Electronics > Projects, Designs, and Technical Stuff
Audio Measurement Pre-Amplifier
toli:
udok, lets not forget what this pre-amplifier is meant to do. It wasn't designed to operate with a 20dBu signal right at the input of the op-amp, its only meant to measure the distortion of that signal. Using a voltage divider at the input to bring it down to a voltage level that is more comfortable for the opamp is completely ok from my point of view. Especially when you consider the fact that the output signal must be 2Vrms or lower to be within the linear range of my sound-card. BTW, the 12V supply isn't limiting this here, the input protection circuit will kick in before this happens, as it wasn't designed to have more than a 2Vrms signal at the opamps input. 20dBu out of range? No problem, switch to a higher range, all the way up to 200Vrms ;)
Absolute reading accuracy is also not too critical here, I'm totally ok with 1-2% measurement error as the readout is meant mostly for measuring output power of amplifiers.
This was meant to be another tool for my work bench. It will be used to measure output of low impedance speaker amplifiers, as well as internal nodes measurements for both transistor and valve amplifier (where high impedance and DC voltage could be present) and more. Therefore, it is AC coupled at the input, and has a 100Kohm input impedance at each input.
The LM4562 you mention is one of my preferred opamps, when it meets the requirements. In fact, if you'll have a look at my blog, you'll see that I've modified the EMU 0404 USB and used LM4562's quite extensively there. This is a low cost, low noise, low distortion opamp, which I use very often. However, for this application where the input is AC coupled and has a 100Kohm resistance at the input, the 10nA (72nA max) input bias current is simply too high for the higher gain settings.
There are obviously other ways to implement every circuit, each with its own pros and cons. I've actually had a few different circuit drawings until I've finally decided that this one seemed like a good compromise between my different requirements. This pre-amplifier was designed for my own use to be able to measure distortion and noise of my other projects. It achieves the target spec I was aiming for, so I'm happy with it :D
With that being said, I do take your (and everyone else's who have replied here) comments into account, and if I ever modify the circuit in the future I will revisit these design decisions once more.
SiliconWizard:
--- Quote from: magic on August 17, 2019, 01:41:53 pm ---Now, check out THD specs of that in-amp :)
--- End quote ---
I know this part pretty well, have used it in several designs actually. It's been a popular solution for XLR input stages in audio gear.
A 0.002% THD+N figure @1kHz (and 1nV/rtHz) is not too shabby really.
I understand the OP was after something much better than 0.002% THD+N. He only talks about THD though, not THD+N. The difference could surprise him, but maybe he just said THD for THD+N.
I was wondering about the overall performance of the circuit.
Not completely convinced about CMRR not being an issue in the OP's design (as he later admitted), but not as big a deal as a SE output as he stated.
Although I completely understand the OP's approach, and the theoretical benefits, I'd still be curious to compare it with a much simpler one with something like a INA163 (with the lowest gain you can get by with) + DRV135 for the output stage.
I admit it would be more of a design for a pure audio device rather than for a measurement equipment, still I'd be curious to see what you can get out of that (especially in terms of noise).
magic:
--- Quote from: udok on August 17, 2019, 05:29:33 pm ---Better use a LM4562 or AD795. They have lower noise
--- End quote ---
No, they are horrible. Their current noise of a few pA/rtHz will annihilate whatever voltage noise advantage they have, in this circuit. There are better bipolar opamps for 1.5k source impedance.
--- Quote from: udok on August 17, 2019, 05:29:33 pm ---no nonlinear input capacitance
--- End quote ---
Samuel Groner's paper shows plenty of problems in bipolar opamps trying to pass 20Vpp from 100k source impedance. Often worse than the old DIFET OPA627. I'm not sure how it translates to this circuit, though.
Their inputs actually still have variable capacitance, between base and collector. Additionally, Early effect modulates their beta, producing nonlinear bias current which impresses nonlinear voltage drop on source resistance. Both of these effects can be mostly eliminated by dynamic cascoding, do you know if your favorite opamp does it? ;)
Less obviously, Early effect in the tail source modulates input stage standing current, which again modulates bias current.
Finally, distorted current output from the input stage is required to compensate for second/third stage nonlinearity. Again, divided by beta, this nonlinearity leaks to the outside world.
--- Quote from: udok on August 17, 2019, 05:29:33 pm ---Maybe it would be better to use an Audio transformer for the balanced to single ended conversion?
--- End quote ---
Haven't those been ditched because of distortion and frequency response inaccuracy?
--- Quote from: toli on August 17, 2019, 07:02:35 pm ---if I ever modify the circuit in the future
--- End quote ---
Consider switching attenuation on the input side, so that impedance seen by the opamp stage is always 1k. Only the DUT will see less load impedance at -20dB than at -40dB.
--- Quote from: SiliconWizard on August 17, 2019, 07:32:59 pm ---I understand the OP was after something better than 0.002% THD+N. He only talks about THD though, not THD+N. The difference could surprise him, but maybe he just said THD for THD+N.
--- End quote ---
Okay, at last some improvement over the parts I posted :-+
But frankly, noise isn't the primary concern in THD measurements, just use longer FFTs, up to the limits of lifetime of the Universe :D
edit
Though you are right, it will interfere with noise floor measurements. Frankly, I would separate the two and use different preamps for each. Amplifying noise is no rocket science, at least at audio frequencies.
SiliconWizard:
--- Quote from: magic on August 17, 2019, 07:49:57 pm ---But frankly, noise isn't the primary concern in THD measurements, just use longer FFTs, up to the limits of lifetime of the Universe :D
--- End quote ---
True, if you're talking about gaussian noise. So if we're talking about the nV/rtHz figures, you're mostly right.
As for THD vs. THD+N, this is yet another story. If using the definitions correctly, THD is only pure harmonic distortion. THD+N includes any form of distortion. Some forms of distortion that you can typically find in audio circuits are non-harmonic. Thus a low, pure THD figure may look misleadingly low.
--- Quote from: magic on August 17, 2019, 07:49:57 pm ---Though you are right, it will interfere with noise floor measurements. Frankly, I would separate the two and use different preamps for each. Amplifying noise is no rocket science, at least at audio frequencies.
--- End quote ---
Yep, and why not.
toli:
Regarding the N of the THD+N, it isn't neglected, its simply low enough to be sufficiently below that of the EMU at the 2V range. For the higher gain ranges the input ref'd noise improves as expected and the added gain brings the external amplified noise above the noise floor of the EMU. This shows things that are otherwise too small in amplitude for the EMU to display, but unfortunately this means that the maximum input signal amplitude must drop accordingly with the range or else it would clip the input of the soundcard.
The EMU 0404 USB is the most limiting part of the setup at the moment, despite the fact it has improved with the opamp replacement to LM4562. However, if I'm being honest, the measurement capabilities I have with this setup are sufficient for my current needs. Extending this further will probably cost significantly more than the 40$ I payed for the EMU 0404 USB on eBay :)
If anyone is familiar with another USB sound interface (must be external since I'm using a laptop) that performs better than this EMU for a reasonable price, do share.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version