just a general note on the value of resistors for lm3875. the feedback resistors in the Alesis amp is 20K, and the ones in the JBL active speaker is about 64K for woofers and 20K for the tweeter. the higher value for the woofers I suppose is to avoid using excessive large capacitors.
I will try to use either a 200K and a 20K resistor to see if there is a difference (I doubt it).
I will try to use either a 200K and a 20K resistor to see if there is a difference (I doubt it).
Whoa, gone for a while and info abounds!
I had to take down my computer set-up to allow for room to tile a floor, so I've been offline for a while. Thanks everyone for keeping the great information flowing.
moamps: Thanks for the cap position recommendation and the comments about R5C5. I think I am on track here, but will double check values and positions, just in case. I also am going to make an RCA to banana wire, as a true probe should not be required for the mV measurements. Might omit a possible cause of oscillation induced by measurement method.
gfiandy, ukram, ron clarke and millwood: Thanks for the information regarding bias current. Seems intuitive, but it helps clarify a possible cause. I'll try to check it out.
sam9: Yes, my current layout is trial based, as my power connectors aren't here and therefore some wiring is temporary. I'll be careful when routing the final wires.
Thanks again for the help. I'm not sure if I'll get a chance to try other solutions quite yet, as the tile floor is still not done. Just had to take a break and check in!
Sandy.
I had to take down my computer set-up to allow for room to tile a floor, so I've been offline for a while. Thanks everyone for keeping the great information flowing.
moamps: Thanks for the cap position recommendation and the comments about R5C5. I think I am on track here, but will double check values and positions, just in case. I also am going to make an RCA to banana wire, as a true probe should not be required for the mV measurements. Might omit a possible cause of oscillation induced by measurement method.
gfiandy, ukram, ron clarke and millwood: Thanks for the information regarding bias current. Seems intuitive, but it helps clarify a possible cause. I'll try to check it out.
sam9: Yes, my current layout is trial based, as my power connectors aren't here and therefore some wiring is temporary. I'll be careful when routing the final wires.
Thanks again for the help. I'm not sure if I'll get a chance to try other solutions quite yet, as the tile floor is still not done. Just had to take a break and check in!
Sandy.
Given that there's barely anything by way of component values to get wrong, I'd start looking at circuit layout and grounding issues.
Firstly, did you try disconnecting the input lead at the PCB altogether and soldering a short between C1 and the PCB ground? If this stops the oscillation, the problem may be related to the input wiring; you can then work backwards towards the input socket.
Your diagram seems to imply that the input ground, and the ground on the volume control, are connected to anywhere that's handy (e.g. the case). If so, this is generally bad form - the input ground, volume pot ground, and the ground ends of R2 and R3 are all 'signal' ground, and sensitive to noise injection. You can connect them together any way you want, but there should be a single connection from this signal ground to the PSU ground. In particular, don't let the ground ends of the decoupling capacitors interfere with the signal ground, and don't use the case as signal ground.
The main cause of the DC offset is the lack of a DC blocking cap between R2 and ground; this gives the amp a gain of 22 (rather than 1) at DC, and makes any input offset errors significantly worse. I assume you've omitted the cap because of some views about its impairing sound quality - if so, 100mV of DC offset may be the price you have to pay. It's probably harmless, but then so is a (well-chosen) cap.
Cheers
IH
Firstly, did you try disconnecting the input lead at the PCB altogether and soldering a short between C1 and the PCB ground? If this stops the oscillation, the problem may be related to the input wiring; you can then work backwards towards the input socket.
Your diagram seems to imply that the input ground, and the ground on the volume control, are connected to anywhere that's handy (e.g. the case). If so, this is generally bad form - the input ground, volume pot ground, and the ground ends of R2 and R3 are all 'signal' ground, and sensitive to noise injection. You can connect them together any way you want, but there should be a single connection from this signal ground to the PSU ground. In particular, don't let the ground ends of the decoupling capacitors interfere with the signal ground, and don't use the case as signal ground.
