|
Ground
Nov 6, 2021 13:29:57 GMT
Post by magicwolf on Nov 6, 2021 13:29:57 GMT
Is it possible to separate the input ground from the output ground in headphone amplifiers? My pcb is made in such a way that the bottom and top of the board are a solid ground polygon, and at the same time the board catches a lot of noise, so I thought to separate the input from the output.
|
|
solderdude
Administrator
measureutternutter
Posts: 4,881
|
Ground
Nov 8, 2021 19:41:24 GMT
Post by solderdude on Nov 8, 2021 19:41:24 GMT
Is it possible to separate the input ground from the output ground in headphone amplifiers? My pcb is made in such a way that the bottom and top of the board are a solid ground polygon, and at the same time the board catches a lot of noise, so I thought to separate the input from the output. There is no reason to separate ground between L and R as the signals are always referenced to the exact same voltage. Only when you have fully balanced inputs and outputs you can physically separate ground (basically building 2 full mono amps in one enclosure) but also in this case the ground whether it is connected or not won't make any difference.
|
|
|
Ground
Nov 9, 2021 9:49:29 GMT
Post by magicwolf on Nov 9, 2021 9:49:29 GMT
>There is no reason to separate ground between L and R Thanks for the answer. I didn't want to separate ground between the channels L and R. I wanted to separate ground between the input and output. For approximately marked in red, where I will have to cut the ground polygon. But to avoid noise at the input, you will most likely have to add a 3R resistor between the ground inputs and outputs. I am interested in whether there will be damage to the input op-amp after such a modding?
|
|
solderdude
Administrator
measureutternutter
Posts: 4,881
|
Ground
Nov 9, 2021 17:41:09 GMT
Post by solderdude on Nov 9, 2021 17:41:09 GMT
There has to be an as low as possible resistance in the entire ground path. That's why it is a plane. Signals have to be referenced to ground.
What would be the idea behind this modification ?
|
|
|
Ground
Nov 17, 2021 13:42:18 GMT
Post by magicwolf on Nov 17, 2021 13:42:18 GMT
There is not one plane ground, but two. One on top, the other on the bottom. I think that as a result, an ground loop is formed and therefore there is a noise. It is impossible to break the ground loop. In addition, as I understand it, the ground should be separated into power regulators and op-amps. It may also make sense to separate the ground from the input.
|
|
solderdude
Administrator
measureutternutter
Posts: 4,881
|
Ground
Nov 18, 2021 11:40:49 GMT
Post by solderdude on Nov 18, 2021 11:40:49 GMT
It is quite common to use ground planes on top and bottom for 2 layer designs. For larger PCB's (is not applicable here) it is also done to prevent the PCB's from warping when heated in the soldering process.
Usually each ground connection is connected on both sides of the PCB to form one lower resistance ground plane (effectively) In some designs (mine for instance) also additional vias are used on strategic places. It does not make sense to separate ground from power supplies, inputs and outputs unless you design it as star-ground. It can make sense to connect return wire grounds of output connectors as closely as possible to the ground of the power supply. It also makes sense not to use several points of a chassis as a ground point and run a single wire to the PCB.
The ground should be connected everywhere unless you create a star-ground requiring each reference to run to a specific point.
When there is noise one should check local decoupling from (very fast) opamps.
|
|
|
Ground
Nov 19, 2021 6:04:10 GMT
Post by magicwolf on Nov 19, 2021 6:04:10 GMT
>It is quite common to use ground planes on top and bottom for 2 layer designs. Many thanks! I understood you. At first I thought that the board was not correctly designed. >It can make sense to connect return wire grounds of output connectors as closely as possible to the ground of the power supply. Unfortunately, I did not understand you ;(. In my case, I have to cut the ground layer at the output as shown in the foto(red) and connect it from the ground from the caps of the power regulators (black)? Did I understand you correctly? >The ground should be connected everywhere unless you create a star-ground requiring each reference to run to a specific point. I agree with you, but this means that I have to develop my board ;(. Perhaps this will have to do. >When there is noise one should check local decoupling from (very fast) opamps. Unfortunately, I also did not understand you. Do you mean to increase the input resitor? Or put a capacitor at the input? Attachments:
|
|
solderdude
Administrator
measureutternutter
Posts: 4,881
|
Ground
Nov 19, 2021 14:21:47 GMT
Post by solderdude on Nov 19, 2021 14:21:47 GMT
No need to cut the groundplane, in fact it would be worse. Simply connect the wire that is supposed to go on the red circle and solder it onto the board between - C10 and -C10e
The input can be connected normally to the PCB as designed.
Decoupling seems to be O.K. (maybe a bit overdone even) Personally I would not use 220nF film capacitors as decoupling but use 10nF (or up to 47nF) ceramic multilayer X7R type capacitors instead. Those are the ones marked 224 next to 220uF elco's.
