I have constructed a Bridge-Parallel GainClone based on the National Semiconductor BPA-200. The input is changed to a LM6172 connected as attached circuit diagram. I have omitted the Cf, feeback capacitors.
The amp functions nicely except that I get a hissing sound (in addition to the music) when connected the amp to the speakers. I have also measured a DC offset of about 0.16V across the output terminal. Is this offset too excessive?
I have also tried changing the LM6172 to OPA2132, the hiss remains. I do not know what could be the problem. Could it be because signal wires (un-shielded) are picking up interference? I have the amps squeezed into a compact chassis. Or could it be due the omission of the feedback capacitors?
What could be the cause of the problem?
The amp functions nicely except that I get a hissing sound (in addition to the music) when connected the amp to the speakers. I have also measured a DC offset of about 0.16V across the output terminal. Is this offset too excessive?
I have also tried changing the LM6172 to OPA2132, the hiss remains. I do not know what could be the problem. Could it be because signal wires (un-shielded) are picking up interference? I have the amps squeezed into a compact chassis. Or could it be due the omission of the feedback capacitors?
What could be the cause of the problem?
Attachments
output DC offset
I intent to deal with the problem of DC offset by tuning each individual opamp with the method in attached diagram. The pot will be tuned till the potential difference between the output pin of the opamp and the ground is zero.
Would this work?
Thanks & Regards,
SW Yeo
I intent to deal with the problem of DC offset by tuning each individual opamp with the method in attached diagram. The pot will be tuned till the potential difference between the output pin of the opamp and the ground is zero.
Would this work?
Thanks & Regards,
SW Yeo
Attachments
sounds like oscillation - this could explain the hiss and the offset
are the chips seperately decoupled ?
.1uf - 1uf across the supply pins of each chip is normal
are the chips seperately decoupled ?
.1uf - 1uf across the supply pins of each chip is normal
You've coupled the inverting stage via 4,7uF capacitors, but the non-inverting output stage directly.So the non inverting stage amplifies the offset of the driver stage.
Try the bypass capacitors for removing the hiss.
Arne
Try the bypass capacitors for removing the hiss.
Arne
Oh i've overseen the input circuit.... Why do you scale down the input by a factor of 10 just to amplify it again by the factor of 10?
Arne
Arne
Not that I know much about these things but it seems you are driving one inverting and one non-inverting amp each with each of the 6172 buffers. Any difference in the amplification of the two buffers will show up as a difference in output from the two bridge outputs. I would run the two inverting amps from the same buffers and the two non-inverting from the other one.
This circuit is unreasonably complicated, strange and simply wrong. Why do you have to imrovise when there are so many working circuits? Even from the application notes. Attenuate befor amplifiying?! If you don't want strange phenomena occuring use separate coupling caps at each stage and a single opamp will suffice for all 4x3875. If you prefer using 2 opamps why don't you do the inversion by the opamps and have all 4x3875 hooked as inverting?
I'm right know in the study process since I'm designing a Gainclone but 10 kohms in series with + input, is that wise? I'm thinking of noise. The feedback network is also a little bit high ohmish for my taste. Lower values will decrease the influence of stray capacitances. You'll get "purer" feedback, more ideal.
EDIT: How about total gain? Way too much! The input buffer should be a buffer, nothing else. And the total gain should be 19-25 (1k + 18 k) according to my taste.
This input buffer with a voltage divider generates only noise. It's just a waste of good properties. If you want some input protection is this just wrong. I doubt that you really need any extra protection. If the chosen opamp can't handle low gain, you have chosen the wrong part. Maybe AD8610/20?
You have also floating inputs (in the picture), must be connected somewhere.
I strongly recommend that you take a closer look at AN-1192 and change only the parts you have control over. This application note is a very good start. I recommend also that you use DC-servo so you can use low resistor values and don't use elco's in the feedback.
I recommend also that you pull down the input cap with 1 Mohms or anything. Don't leave the cap floating, can pick up charges and creates high transients if you connect the amp with the power on.
You should have a lowpass filter at the input. Set the frequency to 50-200 kHz, as a precaution
EDIT: How about total gain? Way too much! The input buffer should be a buffer, nothing else. And the total gain should be 19-25 (1k + 18 k) according to my taste.
This input buffer with a voltage divider generates only noise. It's just a waste of good properties. If you want some input protection is this just wrong. I doubt that you really need any extra protection. If the chosen opamp can't handle low gain, you have chosen the wrong part. Maybe AD8610/20?
You have also floating inputs (in the picture), must be connected somewhere.
I strongly recommend that you take a closer look at AN-1192 and change only the parts you have control over. This application note is a very good start. I recommend also that you use DC-servo so you can use low resistor values and don't use elco's in the feedback.
I recommend also that you pull down the input cap with 1 Mohms or anything. Don't leave the cap floating, can pick up charges and creates high transients if you connect the amp with the power on.
