In another thread, regarding amplifier distortion measurements (started by xrk971 some years ago), I recently posted some questions about distortion measures of my 2A3 SET amplifier. I really appreciate all the answers I got in that thread. But for not polluting that excellent thread with my specific questions, I´m now starting this thread.
Initially my question was however my measures was reliable, but rapidly the discussion were focused on how to improve the measured amplifier and to remove the deviations shown in the FFT-plot.
Now, I would like to rewind the discussion to were I started it, by -step by step, show how i proceed with the measurements and If I did something wrong. By that, conclude if the plots I get is reliable. Later i´m open for a discussions on how to progress, and how the particular DUT (amplifier) can be improved.
Hardware:
I´m using an old HP laptop (with Windows 8.1) run on batteries, put it in flight-mode, not having the risk of interruption from my internet connection. No virus- or other software is installed on this computer. I know that he measuring harnesses aught to be shielded, But I din´t have access to such cable, so I used twisted pairs, tried to keep them as short as possible.
Soundcard is Focusrite Scarlet Solo 3´rd gen which has the following interfaces:
Environment:
All measures are done in my basement on the Swedish countryside, where I have some current consuming gears (electrical heater and led-lights) in my surrounding. Although I have not seen any deviations in my measurements that I could derive to radiated EMI or mains power spikes, so I think the environment is okay for these measurements.
Software:
I have downloaded the latest appropriate REW software.
Focusrite Solo Calibration/loopback:
When soundcard calibration was made, I looped Output 1 (TRS connector using TR tabs) to Input 1 (XLR connector pin 2 & 3). I also connected Output 1 (TRS connector using TR tabs) to Input 2 (TRS connector using TR tabs).
Note: Input 1 (XLR) is a mic input at Solo, and not a line input as in some other sound-cards offer. Hence I use a 10:1 attenuation (voltage divider) when measure speaker output signals (see below for detail). Input 2 is a line input.
In REW Soundcard Preferences, the setup were done as below. Levels were adjusted according to the bars and a calibration file was generated. With the calibration file used the frequency response is ruler flat from some Hz to +30kHz.
The same loop back setup were used for testing equipment noise using REW internal generator and FFT-analysator:
Note that Y-axle is in dBc (level related to the carrier (1KHz) set to 0dBc. Distorsion is low and noise is in the -115dBc region.
Measuring DUT:
When measuring the device I removed the loop back harness, and instead used this one.
The power resistors are two resistors of total 8 Ohm. Banana plugs are connected to speaker outlet respective to a Fluke true rms ac volt meter.
Parallel to the power resistors there is attenuation network of 100K resistor in series, and between them two 10K resistors, making a virtual center tap and a 10:1 attenuation for each leg.
This "pseudo balanced" signal (two hot and one virtual ground) was connected to the XLR at Soundcard Input 1
Soundcard Output 1 (TRS connector using TR tabs) connected to av RCA male attached to the DUT input. and also to Input 2 (TRS connector using TR tabs)
Turning the DUT on with no stimuli added. this is what the FFT-plot shows. As seen there is an annoying peak at 50Hz and it´s harmonics. Note that there are also peaks in the 8kHz - 13kHz region. I have no idea what causes that?
Measurement 1W output (2.82Vrms /8 Ohm)
With 1kHz stimuli added the plot shows as below. 50Hz peak increased to -60dBc and one can see that the 1kHz stimuli (and its harmonics) are modulated.
Anyone seeing anything odd or faulty done in this so far?
Initially my question was however my measures was reliable, but rapidly the discussion were focused on how to improve the measured amplifier and to remove the deviations shown in the FFT-plot.
Now, I would like to rewind the discussion to were I started it, by -step by step, show how i proceed with the measurements and If I did something wrong. By that, conclude if the plots I get is reliable. Later i´m open for a discussions on how to progress, and how the particular DUT (amplifier) can be improved.
Hardware:
I´m using an old HP laptop (with Windows 8.1) run on batteries, put it in flight-mode, not having the risk of interruption from my internet connection. No virus- or other software is installed on this computer. I know that he measuring harnesses aught to be shielded, But I din´t have access to such cable, so I used twisted pairs, tried to keep them as short as possible.
Soundcard is Focusrite Scarlet Solo 3´rd gen which has the following interfaces:
- Input 1 (XLR). Ballanced Mic input with an internal pre amplifier.
- Input 2 (TRS). Ballanced Line level input.
- Output 1 (TRS). Ballanced Line level output
Environment:
All measures are done in my basement on the Swedish countryside, where I have some current consuming gears (electrical heater and led-lights) in my surrounding. Although I have not seen any deviations in my measurements that I could derive to radiated EMI or mains power spikes, so I think the environment is okay for these measurements.
Software:
I have downloaded the latest appropriate REW software.
Focusrite Solo Calibration/loopback:
When soundcard calibration was made, I looped Output 1 (TRS connector using TR tabs) to Input 1 (XLR connector pin 2 & 3). I also connected Output 1 (TRS connector using TR tabs) to Input 2 (TRS connector using TR tabs).
Note: Input 1 (XLR) is a mic input at Solo, and not a line input as in some other sound-cards offer. Hence I use a 10:1 attenuation (voltage divider) when measure speaker output signals (see below for detail). Input 2 is a line input.
In REW Soundcard Preferences, the setup were done as below. Levels were adjusted according to the bars and a calibration file was generated. With the calibration file used the frequency response is ruler flat from some Hz to +30kHz.
