I don’t expect that any opamp can be used as power supply. OPA551 is unity gain stable and trying the same with its direct ‘cousin’ OPA552, wouldn’t end with the same result.
I’ve noticed OPA551 exceptional stability while working on first prototype of super regulator with additional pass MOSFET, for another purpose.
I’ve noticed OPA551 exceptional stability while working on first prototype of super regulator with additional pass MOSFET, for another purpose.
Have you noticed Elvee , that again "Elvee's Law" was broken with the 1uF // with 330uF in tombo56's design ?
It is hard to stay in your 30 to 100 ratio. I've broke it too , with 0,1uF' x 3 and around 200 uF.
It is hard to stay in your 30 to 100 ratio. I've broke it too , with 0,1uF' x 3 and around 200 uF.
This “law” is very sensible if you want to avoid impedance peaks and it provides safety margin in design. However, sometimes you must break it and face the consequences. It is luck that there are none in this case.
Since the OPA551 is quite an expensive beast , I wonder how your design would compare to the LT30xx "super" LDO's .
They will have better PSRR at high frequencies, lower noise (not that it matters in this specific case). LT3045 is more expensive than OPA551. Maximum input voltage is to low for my case.
For distortion measurements I use basic level equipment: REW + Focusrite Solo gen. 3.
With Focusrite, there is not to wide window of output/input signal levels where its distortion is at minimum.
0,0003 % is that minimum.
With Focusrite, there is not to wide window of output/input signal levels where its distortion is at minimum.
0,0003 % is that minimum.
I wonder how valid those stability testing results of yours are. How fast is that Hantek scope and was the square wave signal applied directly to the input of OPA828 or to the RCA input with the following low pass filter? Why exactly 32 Ohm and such large capacitances for the testing?
Oscilloscope has 60 MHz bandwidth. All measurements were done without C5, so no input filtering. Amplifier was signal/load tested to above 10 MHz and output checked up to 60 MHz.
This is HP forum section and most headphones are 32 Ω, so why not to test with that impedance. Excessive capacitive load can reveal amplifier weakness or instability.
If you suspect any flaw in my measurements and methods then you should point to it. I wonder a lot of things as well. 😉
For headphone amplifiers I usually use 470pF, 1nF and 10nF in parallel with 50 Ohm, 100 Ohm and 300 Ohm. Makes 9 combinations in total. I also mostly use square waves or a square wave of lower amplitude on top of a higher amplitude sinewave. But with my Analog Discovery 2 and it's 30MHz-ish -3dB bandwith I was never really all that confident in my results. To gain more confidence I made sure that my next standalone scope had more BW and started testing more different combinations of load resistance and capactiance as well as different supply voltages. Still not 100% confident in the results, but better than before 😉
Well, I’m new in this forum section and can’t contribute much to the testing methodology of headphone amplifiers. I can follow established methods if there are any specific. Though, it shouldn’t be different form the preamplifiers or, partially, power amplifiers testing.
To be honest I wish there was something like an established method people would follow around here. In that regard you have already exceeded the pretty much non-existing standard since many people seem to rely on simulation or what some forum member or app note said should work. Which might be fine for the majority of cases, but a healthy portion of scepticism doesn't hurt.
How is that for dumb . Connected the HPamp , but the low voltage detector switched it off . Disconnected 1 channel and no problem , again with 2 , switched off . Replaced some caps that might have been too hot when soldering , still the same. then I remembered , I used a 2 step power up , and the second mosfet switch doesn't turn on if the voltage says below the detector's level . The first step has a 150 ohm limiter and because of the much higher current draw of the LME49600 ( compared to the test amp I used before) , I can'get the voltage up enough after the 8 sec startup grace period. 2 afternoons wasted ...
Well it works . The offset is 0,17 mV and a slightly disappointing -0,50 mV . Both are fairly stable with no signal , but with signal it varies a bit.
Typical offset of the OPA828 is 50 uV x 5 gain , should only give 0,25 mV offset. Still pretty good and below the 1 mV that I put as maximum.
The naked LME's get a bit too hot for my liking in 28 degr C room . The OPA828's are too small to feel , and too risky to touch with my tongue .
The trouble with Elvee's oscillation sniffer , is that it only detects above the schottky diode's V forward. Small oscillation only with oscilloscope.
Well it works . The offset is 0,17 mV and a slightly disappointing -0,50 mV . Both are fairly stable with no signal , but with signal it varies a bit.
Typical offset of the OPA828 is 50 uV x 5 gain , should only give 0,25 mV offset. Still pretty good and below the 1 mV that I put as maximum.
The naked LME's get a bit too hot for my liking in 28 degr C room . The OPA828's are too small to feel , and too risky to touch with my tongue .
The trouble with Elvee's oscillation sniffer , is that it only detects above the schottky diode's V forward. Small oscillation only with oscilloscope.
True, but uncontrolled oscillations tend to be largish, unless they are shunted by a low impedance.
Even then there is no actual hard threshold: the detection transfer is exponential, meaning even a few tens of mV pp will register in some way. the key is to make differential measurements: measure the voltage without power applied, then with the circuit powered.
If you see more than a few mv difference, you know you have a problem.
I have described an RF sniffer elsewhere, capable of detecting < 10mv pp using only regular 1n4148's.
If you are interested, I may dig it out, but basically, a diode detector based on point-contact Ge diodes should
be sufficient: they are low bandgap AND schottky, meaning they are very sensitive
The output value does not bear a reliable relation with the the actual RF voltage, but it doesn't really matter
Even then there is no actual hard threshold: the detection transfer is exponential, meaning even a few tens of mV pp will register in some way. the key is to make differential measurements: measure the voltage without power applied, then with the circuit powered.
If you see more than a few mv difference, you know you have a problem.
I have described an RF sniffer elsewhere, capable of detecting < 10mv pp using only regular 1n4148's.
If you are interested, I may dig it out, but basically, a diode detector based on point-contact Ge diodes should
be sufficient: they are low bandgap AND schottky, meaning they are very sensitive
The output value does not bear a reliable relation with the the actual RF voltage, but it doesn't really matter
Measuring a few mV after an oscillator sniffer can also be pickup by the DVM of noise , so I'm not too worried about it .
But making differential measurements is a good start . Thanks Elvee.
The HPamp draws 35 mA per channel idling , about what I had calculated , measured over all 4 , 10 ohms of the CRC .
But making differential measurements is a good start . Thanks Elvee.
The HPamp draws 35 mA per channel idling , about what I had calculated , measured over all 4 , 10 ohms of the CRC .
A solution looks best for me. I have used LME49600 just with other opamp. It’s just a very good circuit, nothing to wory about
Does anyone find that changing LME49600 bandwidth has an audible effect (assuming of course that stability is maintained for all test cases)?
I doubt that LME bandwidth change can affect sound in any way. Composite amplifier bandwidth is determined mostly by OPA828. LME, even in low bandwidth mode, has almost 3x the bandwidth and it’s slew rate is the same for both bandwidth modes. So, only resulting composite bandwidth remains as change. And there will be only insignificant change of maybe 100 kHz to 2.5 MHz – 3.5 MHz (gain dependent) of composite amplifier bandwidth.
BTW, based on latest buffer measurement driving line transformer in the amplifier, I’m confident that it delivers IRL, datasheet declared performance of 0.00003 %. Distortion present is almost completely Focusrite internal distortion.
BTW, based on latest buffer measurement driving line transformer in the amplifier, I’m confident that it delivers IRL, datasheet declared performance of 0.00003 %. Distortion present is almost completely Focusrite internal distortion.