifier| - AliExpress
I generally use the one designed by Jhofland called the Best Thing since Sliced Bread (BTSB) buffer. It has balanced/SE in and Bal and SE out and DIP switches to select gain of 0/6/14/20dB and has built in +/-15v PSU and only needs 12v wall wart. There are through hole and SMT version. Noise floor is exceedingly low around -130dB and distortion is also low (based on state of the art OPA1656 and LME49724 for v1.3 and OPA1656 and OPA1637 for v1.2p).
Thanks XRK! This is exactly what I need. I see that this is a good tool for general T&M needs but also good for e.g. to be built in the F6 as an preamp with settable gain to bring F6 input sensitivity to "standard" level.
Thanks XRK! This is exactly what I need. I see that this is a good tool for general T&M needs but also good for e.g. to be built in the F6 as an preamp with settable gain to bring F6 input sensitivity to "standard" level.
The java driver in REW is worthless for my soundcard, it shows huge peaks of the harmonics that don't exist in the device on test but are added in the java driver.
The best way to get it working in REW was to use the ASIO driver.
But with the ASIO driver you can't select a different device for input and output.
A workaround is to use another generator/player as output for the test tone, for example foobar with a 192khz 24bit test tone.
The best way to get it working in REW was to use the ASIO driver.
But with the ASIO driver you can't select a different device for input and output.
A workaround is to use another generator/player as output for the test tone, for example foobar with a 192khz 24bit test tone.
That's a good tip, thank you. Maybe this way I'll get ASIO working with EMU 0404.
Definitely use a 24b/192khz 1khz testtone, that puts the resolution at 144db,
16b/44.1khz is only 96db
16b/44.1khz is only 96db
The java audio driver itself does not add any harmonics. It's only 16bit though and outputs to windows audio mixer which most likely resamples in your case.
Another option is booting to linux where the java driver outputs bitperfectly up to 32bits to hardware devices.
There are many ways to generate any tone at any frequency, samplerate, bit width, channel count, sox being one of the most flexible ones. For USB I would try to avoid the exactly 1kHz as many USB devices experience 1kHz artefacts due to the USB transport.
There is something else wrong with the java driver,
the java driver showed the harmonics on -10db,
with ASIO it was -90db like it should be.
Hi Neal,
I too am working on learning how REW works with a 2i2 and Akitika.
One thing to check is the signal received by REW - I found that the choice of input termination for the 2i2 determines the preamp gain (page 15 manual, verified by trying the two input types). With an XLR input, the software assumes it is a microphone input; with TRS, line level. You can see the signal levels by selecting "Signals" in REW. I have found it useful to verify the Windows Sound settings and check the signal levels before assuming RTA is providing reasonable results. Also, I found REW version 5.20.4 easier to use in terms of getting reasonable results.
I will post some results separately.
Regards,
Mike