The schematic is in the first post of this very thread. There is a cap
at the input of the buffer. The offset is almost zero. I do not know what is the
problem. The volume control is a 64 step relay driven attenuator. Could
one of the relay be a problem?
Thanks
Roushon
Sorry for missing the first post - been quite a while since it was posted.
There is no bandwidth limiting anywhere in that circuit. Did you build it without any? It will be very sensitive to RF pickup without. It is possible to get away with it, but it may still oscillate a bit when it is connected like this.
I am assuming each half of the circuit works on its own and only when connecting the two you get the interference issue? Is the circuit properly grounded - I had a similar problem with a ground loop a while back - insulating the input sockets from the chassis fixed it for me.
Why would you need a T-network in the feedback if you already have a source buffer? Try it without - a high input impedance and poor grounding can lead to instability. I would try the entire circuit without a servo first - I know you did this, but maybe just taking it one step at a time would help - first just the LM3886, then add the buffer and an input cap and finally the servo. this will help you pinpoint the issue. Good luck. I would suggest reading through some of the suggestions made in the first few days, specially by Hurtig. Just throwing things together will likely not work - get one block to work at a time first.
There is no bandwidth limiting anywhere in that circuit. Did you build it without any? It will be very sensitive to RF pickup without. It is possible to get away with it, but it may still oscillate a bit when it is connected like this.
I am assuming each half of the circuit works on its own and only when connecting the two you get the interference issue? Is the circuit properly grounded - I had a similar problem with a ground loop a while back - insulating the input sockets from the chassis fixed it for me.
Why would you need a T-network in the feedback if you already have a source buffer? Try it without - a high input impedance and poor grounding can lead to instability. I would try the entire circuit without a servo first - I know you did this, but maybe just taking it one step at a time would help - first just the LM3886, then add the buffer and an input cap and finally the servo. this will help you pinpoint the issue. Good luck. I would suggest reading through some of the suggestions made in the first few days, specially by Hurtig. Just throwing things together will likely not work - get one block to work at a time first.
i've been running your typical BPA-200 since mid 2008 with zero problems. I used the exact chips for the servo's that national semiconductor called for as well as all of the .1% resistors. The amplifier is DC coupled to an active crossover.
Amplifier produced just under 220W rms into an 8 ohm resistive load and has close to 0mv offset. Its one of my favorite amps actually.
Amplifier produced just under 220W rms into an 8 ohm resistive load and has close to 0mv offset. Its one of my favorite amps actually.
Roushon,
I've Spiced the poweramp part of your schematic - the LF gain peaks around 62dB at 100Hz. Is this right? If so, that's your problem when paralleling the two amps - the gains won't match closely enough at low frequencies, leading to the two halves fighting each other. To fix this, you'll need to get some very close tolerance caps for your 470ns in the servo, or redesign the circuit substantially.
I've Spiced the poweramp part of your schematic - the LF gain peaks around 62dB at 100Hz. Is this right? If so, that's your problem when paralleling the two amps - the gains won't match closely enough at low frequencies, leading to the two halves fighting each other. To fix this, you'll need to get some very close tolerance caps for your 470ns in the servo, or redesign the circuit substantially.
Thanks a lot...
Thanks a lot to all of you for your kind suggestions especially to Sangram.
I agree, I have overdone the circuit at the very starting. I should elaborate
on the final design.
1. There is only one buffer.
2. There is a 220pF silver mica cap between the inputs of LM3886
3. There is a zobel (4.7Ohm+0.1mfd) at the output of LM3886
4. Each and every ground returned to the star ground, this made my
life a bit difficult as there are four amps and two buffers.
5. There is a relay driven 64 step attenuator after the buffer. I
wanted to use this before the buffer but the datasheet of OPA627
says that input resistance of more than 100Ohms introduces noise.
The input resistance of the OPA is now 82Ohms.
6. There is input protection using 2N4117A to the Buffer following
suggestions in the datasheet of OPA627.
I have noticed that the AM radio audio noise appears as soon as
I connect the attenuator. The buffer has no function to this noise.
The input as was well as all grounds are returned to the star, so there
is absolutely no chance of a ground loop unless there is some accidental
ones. This weekend I plan to rewiring the amp. About the t-network
I will try without it. The heat sink is enough I feel as there is one big heat sink directly on the IC and the whole heatsink is screwed with the chasis.
But I have not touched to check sudden increase in heat. I will check that
too. The amp is usually working except in high volume. And the sound is fantastic. Regarding 470n cap I am not sure how close they are. But
I see there is no problem regarding Offset.
Regards
Roushon.
Thanks a lot to all of you for your kind suggestions especially to Sangram.