Also I would not mount VR1 and VR2 nor the 10M resistors which is begging for noise ! R13 I would lower to max 4.7 ohm. The volume control I would make 10k log.
|
|
|
Ground
Nov 19, 2021 18:22:58 GMT
Post by magicwolf on Nov 19, 2021 18:22:58 GMT
>Simply connect the wire that is supposed to go on the red circle and solder it onto the board between - C10 and -C10e Thanks! I'll try. >Personally I would not use 220nF film capacitors as decoupling but use 10nF (or up to 47nF) ceramic multilayer X7R type capacitors instead. I use WIMA 100nf. >Those are the ones marked 224 next to 220uF elco's. C10 and Ce10 - 1000uF, C11,Ce11,C15,Ce15... - 47uF >Also I would not mount VR1 and VR2 nor the 10M resistors which is begging for noise Yes, I don't use it. Thanks! So I did it correctly, although the DC output voltage is 100mV. >The volume control I would make 10k log. I use 50K DACT: aliexpress.ru/item/32714092387.html>R13 I would lower to max 4.7 ohm. I understand you, but unfortunately the this schema is designed in such a way that the output must have a resistor of at least 50 ohms ;( IMHO. Otherwise, the output op-amp will not amplify the voltage. When the output is a resistor with 50 ohms or higher, the sound is more interesting. I don't understand how to lower the output impendence. Here is the schema Micro-Cap 12: disk.yandex.ru/d/hEiG9wldGm3PHQHere are the measurements on E-MU 1616m (5R resistor output): disk.yandex.ru/i/-bWdCh7jmNk-dwI use input op-amp LME49710, output - OP01: aliexpress.ru/item/1005003146963494.html
|
|
solderdude
Administrator
measureutternutter
Posts: 4,881
|
Ground
Nov 19, 2021 19:45:02 GMT
Post by solderdude on Nov 19, 2021 19:45:02 GMT
100mV is way too much. I would try to find out if you can trim the OP01.
Maybe go for OPA1611 as OP device.
What would be the reason the amp would need to have at least 50 ohm series resistance ?>
|
|
|
Ground
Nov 20, 2021 15:21:30 GMT
Post by magicwolf on Nov 20, 2021 15:21:30 GMT
>100mV is way too much. ;( I.e. can I damage the headphones?
>Maybe go for OPA1611 as OP device. Instead of the 1st or instead of the second?
>Simply connect the wire that is supposed to go on the red circle and solder it onto the board between - C10 and -C10e I tried - the same noise ;( Maybe the problem is that I do not have a metal chassis?
>What would be the reason the amp would need to have at least 50 ohm series resistance ? I measured voltages (Vrms) at different output resistor (100R, 50R, 4R7). Initially, I found the position of the potentiometer in which there was no clipping (sine 1KhZ). Then I measured at the input to the amplifier, at the input of 1 op-amp, at the output of 1 op-amp, at the output of 2 op-amp and at the output of the amplifier: Input Input 1 Output 1 Output 2 Output 100R - 1,63 1,27 3,04 5,94 1,41 50R - 1,63 0,85 1,98 3,96 1,70 4R7 - 1,63 0,55 1,27 2,69 2,33 Measurements were made at a load of 32 ohms. It can be seen from the measurements that at a resistor of 100 ohms, a signal almost equal to the input (in voltage) is applied to the input of the first op-amp. The output of the second op-amp has the maximum voltage of all three variants. When the output is 100 ohms, the sound is much more interesting. I do not know how to lower the output impedance and at the same time maintain the operating mode of the amplifier, which I like. Perhaps this is due to the potentiometer, but I had an alps before and it was the same.
|
|
solderdude
Administrator
measureutternutter
Posts: 4,881
|
Ground
Nov 20, 2021 19:18:39 GMT
Post by solderdude on Nov 20, 2021 19:18:39 GMT
Some IEM's might be killed this way.
OPA1611 as the second device. You can lower output R to a few ohm.
connecting the output wire to the spot on the PCB might give a theoretical slight advantage that may be measurable. I have no idea where the noise comes from. I would suggest using different opamps first. Try the amp stand alone (nothing connected) in a different room. Sometimes the noise is coming from wireless devices in the proximity.
When doing the measurements you should not touch the volume control which is what you seem to have done. What your measurements show is that the max. output current of the amp is limited to 73mA (100mA peak) but that's all it shows. Strangely enough the OP01 datasheet states 200mA (140mA eff) and it seems to be half of that.
