Well after tearing the soldered board apart, I just thought I’d breadboard @Nelson Pass idea on the 1k input resistor and @richb bigger caps. Unfortunately neither of those changed performance significantly. I am using opa1612 (bipolar) and opa1656(fet) opamps. Bith are unity gain stable. Much better performance than an lm4652. Also tried opa1622 which was not as good as the opa1612 or opa1656.
At this point the temp gain over ambient with no signal is about 27 deg c. Adding a signal @2vrms brings it up another 5 deg c.
When watching dc voltage change across the 10k resistor from in+ to gnd, it varies only a few hundredths of a mv. I coukdn’t see anything odd in the fft on the scope or the qa402.
The temp doesn’t vary. Is there really any oscillation going on?
@EmuMannen hearing the difference doesn’t apply here. This is for testing ultra low noise circuits with thd+n near -120db so the noise from every resistor or higher values of resistors really matters which is why I’m agonizing way too much over this.
At these temperatures I’m less than a db off what amir at Audioscience forum measured on the d10s dac. The problem is not the harmonics but a db or a little more in noise. Although I need to look at the smsl su-9 which has much better performance than the d10s and see how things look.
At this point the temp gain over ambient with no signal is about 27 deg c. Adding a signal @2vrms brings it up another 5 deg c.
When watching dc voltage change across the 10k resistor from in+ to gnd, it varies only a few hundredths of a mv. I coukdn’t see anything odd in the fft on the scope or the qa402.
The temp doesn’t vary. Is there really any oscillation going on?
@EmuMannen hearing the difference doesn’t apply here. This is for testing ultra low noise circuits with thd+n near -120db so the noise from every resistor or higher values of resistors really matters which is why I’m agonizing way too much over this.
At these temperatures I’m less than a db off what amir at Audioscience forum measured on the d10s dac. The problem is not the harmonics but a db or a little more in noise. Although I need to look at the smsl su-9 which has much better performance than the d10s and see how things look.
A good scope (at least one of around 30Mhz bandwidth or more) will show oscillation in seconds.
https://www.diyaudio.com/community/...-checked-to-see-its-stable-havent-you.191389/
This thread started with -14.something V offsets and overheating op-amps, lately the DC bias points are reasonable and the self heating is not excessive either; although I don't understand post #55. It could very well be that the problem has been solved, but it won't hurt to check with an oscilloscope.
So I used a different dip8 adapter for the opa1612 and opa1656 and made sure I placed and soldered them as perfectly as possible. The temp @+-15v is down to a rise of 26deg c over ambient. Thd+n is looking pretty close to direct connect. The oscilloscope is not showing anything like the images @Mooly pointed me to.
Even when running the SilentSwitcher from battery, and putting the circuit in a small aluminum chassis grounded to signal/SilentSwitcher ground, I get some mains harmonics that I don’t get when directly connecting the dac to the adc( everything in the setup is running on battery).
I’ll have to wrestle with some better shielding after I solder it up. Should I run a separate ground wire to the chassis for the 2 10uf power supply bypass capacitors?
Thx everyone for all of your time and great suggestions. This is an amazing forum!
Even when running the SilentSwitcher from battery, and putting the circuit in a small aluminum chassis grounded to signal/SilentSwitcher ground, I get some mains harmonics that I don’t get when directly connecting the dac to the adc( everything in the setup is running on battery).
I’ll have to wrestle with some better shielding after I solder it up. Should I run a separate ground wire to the chassis for the 2 10uf power supply bypass capacitors?
Thx everyone for all of your time and great suggestions. This is an amazing forum!
Tried that but the single greatest contributor to the mains harmonics going up considerably is when the dac is connected to laptop power( again no ac in the setup). Is it just the usb cable from the laptop picking up stray ac?
Also a pic of setup was requested earlier and I finally resoldered it.
Also a pic of setup was requested earlier and I finally resoldered it.
I suppose its possible but it equally might be something else. It all comes down to detailed logical tests to see where the noise enters.
For example in your picture you could try unsoldering the orange and yellow input leads and tagging those to the ground of the sockets. Having done that couple it all up (so now just the ground connection is connected via the leads) and see if the noise is still there.
Those sockets do not look isolated (insulated) from the chassis which may play a part.
For example in your picture you could try unsoldering the orange and yellow input leads and tagging those to the ground of the sockets. Having done that couple it all up (so now just the ground connection is connected via the leads) and see if the noise is still there.
Those sockets do not look isolated (insulated) from the chassis which may play a part.
The decoupling caps are 10uf np. I didnt have anything that large in film. The larger ones are 220uf np ( but thise do nothing except when I usethe ac supply). All these ac mains harmonics tests have been running on 2 9v batteries.
@Mooly Yellow is input and orange is output( its 1 channel only). So you mean ground the yellow, plug the dac ouput into that jack( where it does normally go) and check for
mains harmonics?
The body of those jacks are grounded. Should I undo that?
@Mooly Yellow is input and orange is output( its 1 channel only). So you mean ground the yellow, plug the dac ouput into that jack( where it does normally go) and check for
mains harmonics?
The body of those jacks are grounded. Should I undo that?
I notice your 'local' decoupling caps are 100x the value being recommended, and the wrong type. 100nF (0.1uF) are worth buying in bulk for this purpose in future projects.