I agree, I have overdone the circuit at the very starting. I should elaborate
on the final design.
1. There is only one buffer.
2. There is a 220pF silver mica cap between the inputs of LM3886
3. There is a zobel (4.7Ohm+0.1mfd) at the output of LM3886
4. Each and every ground returned to the star ground, this made my
life a bit difficult as there are four amps and two buffers.
5. There is a relay driven 64 step attenuator after the buffer. I
wanted to use this before the buffer but the datasheet of OPA627
says that input resistance of more than 100Ohms introduces noise.
The input resistance of the OPA is now 82Ohms.
6. There is input protection using 2N4117A to the Buffer following
suggestions in the datasheet of OPA627.
I have noticed that the AM radio audio noise appears as soon as
I connect the attenuator. The buffer has no function to this noise.
The input as was well as all grounds are returned to the star, so there
is absolutely no chance of a ground loop unless there is some accidental
ones. This weekend I plan to rewiring the amp. About the t-network
I will try without it. The heat sink is enough I feel as there is one big heat sink directly on the IC and the whole heatsink is screwed with the chasis.
But I have not touched to check sudden increase in heat. I will check that
too. The amp is usually working except in high volume. And the sound is fantastic. Regarding 470n cap I am not sure how close they are. But
I see there is no problem regarding Offset.
Regards
Roushon.
Assuming the circuit is fine and stable without the relay attenuator,
Can you show us a wiring diagram of the overall circuit including the source and input, plus how all the grounds are connected? Maybe you will find a clue there. Obviously something is picking up RF.
I would still move the relay attenuator to before the 627. In passive mode the input resistance of the buffer is equal to the source output resistance, so an attenuator will not create problems when inserted in the middle. OTOH, when the servo is correcting the variable offset typical of changing input impedance when the relays change step, it could create issues at some steps (typically half volume is worst case for most amps). AFAIK the attenuator is basically looking into an input impedance of about 11K, and this may cause issues when the ground leg resistor reaches that value.
I would simply knock the servo off the circuit unless you absolutely have to have it. IME a parallel chipamp has no issues with offset even without - I run a pair all day (non-inverting) with no fancy servos/input buffer/attenuator, and it runs cool and stable, with great sound and ear-splitting volumes.
Also, I would change the 220pf input cap to about 22pF, or connect it to the ground instead of the + input. It just seems too high for me, and since you are connecting a servo to the positive input, some of that may be finding its way into the - input as well.
Can you show us a wiring diagram of the overall circuit including the source and input, plus how all the grounds are connected? Maybe you will find a clue there. Obviously something is picking up RF.
I would still move the relay attenuator to before the 627. In passive mode the input resistance of the buffer is equal to the source output resistance, so an attenuator will not create problems when inserted in the middle. OTOH, when the servo is correcting the variable offset typical of changing input impedance when the relays change step, it could create issues at some steps (typically half volume is worst case for most amps). AFAIK the attenuator is basically looking into an input impedance of about 11K, and this may cause issues when the ground leg resistor reaches that value.
I would simply knock the servo off the circuit unless you absolutely have to have it. IME a parallel chipamp has no issues with offset even without - I run a pair all day (non-inverting) with no fancy servos/input buffer/attenuator, and it runs cool and stable, with great sound and ear-splitting volumes.
Also, I would change the 220pf input cap to about 22pF, or connect it to the ground instead of the + input. It just seems too high for me, and since you are connecting a servo to the positive input, some of that may be finding its way into the - input as well.
220pf to 330pf is the standard RF filter capacitor value that goes across the inputs of ICs like the LM3886. There is no way anything as low frequency as the output of a servo will go through a 220pf capacitor.
Do you have one zobel for each LM3886?
A gain of 62dB @ 100Hz sounds very wrong to me.
Do you have one zobel for each LM3886?
A gain of 62dB @ 100Hz sounds very wrong to me.
all seems well
I did the rewiring of all the power cables placing them far away
from the bass of ground cables. The problem seems to be solved. Why?
Now I can play and raise volume as much as I want, no noise appears. But,
there is still the very slight RF noise in one channel making the sound a bit
dull, the noise is there even when there is no signal. The noise does not
increase with volume. The other channel is crystal clear, in fact when there
is no signal it is silent.
I am sorry, I do not have a digital camera at the moment, so not able to
show any picture. In the near future I will post.
There is zobel for each LM3886.
It is puzzling as all the amps are identical and placed symmetrically on the
chasis around the star ground. Still two of them picking RF and the other
two are not.
Thanks
Roushon
I did the rewiring of all the power cables placing them far away
from the bass of ground cables. The problem seems to be solved. Why?