The main cause of the DC offset is the lack of a DC blocking cap between R2 and ground; this gives the amp a gain of 22 (rather than 1) at DC, and makes any input offset errors significantly worse. I assume you've omitted the cap because of some views about its impairing sound quality - if so, 100mV of DC offset may be the price you have to pay. It's probably harmless, but then so is a (well-chosen) cap.
Cheers
IH
Double Checking
The kitchen tile fiasco is now over and I can get back to fun hobbies, so its on to implementing changes to the RJM GC to minimize DC offset.
Per Jan and others, I am going to try a DC blocking cap, but wanted to make sure I'm planning correctly.
I have attached a pic of part of the schematic which is affected. Have I placed Cdc correctly to act as a DC blocker? In my mind, R4 and Cdc are forming an RC filter, but I am unsure of how the paralleled R3 affects that RC. Should Cdc be in series with R3?
If my placement assumption is correct and I am shooting for a time constant of 100msec, per Jan's suggestion, for the 22.1K R4, I would use a 470 uf cap. I think this must be a non-polarized cap as well, but I am honestly not sure. The op-amp output, referencing the power ground, can be both +/-, hence an NP cap, correct?
Thanks for everybody's continued help.
Sandy.
The kitchen tile fiasco is now over and I can get back to fun hobbies, so its on to implementing changes to the RJM GC to minimize DC offset.
Per Jan and others, I am going to try a DC blocking cap, but wanted to make sure I'm planning correctly.
I have attached a pic of part of the schematic which is affected. Have I placed Cdc correctly to act as a DC blocker? In my mind, R4 and Cdc are forming an RC filter, but I am unsure of how the paralleled R3 affects that RC. Should Cdc be in series with R3?
If my placement assumption is correct and I am shooting for a time constant of 100msec, per Jan's suggestion, for the 22.1K R4, I would use a 470 uf cap. I think this must be a non-polarized cap as well, but I am honestly not sure. The op-amp output, referencing the power ground, can be both +/-, hence an NP cap, correct?
Thanks for everybody's continued help.
Sandy.
Attachments
Re: Double Checking
Cdc should indeed be in series with R3, not the way you have drawn it. The idea is that Cdc has an impedance which rises towards infinity at DC (Z = 1/2.pi.f.C), meaning that at DC the output is connected directly to the input. This gives unity gain (i.e. less than the x22 AC gain), which should give a lower DC offset.
The time constant in this case will be R3.Cdc; this is because the response will start to roll off when Cdc's impedance is comparable to that of R3.. If R3 = 1k, Cdc of 100uF will give 100ms.
Cheers
IH
Sandy H. said:
Cdc should indeed be in series with R3, not the way you have drawn it. The idea is that Cdc has an impedance which rises towards infinity at DC (Z = 1/2.pi.f.C), meaning that at DC the output is connected directly to the input. This gives unity gain (i.e. less than the x22 AC gain), which should give a lower DC offset.
The time constant in this case will be R3.Cdc; this is because the response will start to roll off when Cdc's impedance is comparable to that of R3.. If R3 = 1k, Cdc of 100uF will give 100ms.
Cheers
IH
Thanks!
Ian: That's why I always ask. I understand just enough to get myself in trouble with most of this. Thanks for the help. In my situation, since R3 is 680 Ohm, the proper cap would be 147 uf. A search said that Digikey only had 100 and 220 uf bi-polar caps, so if I go with 220uf, I'll be around 6Hz and if I choose 100uf, it'd be 15Hz. I guess I'll see what Radio Shack has and go for it. If nothing else, it'll get me closer.
Thanks!
Sandy.
Ian: That's why I always ask. I understand just enough to get myself in trouble with most of this. Thanks for the help. In my situation, since R3 is 680 Ohm, the proper cap would be 147 uf. A search said that Digikey only had 100 and 220 uf bi-polar caps, so if I go with 220uf, I'll be around 6Hz and if I choose 100uf, it'd be 15Hz. I guess I'll see what Radio Shack has and go for it. If nothing else, it'll get me closer.
Thanks!
Sandy.
Attachments
True
True and I honestly didn't consider that. I'd imagine, however, a -3db point of 10Hz is just arbitrarily a number outside of what is considered typical reproduction and a number like 14 Hz wouldn't be all that much worse.