The same loop back setup were used for testing equipment noise using REW internal generator and FFT-analysator:
Note that Y-axle is in dBc (level related to the carrier (1KHz) set to 0dBc. Distorsion is low and noise is in the -115dBc region.
Measuring DUT:
When measuring the device I removed the loop back harness, and instead used this one.
The power resistors are two resistors of total 8 Ohm. Banana plugs are connected to speaker outlet respective to a Fluke true rms ac volt meter.
Parallel to the power resistors there is attenuation network of 100K resistor in series, and between them two 10K resistors, making a virtual center tap and a 10:1 attenuation for each leg.
This "pseudo balanced" signal (two hot and one virtual ground) was connected to the XLR at Soundcard Input 1
Soundcard Output 1 (TRS connector using TR tabs) connected to av RCA male attached to the DUT input. and also to Input 2 (TRS connector using TR tabs)
Turning the DUT on with no stimuli added. this is what the FFT-plot shows. As seen there is an annoying peak at 50Hz and it´s harmonics. Note that there are also peaks in the 8kHz - 13kHz region. I have no idea what causes that?
Measurement 1W output (2.82Vrms /8 Ohm)
With 1kHz stimuli added the plot shows as below. 50Hz peak increased to -60dBc and one can see that the 1kHz stimuli (and its harmonics) are modulated.
Anyone seeing anything odd or faulty done in this so far?
Attachments
The cables are not shielded.
The shielded and good interconnection cable should be used.
Also grounding is important (avoiding grounding loop between the Solo and the DUT)
The shielded and good interconnection cable should be used.
Also grounding is important (avoiding grounding loop between the Solo and the DUT)
Try measuring a different amplifier and compare the results. That will give you an indication whether your test procedure or your amplifier or maybe both are causing the 50 Hz related noise.
Are you using the balanced input? It looks like you need a ground connection on that input. You are picking up 50Hz noise from power line.
Your pseudo symm divider will work with symm full bridge outputs (TPA3255 etc) exclusively
A single ended amp output must be connected strictly unbalanced to the soundcard input.
A single ended amp output must be connected strictly unbalanced to the soundcard input.
Last edited:
One of the (few) cons of Diyaudio is that it´s hard to summarize the long threads, reading afterwards. It´s very easy to misunderstand if discussions is´t followed then they are present. Obviously I have misunderstand some of the older posts regarding how the measuring harnesses shall be connected.
Just to make one thing clear. I only measure single end amplifiers and if not exclusively, mostly low power tube amplifiers.
Now, this is how my new harness is connected.
And this is the plot when amplifier is turned on, with no stimuli. Note the 50Hz peak are still there at -80dBu. But H2 is -54dBu.
When 1kHz stimuli and 1W output, the 50Hz peak is -57dBc. Pls. note that the previous plot was measured in dBu, and is now changed to dBc (decibel to Carrier).
I also made some frequency response measurements. Besides some small ripples, I see no abnormalities here.
Distortion in frequency domain also look okay to me.
Just to make one thing clear. I only measure single end amplifiers and if not exclusively, mostly low power tube amplifiers.
Now, this is how my new harness is connected.
And this is the plot when amplifier is turned on, with no stimuli. Note the 50Hz peak are still there at -80dBu. But H2 is -54dBu.
When 1kHz stimuli and 1W output, the 50Hz peak is -57dBc. Pls. note that the previous plot was measured in dBu, and is now changed to dBc (decibel to Carrier).
I also made some frequency response measurements. Besides some small ripples, I see no abnormalities here.
Distortion in frequency domain also look okay to me.
Attachments
There is a great article in this issue that will help you.
https://audioxpress.com/news/Design...-t-Miss-audioXpress-April-2016-Now-Available-
https://audioxpress.com/news/Design...-t-Miss-audioXpress-April-2016-Now-Available-
Provided the cold end of your unbalanced speaker cold end (gnd) is the upper wire -
I would connect pin3 with pin1 of balanced input.
I would connect pin3 with pin1 of balanced input.
I previously made that virtual ground network between amp output and Solo input. But I was told that this was only suitable for H-bridged amplifiers and not for single end.Look at this circuit: measurement setup
Now I don’t know whom to listen to?
Do you mean my circuit or the one linked to in post #10 ?This circuit shows a balanced amp output - pointing into the wrong direction
If mine. I don’t see what’s wrong. XLR pin 1 is gnd. Hence pin 2 sees 1:10 of Voltage feed by the amplifier over the 8 Ohm load.
No. This setup shows balanced probing. Like two-channel differential 'oscope probing. No matter what DUT output is. In addition this setup break GND loop.This circuit shows a balanced amp output - pointing into the wrong direction
And for normal FR/FFT measurement reference channel is not neccesary. Simple Soundcard Out-> DUT In and DUT Out->Soundcard In. You have one TSR line-level in, it might be better than mic XLR.
This is a misconception.No. This setup shows balanced probing. Like two-channel differential 'oscope probing. No matter what DUT output is. In addition this setup break GND loop.
Probing an unbalanced output with a balanced attenuator is calling for trouble.
I tumbled into this trap too some time ago.
I would appreciate if anyone could share a fft-plot of a SET amplifier with AC heater for me to compare with.
I have a question. Could one measure amplifier noise floor with this setup (noise amplitude spectral density)?
Last edited:
- Home
- Design & Build
- Software Tools
- Amplifier distortion measurements using REW and Focusrite Solo