Now I can play and raise volume as much as I want, no noise appears. But,
there is still the very slight RF noise in one channel making the sound a bit
dull, the noise is there even when there is no signal. The noise does not
increase with volume. The other channel is crystal clear, in fact when there
is no signal it is silent.
I am sorry, I do not have a digital camera at the moment, so not able to
show any picture. In the near future I will post.
There is zobel for each LM3886.
It is puzzling as all the amps are identical and placed symmetrically on the
chasis around the star ground. Still two of them picking RF and the other
two are not.
Thanks
Roushon
The two channels have different power supply. Could the PS of the
faulty channel introduce RF? Both the amps in the faulty channel
gives the RF noise individually also.
Actually I am trying to find the solution without making
a drastic change in the whole project because that will be really a
hard work.
Thanks
Roushon
faulty channel introduce RF? Both the amps in the faulty channel
gives the RF noise individually also.
Actually I am trying to find the solution without making
a drastic change in the whole project because that will be really a
hard work.
Thanks
Roushon
Seemed strange to me too. Roushon, is there an error in the schematic you've posted? You have the LM3886 set up for a gain of X1000 (1k input, 10k feedback and 100X T divider) which results in a response droop at the top end - it can't manage this gain at 20kHz due to only having a GBW of 8MHz typically.
Thanks Abraxalito for pointing out the error again which I overlooked last
time. I will remove the T-network for improvement of the sound quality.
Regards
Roushon.
If you just remove it, the LM3886 will go unstable, so I'd just remove the 100R to ground. This will leave you with a gain of X20 (two 10ks in the feedback). You'll need to bring the volume control up quite a long way to compensate for the loss of so much gain, but at least your HF response won't roll off.
I meant removing the 100R only.
I noticed one thing yesterday, (without any signal at the input) if I give the
maximum volume then the RF vanishes.
Still, with the t-network, the other channel is silent (absolutely no RF or
any other noise). Don't know why? May be, the source of the RF is near
the other channel which I need to find out.
Roushon.
t-network removed
Finally removed the t-network. Now there is only loud noise from
all the four amp individually (without any input connection). Voltage droping by around 20v (both + and -) the moment the mute goes off (without speaker). What to do? Still not able to show any picture. Very strange
situation, only removed the 100R and no other change, now the feedback is 20K for all the four amp.
Regards
Roushon.
Finally removed the t-network. Now there is only loud noise from
all the four amp individually (without any input connection). Voltage droping by around 20v (both + and -) the moment the mute goes off (without speaker). What to do? Still not able to show any picture. Very strange
situation, only removed the 100R and no other change, now the feedback is 20K for all the four amp.
Regards
Roushon.
So the pairs of amps are no longer connected together via the 0.1ohm sharing resistors? Sounds like major oscillations are occurring which send your LM338s into current limiting. Taking out the 100Rs has made the LM3886s more critical of any layout/wiring problems you might be having. Having SPICEd this new circuit, you'll have a very nice flat frequency response (servo gives -3dB at 1Hz) and won't need to worry about matching those 470ns when you eventually get it stable....
Still fighting to get things right. Removing the 100R made everything fall
apart, but I am sure it will pay off. Tried to pull down the input of LM3886 by
10K to ground, did not work. Next removed the attenuator, connected the
amps directly to the output of buffer and used the sginal source (an AIWA CD
player) volume control to raise the volume. Then absolutely no noise and
both the channel played smoothly. Now (reluctantly) trying to shift the
attenuatur to the input of the buffer. Can I keep the attenuator in between
the buffer and amp after making some changes? I will attach the attenuator
circuit in the next post.
Regards
Roushon
attenuator shifted
Finally checked after shifting the attenuator at the input of the buffer. No problem
now, the sound is excellent except for two minor issues:
1. Need to raise the volume almost halfway to make the amp audible, which
I do not mind much.
2. On and off a motorboating sound appears in one channel (this is the
channel which was picking the RF before removing the t-network). I hope a
rewiring will solve this problem.
Thanks a lot to everybody who posted in this thread for help and discussion
and of course I am very grateful to this forum where I began and started
learning from scratch.
Regards
Roushon
Finally checked after shifting the attenuator at the input of the buffer. No problem
now, the sound is excellent except for two minor issues:
1. Need to raise the volume almost halfway to make the amp audible, which
I do not mind much.
2. On and off a motorboating sound appears in one channel (this is the
channel which was picking the RF before removing the t-network). I hope a
rewiring will solve this problem.
Thanks a lot to everybody who posted in this thread for help and discussion
and of course I am very grateful to this forum where I began and started
learning from scratch.
Regards
Roushon
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.