You should have a lowpass filter at the input. Set the frequency to 50-200 kHz, as a precaution
yeti said:
I have read that opamp functions better with with higher gain. Even though LM6172 can operate at unity gain. I thought it might sound better this way.
analog_sa said:
The circuit is basically the same as the Figure 7 of NS BPA-200, other than the input opamp and the difference in the value of the resistors. Using two opamp/ or a dual opamp here, one opamp driving two LM3875s in parallel each with a input resistance of 10k Ohm, thus a input resistance of 5k ohm. 5k ohm of input resistance is a easiler load for the buffering Opamp compare to 2.5kOhm of 4 LM3875s amps in parallel. I do not how much this affect the sound though.
I think this might be an interference problem, input is too close to output. I remember there wasn't any hissing when I was testing the amp with the input buffer out of the chassis. The hiss only appears after I put the input buffer into the chassis. The chassis is quite packed.
Under this situation, would the use of shielded signal wire helps? I am using unshielded wire of very thin guage at the moment.
Can anyone advice me if the measure to get rid of the output dc offset voltage is going to work?
Thanks & Regards,
SW Yeo
Floating inputs
You must ensure that the two opamps that are not used don't have floating inputs.If they oscillate at RF it will affect all the other opamps and that can be audible in the audio range due to intermodulation products.
Connect the unused - ve input to the output of the respective amp. That brings down gain to 1 . Then you can ground the + input - possibly through a low value resistor. This is possible only if the amp is sttable at unity gain. If not, set it at some stable gain value.
You must ensure that the two opamps that are not used don't have floating inputs.If they oscillate at RF it will affect all the other opamps and that can be audible in the audio range due to intermodulation products.
Connect the unused - ve input to the output of the respective amp. That brings down gain to 1 . Then you can ground the + input - possibly through a low value resistor. This is possible only if the amp is sttable at unity gain. If not, set it at some stable gain value.
Re: Floating inputs
I don't actually get you. All the opamps are used.
The hiss is such that there is still music being heard.
ashok said:
I don't actually get you. All the opamps are used.
The hiss is such that there is still music being heard.
those input resistors don't determin the input impedance - they're there to make the inside transitors work ok......use 1 kohm. the lm38xx input impedance is much higher btw...and don't use non inverting and inverting in a balanced circuit ( both "legs have to be exactly the same), in your case its much better to use non inverting mode and make a good line driver and do the inversion there....opa627 - lpf - drv134 and good powersupply.
Re: Re: Floating inputs
By hiss, do mean noise or what?
I think you got it wrong by having attenuation and in the same time gain. It's like controlling speed in a car with the brakes.
One measure to rule out the LM3875 och to shorten all inputs of the LM3875 to ground, + input for the non-inverting and - input for the inverting. If you still have noise you have the problem there.
OK, the schematics wasn't clear. I have only one symbol for an opamp and addtional symbols for power. You haven't written anywhere IC1, IC2 etc so it was a little bit hard to follow and understand what you ment.overmind said:
By hiss, do mean noise or what?
I think you got it wrong by having attenuation and in the same time gain. It's like controlling speed in a car with the brakes.
One measure to rule out the LM3875 och to shorten all inputs of the LM3875 to ground, + input for the non-inverting and - input for the inverting. If you still have noise you have the problem there.
Schematic was not clear.
Like Peranders said , the circuit diagram is misleading. The two opamps on the left top and bottom look like they are floating.
I was referring to them. Now you say it only shows the supply connections.
Cheers.
Like Peranders said , the circuit diagram is misleading. The two opamps on the left top and bottom look like they are floating.
I was referring to them. Now you say it only shows the supply connections.
Cheers.
You have only 1000 uF as decoupling. 100 nF (polyester, ceramic) close to the pins of LM3875 is strongly recoemmended if not totally necessary.
A circuit without decoupling is asking for trouble whatever the circuit is.
A circuit without decoupling is asking for trouble whatever the circuit is.
decoupling caps are crucial to be close.... the power supply caps are not crucial.....you need a low impedance to ground etc...etc...
peranders said:
JOE DIRT® said:
Each LM3875 opamp is hardwired and the components are placed closely. I have tried placing the by-passing cap as close to the pins as possible. The hiss is still there. I am using Wima 0.22uF 160V MKP10 caps for by passing.
I have constructed the LM3875 opamps as 4 modules mounted on heat sinks. The ground of each module is joined by the a straight piece of magnet wire, diameter 1.2mm.
Each of the these module works nicely individually, hissing appearing as they are paralleled and bridged.
I am getting very loss. What could be the problem?
Re: Re: Schematic was not clear.
Have you tested to have output filter (L//R) and Zobel network.
Have you tested just to connect to one half, to rule out the bridging thing? Have you tried only to use two IC, noone in parallel, to rule out the paralleling?
Hardwiring and 10 A devices is not trivial I suppose.
I suspect this is oscillation. Put a radio near and see if you can get it silent or pick up noise.overmind said:
Have you tested to have output filter (L//R) and Zobel network.
Have you tested just to connect to one half, to rule out the bridging thing? Have you tried only to use two IC, noone in parallel, to rule out the paralleling?
Hardwiring and 10 A devices is not trivial I suppose.
- Status
- Not open for further replies.