OK, so unsolder yellow and tag it to the green ground wire. See if its silent like that.
If not then remove the sockets from the metal chassis and just have them free as a test.
Actually found some 100nf but they made no difference at least with battery power. Will test with bench supply later.
@Mooly shorted input but hard to tell ac harmonics from noise. However, taking the whole thing out of the aluminum chassis and thereby the signal grounds out of the chassis and picked up about .8 db in noise improvement.
First pic is direct connect rca into left and right channels of dac but signal only output on left channel into xlr into adc. Second pic is circuit no longer in the aluminum chassis. Still have thise few ac spikes but thd and thd+n are very close. Maybe I should put it in a cardboard box!
@Mooly shorted input but hard to tell ac harmonics from noise. However, taking the whole thing out of the aluminum chassis and thereby the signal grounds out of the chassis and picked up about .8 db in noise improvement.
First pic is direct connect rca into left and right channels of dac but signal only output on left channel into xlr into adc. Second pic is circuit no longer in the aluminum chassis. Still have thise few ac spikes but thd and thd+n are very close. Maybe I should put it in a cardboard box!
You really need to prove that to yourself in the first place.
If it's measurable then your circuit has a problem cause no AP can measure it with lm4562.If you can hear it in a properly designed circuit, then you need to prove it in a blind test with 10...1 million participants.
Every single time I hear that a unity gain buffer with a modern op-amp sounds better than another unity gain stage employing another modern op-amp, both being unity gain stable and recommended for audio use, i can only think about bad pcb practice, non compensated circuits or a possible impedance mismatch.I simply can't figure of any other possibility left.
Aren't these circuits supposed to be shielded, surrounded by ground planes,having decoupled supplies, etc?... in the old times there was always a 100pF...1nF cap shunting the inputs to ground or a very simple first order rc filter aimed at killing rf on the input...I see no reason making advanced measurements on breadbord circuits unless you really did the best job you can on it.If i see double sided breadboard pcb I look for vias stitching in the first place cause all the pcb vias left untouched are simple little antenas and capacitors distributed all over your pcb so if you know that your circuit is RF susceptible like having 55 Mhz bandwith...then you should really consider vias stitching on prototyping boards.
Last edited:
@dreamth this is not related to audibility at all. It is a piece of a very low noise equipment test setup. As you can see the results with my circuit in place are only .6db worse than direct connection of the dut to the adc so I’m calling it good untiand if I decide I need to do a pcb, better grade parts, etc.
I have gone through a wide variety of bypass and coupling configurations and the current one is closest to direct connection of the dut and adc.
At this point I can now even use the lowest voltage input setting on the adc which has an input impedance of 225 ohms without concern of output impedance of circuits or devices being too high for it.
@Mooly interestingly the ac harmonics are predominately odd order. I also discovered that the output of the dac, not connected to my buffer, has a dc offset that has a full wave time of 16 seconds and goes up and down at a very regular rate from .6mv to 1.4mv. Any ideas what that is? It is reflected in the buffer output.
I have gone through a wide variety of bypass and coupling configurations and the current one is closest to direct connection of the dut and adc.
At this point I can now even use the lowest voltage input setting on the adc which has an input impedance of 225 ohms without concern of output impedance of circuits or devices being too high for it.
@Mooly interestingly the ac harmonics are predominately odd order. I also discovered that the output of the dac, not connected to my buffer, has a dc offset that has a full wave time of 16 seconds and goes up and down at a very regular rate from .6mv to 1.4mv. Any ideas what that is? It is reflected in the buffer output.
Is its frequency dependent on the DAC clock frequency? Any relation with the resolution of the incoming signal? What kind of DAC is it?
A sigma-delta DAC could produce a 62.5 mHz (millihertz) tone if there were a very small DC offset at the digital input, like the 0.5 LSB or 1 LSB offsets you can get from rounding or inverting signals, but I would then expect the level to be well below 0.8 mV peak to peak.
A sigma-delta DAC could produce a 62.5 mHz (millihertz) tone if there were a very small DC offset at the digital input, like the 0.5 LSB or 1 LSB offsets you can get from rounding or inverting signals, but I would then expect the level to be well below 0.8 mV peak to peak.
Instead of directly grounding the +input pin of the unused side, I would ground it through a 1K or so resistor. Also I would check to be sure I didn’t have too many stupid pills today and confused the inverting and non-inverting inputs on the unused side or where the ground connections actually are!
Opamps are interesting in that they do not directly connect to ground! The power supply center is considered ground and the bypass capacitors are connected to it. So the opamp’s signal reference connects to the ground otherwise you will see strange results.
When an opamp is marginally stable it will work better with a gain of 3 to 10. To keep the system gain at unity an input resistor voltage divider is used to reduce the input by the same amount as the gain! Seems silly but does give better stability.
Opamps are interesting in that they do not directly connect to ground! The power supply center is considered ground and the bypass capacitors are connected to it. So the opamp’s signal reference connects to the ground otherwise you will see strange results.
When an opamp is marginally stable it will work better with a gain of 3 to 10. To keep the system gain at unity an input resistor voltage divider is used to reduce the input by the same amount as the gain! Seems silly but does give better stability.