Electronics > Projects, Designs, and Technical Stuff
Audio Measurement Pre-Amplifier
toli:
Hi All,
I wanted to share a project I have been working on for a few months in my (limited) spare time. To allow me to make easier and better measurements of audio equipment (and a few other things as I'll mention later) I wanted to design a measurement pre-amplifier for audio band frequency. I have initially considered building something that is "readily available" like the soundcard interface project from pmillett. However, I wanted something that is very versatile that will have a high maximum voltage range, low noise for low signal measurement, SE and BAL support for both input and output signals even for higher voltage swings at the input. Therefore I've decided to design one that will meet my specific needs.
I plan on using it for both testing audio gear for distortion and other possible issues (noise for instance), as well as being able to measure low noise densities for things like voltage regulators and a few other low noise projects I have in mind where this will come in handy.
To support all of this I went for a balanced input structure, and 6 different gain range (selected by relays) from -40dB to +60dB in 20dB steps. For the most sensitive ranges (+40dB and +60dB gain) the input referred noise is 7nV/rtHz (assuming it is driven from a low impedance source obviously). The input resistance is 100Kohm at each terminal, so 200Kohm differential.
The schematic for the main part of the circuit is shown below:
The schematic is actually fairly straight forward with a resistive voltage divider with 3 different outputs for 0dB/-20dB/-40dB settings, then an input protection circuit to keep the first stage opamp's safe. This is followed by 2 more stages with gain (0dB/+20dB in the first stage, -6dB/+14dB/+34dB in the second stage), and finally enters the output line driver which adds its own +6dB of gain.
There is also a T-RMS -> DC converter chip there to be able to measure the input signal amplitude (this isn't shown in this snip of the schematic).
The PS section shown below:
It starts with some input protection, into an isolated DC-DC module, followed by low-noise linear regulators and some filters. This is done to keep the supply noise+ripple low enough for the higher gain settings where noise coupling could be an issue.
I've used an additional smaller board to mount some of the controls to be able to place everything into a moderate size case when I'm finished.
The complete schematic can be downloaded here:
http://tolisdiy.com/wp-content/uploads/2019/08/MeasurementPreAmp_Schematic_TolisDIY-1.pdf
The assembled pre-amplifier can be seen below:
It is comprised of the main board that does all of the heavy lifting, a smaller control board for range selection and mounting of the range select LED's, and a panel mount DC voltmeter that I got from AliExpress obviously :)
I have done some measurements for THD, noise, and a few other parameters. I am measuring i t with a modded EMU 0404 USB soundcard. To extent linearity of the measurement setup I use a low distortion oscillator, and a notch filter for 1KHz measurement. I am actually very happy with the performance, they match my expectations and calculated values quite well. I think I could potentially get more out of with a higher end soundcard or interface, but if I'm being honest, the current level of distortion is just excellent for my hobby use.
THD vs. frequency, unfortunately this is limited by the EMU 0404 USB:
Spectrum with external oscillator at 1KHz 2Vrms SE signal, through the pre-amplifier and into the soundcard, again, THD is limited by the EMU 0404 USB:
With a notch filter added before the EMU:
The notch filter transfer function can be seen here:
http://tolisdiy.com/wp-content/uploads/2019/08/fr_measure.png
Overall it seems that THD on the order of 0.0001% can be measured even with a full-scale signal. I'm saying "on the order of" and not giving an exact number because even with the notch filter, it seems like the residual distortion of the EMU is preventing me from getting an accurate reading of the distortion of the pre-amplifier.
Finally, I've measured the input noise density on the higher gain settings (terminated with 50ohm to reduce coupling and measure the noise assuming a low impedance source is connected), with 0dBFS being 4mVrms:
This translates to ~8nV/rtHz, which is very close to the calculated value of 7nV/rtHz that I've expected based on the datasheed of the input amplifier (3.3nV/rtHz typical value) and the value of resistors in the input protection and the feedback loop of the first stage.
This project isn't done yet, as I still need to design the front and rear panels for the aluminium case, to make it look nicer, easier to use, and reduce external coupling especially on the more sensitive ranges.
However, this project finally reached a point where I've been able to verify its performance meet the target I have set when I've started it, so I figured I should share it on the forum :blah:
I have tried to keep this post shorter and to the point, which is why I have omitted many measurements results, and comments about the circuit design. However, if you'd like to read in more detail, I have posted all of it with much more detail on my blog.
magic:
1ppm ain't bad :)
I think this is territory where a low-noise bipolar like 5534 or OPA2210 could give those JFET opamps a good run for their money. But the -20dB attenuator would cause problems because of its 9k output impedance.
Speaking of which, what's THD at 10kHz on -20dB range? The datasheet of ADA4625 is silent on certain important issue.
toli:
You are right, the need for a low current at the input of the op-amp is why I had to opt for something with a FET input. Otherwise going for a BJT input device would have been better and noise could have also been even better on the 2mV/20mV ranges.
Measuring at 10KHz is something I can do only with the EMU 0404 as my external oscillator and notch are tuned for 1KHz only. Therefore at 10KHz I'm limited to ~2Vrms and a measurement limit of ~0.001% THD which is the best the EMU can do at this frequency. This means that on the 20V (-20dB) range the input amplitude will be much smaller than full-scale simply because I don't have anything on the bench that can generate a low distortion 20V 10KHz signal. Given these limitations, I've measured this, and the measurement is still limited by the EMU 0404 performance at ~0.001% at 10KHz.
magic:
Okay, that's not terrible. The problem is of course capacitance of the opamp's input JFETs (and, frankly, your protection diodes too but it's smaller in comparison) which varies with signal level. Its impedance forms a divider with the attenuator resistors and the division ratio varies with input voltage.
It may still become somewhat worse when the full 20Vrms input range is utilized.
The protection diodes could be bootstrapped if it ever becomes a problem. As for opamps, they exhibit this problems in various amounts and it's rarely documented. Maybe the ADA4625 isn't too bad. At least its inputs seem to be cascoded.
http://www.ti.com/lit/an/slyt595/slyt595.pdf
toli:
You are right on point there magic :D
This is the exactly why I went for discrete small signal diodes that have very low capacitance in the input protection circuit, and I've biased them with 5V of reverse voltage so that their contribution is minimal to the input capacitance of the opamps even with maximal expected signal swing.
The results of 0.001% are limited by the EMU as I've mentioned, so I can't really say what the exact limit under these conditions is.
What is possible to do relatively easily is to add a series resistor of 10K at the input and measure this on the 2V range. This will have little attenuation, so it will allow measurement with maximal swing at the input of the opamps. While this will still be limited to 0.001% by the EMU limits, this will be done with a full swing input at the input of the opamps which will stress them significantly more. I actually think that's valuable information as it will give me some indication of the limits in a range that I can't verify otherwise. So this discussion is already bearing fruits :-+
In fact, its actually so easy to do, that I went ahead and added 10K in series with each of the inputs in a very improvised fashion and measured it now (see attached image for the improvised connection).
For the full swing of 2V at 10KHz (1.97Vrms to be exact), the THD is now about 0.0012% (vs 0.001% in bypass mode). So ~0.001% is about the best THD that could be measured with a full-swing 20Vrms signal on the 20V range at 10KHz. This will obviously be better in other ranges, so if THD is all I would like to measure, I can always switch to the 200V range even for a 20V signal which will both reduce the output impedance of the divider and reduce swing at the opamp input ;)
Navigation
[0] Message Index
[#] Next page
Go to full version