Thanks for bringing it up, though. That's the kind of thing I kick myself for sometimes. . . being too short sighted and closed minded, ignoring the forest for the trees!
Sandy.
True and I honestly didn't consider that. I'd imagine, however, a -3db point of 10Hz is just arbitrarily a number outside of what is considered typical reproduction and a number like 14 Hz wouldn't be all that much worse.
Thanks for bringing it up, though. That's the kind of thing I kick myself for sometimes. . . being too short sighted and closed minded, ignoring the forest for the trees!
Sandy.
Re: True
Making this capacitor bigger than is strictly required (for -3dB at 20Hz) is generally considered a good thing. Electrolytic capacitors are known to produce distortion when there is a significant AC voltage across them (i.e. when they're not just behaving as coupling capacitors). This will be the case at the rolloff point and below, so it's best to put the rolloff point comfortably below any frequencies you expect in the music.
100ms time constant is equivalent to (1 / 2.pi. 0.1s) = about 1.6Hz, which sounds adequate to me. I'd go for either 680R and 100uF (68ms time constant), or 680R and 220uF (about 150ms).
Cheers
IH
Sandy H. said:
Making this capacitor bigger than is strictly required (for -3dB at 20Hz) is generally considered a good thing. Electrolytic capacitors are known to produce distortion when there is a significant AC voltage across them (i.e. when they're not just behaving as coupling capacitors). This will be the case at the rolloff point and below, so it's best to put the rolloff point comfortably below any frequencies you expect in the music.
100ms time constant is equivalent to (1 / 2.pi. 0.1s) = about 1.6Hz, which sounds adequate to me. I'd go for either 680R and 100uF (68ms time constant), or 680R and 220uF (about 150ms).
Cheers
IH
Putting in my digikey order today. . .
For a bi-polar 100uf electrolytic to help with my DC offset issues. I am also going to buy the components to make an inverted model, as I have a chip left over.
Can anyone think of other related parts for the GC, either inverted or non-inverted which I might need. I only ask, as the RJM schematic didn't show the DC Blocking cap, so I didn't order it. I don't mind buying a few extra components if it means doing that vs being off a project for 3 or 4 days while the one off parts come in.
Let me know if you have suggestions from a functionality or tweaking point of view.
Thanks for the help.
Sandy.
For a bi-polar 100uf electrolytic to help with my DC offset issues. I am also going to buy the components to make an inverted model, as I have a chip left over.
Can anyone think of other related parts for the GC, either inverted or non-inverted which I might need. I only ask, as the RJM schematic didn't show the DC Blocking cap, so I didn't order it. I don't mind buying a few extra components if it means doing that vs being off a project for 3 or 4 days while the one off parts come in.
Let me know if you have suggestions from a functionality or tweaking point of view.
Thanks for the help.
Sandy.
Update
I added a 220uf cap per the above recommendations from R3 to ground and indeed the offset went to 1mV. I also finished the umbilical and did some slight re-routing inside. There is no hum present either.
My next step is to modify the attenuator portion of the circuit to limit my output to about 4w into 8 Ohm, as this is only my measurement amp and I wanted to keep the package (and heatsink) small, so don't want to allow it to be driven hard. Also, I want to chose the pot so as to be able to get more control over those 4w to be able for more precise set-up for measurements. I think I understand how to modify the shunt with additional resistors to do this. I'll post back when successful.
Thanks for the help.
Sandy.
I added a 220uf cap per the above recommendations from R3 to ground and indeed the offset went to 1mV. I also finished the umbilical and did some slight re-routing inside. There is no hum present either.
My next step is to modify the attenuator portion of the circuit to limit my output to about 4w into 8 Ohm, as this is only my measurement amp and I wanted to keep the package (and heatsink) small, so don't want to allow it to be driven hard. Also, I want to chose the pot so as to be able to get more control over those 4w to be able for more precise set-up for measurements. I think I understand how to modify the shunt with additional resistors to do this. I'll post back when successful.
Thanks for the help.
Sandy.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.