Clearly the lowest output R has the highest output power in 32ohm (assuming is started clipping at that voltage) With 4.7 ohm output R you reached 170mW with 50 ohm output R you reached 90mW with 100 ohm output R you reached 60mW
background noise should be lowest with 100 ohm output R (due to voltage division) The point is there is no need for any amp with 4.7x gain to exhibit any noise at all even in sensitive headphones at 0 ohm output R.
|
|
|
Ground
Nov 24, 2021 6:19:40 GMT
Post by magicwolf on Nov 24, 2021 6:19:40 GMT
>Some IEM's might be killed this way. Thanks for the caution! >OPA1611 as the second device. You can lower output R to a few ohm. Okay I'll try! Once I used them both in the voltage amplification stage and the current amplification stage. But I didn't like them, they have a very neutral sound. >connecting the output wire to the spot on the PCB might give a theoretical slight advantage that may be measurable. I measured, unfortunately the miracle did not happen ;((( >I have no idea where the noise comes from. Thank you so much for this remark, I understand you!!! It looks like the problem is in the transformers. I put 6 transformers (3 - DAC, 2 - rasberry pi 3, 1 - amp) in a separate chassis - apparently it's because of them ;((. For the experiment, I placed the amplifier in a separate chassis with own transformer and did an RMAA tests - the noise peaks decreased. When I inserted a metal sheet between the transformer and the amplifier board, the noise decreased even more. I decided that the noise peaks were due to transformers. I don't know what to do with it yet. I recently bought a channel switch: aliexpress.ru/item/4000103427284.htmlHe also had noises, I found ground loops and eliminated them - the noises decreased dramatically (The peaks disappeared and the noises themselves became less). So I thought I could do something similar with the amplifier board - apparently nothing can be done. >I would suggest using different opamps first. I tried, alas, it does not work. >Try the amp stand alone (nothing connected) in a different room. Sometimes the noise is coming from wireless devices in the proximity. Yes, I tried different variants. Thank you!!! >When doing the measurements you should not touch the volume control which is what you seem to have done. Thanks! I don't touch and the volume control (cap) is made of plastic. >What your measurements show is that the max. output current of the amp is limited to 73mA (100mA peak) but that's all it shows. When the resistor 100R is soldered, the input voltage at the first op-amp is almost the same as the output voltage at the output of the DAC, i.e. the input signal at the first op-amp has minimal distortion. I think this is the first reason for the good sound in this mode. The second reason is the output voltage on the second op-amp - it seems to me that for this op-amp this is the optimal voltage and therefore the sound of this op-amp is very pleasant. _IMHO_ >background noise should be lowest with 100 ohm output R (due to voltage division) The background noise doesn't change, I don't understand why. >The point is there is no need for any amp with 4.7x gain to exhibit any noise at all even in sensitive headphones at 0 ohm output R. I agree with you, so I replaced the resistor R6 (on the ground in the feedback of the first op-amp) with 800R. You can also solder the resistor R11 (the second op-amp to the ground). What do you think?
|
|
solderdude
Administrator
measureutternutter
Posts: 4,881
|
Ground
Nov 24, 2021 9:26:42 GMT
Post by solderdude on Nov 24, 2021 9:26:42 GMT
>OPA1611 as the second device. You can lower output R to a few ohm. Okay I'll try! Once I used them both in the voltage amplification stage and the current amplification stage. But I didn't like them, they have a very neutral sound. So you are not looking to make an amplifier but rather an effect box ? In that case having a bigger output R can change the sound with certain headphones. One of the reason G1217 designs have selectable output R in rather big steps. >connecting the output wire to the spot on the PCB might give a theoretical slight advantage that may be measurable. I measured, unfortunately the miracle did not happen ;((( Shows the groundplane is well constructed. >When doing the measurements you should not touch the volume control which is what you seem to have done. Thanks! I don't touch and the volume control (cap) is made of plastic. I meant not change the settings which appears to have been the case. >background noise should be lowest with 100 ohm output R (due to voltage division) The background noise doesn't change, I don't understand why. Could be: You are looking at the noise floor of the measuring equipment. Due to voltage devision (and you clearly adjusting the input levels) and loading with 32 ohm there MUST be a better S/N ratio on the 32 ohm load when the amp itself is noisy of at least 12dB. I agree with you, so I replaced the resistor R6 (on the ground in the feedback of the first op-amp) with 800R. You can also solder the resistor R11 (the second op-amp to the ground). What do you think? It will lower the overall gain of the amp and subsequently lower the self noise of the opamps. That doesn't seem to be the issue though.
|
|
|
Ground
Nov 24, 2021 17:05:00 GMT
Post by magicwolf on Nov 24, 2021 17:05:00 GMT
>So you are not looking to make an amplifier but rather an effect box ? Well, a little more interesting sound >Could be: You are looking at the noise floor of the measuring equipment. >Due to voltage devision (and you clearly adjusting the input levels) and loading with 32 ohm there MUST be a better S/N ratio on the 32 ohm load when the amp itself is noisy of at least 12dB. Unfortunately, I didn't understand you well . Maybe I'm taking measurements incorrectly? Just in case, the measurements of the EMU 1616M are unbalanced output -> unbalanced input: >That doesn't seem to be the issue though. Most likely yes. Here is a photo of the amplifier:
|
|