Wimhurst, I tried the Zobel, but no difference. So I still think about the vertical mounted transistors (or the bigger caps, what would really surprise me).
Which zener diodes do you use?
Hi Stammheim,
How long are the leads on the vertical transistors? Are they full length to allow for easy soldering?
If so, then that could well be the problem. Having each lead full length would add about 2" round trip to the sensitive G-S path, potentially causing issues. All my amps were built and testing with very short leads and the fets mounted horizontally with an L-bend right at the thin part of the leg.
You could further experiment and try swapping the large bypass caps to the bottom and mount the transistors vertically with short leads. That should also be fine.
Regards,
Owen
How long are the leads on the vertical transistors? Are they full length to allow for easy soldering?
If so, then that could well be the problem. Having each lead full length would add about 2" round trip to the sensitive G-S path, potentially causing issues. All my amps were built and testing with very short leads and the fets mounted horizontally with an L-bend right at the thin part of the leg.
You could further experiment and try swapping the large bypass caps to the bottom and mount the transistors vertically with short leads. That should also be fine.
Regards,
Owen
Many thanks.
Well, here a picture of my built.
I understand, but legs shouldn't be longer than using the L-shape.
Any interferences from the few mm transistors <-> caps?
I also tried to use a copper-shield between them, but no difference.
By the way... sorry for wrong information... the zobel halves the interferences, but it still stayes.
Well, here a picture of my built.
I understand, but legs shouldn't be longer than using the L-shape.
Any interferences from the few mm transistors <-> caps?
I also tried to use a copper-shield between them, but no difference.
By the way... sorry for wrong information... the zobel halves the interferences, but it still stayes.
Last edited:
If the amplifier looks stable with horizontal mounted transistors but unstable with vertical mounted transistors the design is bad or on the edge. Try adding a 10 - 20 pF miller cap parallel to check if the oscillations stop.
Lateral mosfets are very prone to oscillations - every millimeter of wire counts in.
Lateral mosfets are very prone to oscillations - every millimeter of wire counts in.
These Alfets are a design evolution to stop that oscillation problem. Removing the internal protection diodes was part of this the way Semelab achieved this.
Just in case, I did anyway add a 22pF cap across each Alfet Gate and Drain when I used them in a 1986 amplifier I restored last year.
Just in case, I did anyway add a 22pF cap across each Alfet Gate and Drain when I used them in a 1986 amplifier I restored last year.
Last edited:
Thanks for support, guys!
I'd like to avoid soldering a bigger capacity as the Miller cap to keep the slew rate as big as possible.
I’ll try to shorten the Mosfets legs these days, which unfortunatelly requires modifying the mechanical mounting on the radiator.
OK, I’ll give the Miller caps a chance then, when shortening legs won’t help.
For better understanding: What exactly is the Mosfets` problem with „too long wires“? I mean nobody speaks of several meters, but maybe 5mm, which could be reduced using horizontal Mosfets…
Regards
Stammheim
I'd like to avoid soldering a bigger capacity as the Miller cap to keep the slew rate as big as possible.
I’ll try to shorten the Mosfets legs these days, which unfortunatelly requires modifying the mechanical mounting on the radiator.
OK, I’ll give the Miller caps a chance then, when shortening legs won’t help.
For better understanding: What exactly is the Mosfets` problem with „too long wires“? I mean nobody speaks of several meters, but maybe 5mm, which could be reduced using horizontal Mosfets…
Regards
Stammheim
Hi Stammheim,
As Andrew mentioned, the major culprit is the gate lead which represent the "L" in the RLC circuit formed by the gate stopper (R) the lead inductance of the PCB plus the leg of the FET (L) and the capacitance of the gate itself (C).
The longer the gate leg is, the higher the L value is, and the more likely the circuit is to oscillate. Increasing R can help to damp this, or prevent it, but too large a gate stopper can cause other issues.
Try shortening all the leads, or at least the gate lead and see if that helps things.
Regards,
Owen
As Andrew mentioned, the major culprit is the gate lead which represent the "L" in the RLC circuit formed by the gate stopper (R) the lead inductance of the PCB plus the leg of the FET (L) and the capacitance of the gate itself (C).
The longer the gate leg is, the higher the L value is, and the more likely the circuit is to oscillate. Increasing R can help to damp this, or prevent it, but too large a gate stopper can cause other issues.
Try shortening all the leads, or at least the gate lead and see if that helps things.
Regards,
Owen
Owen
It would seem that I was not following this thread for a while, possibly as the site stopped sending me new post updates.
Why did you lower the gate resistor values to 51 Ohms for both, from the 330 and 220? Was there a sonic or measurable benefit? Lower noise?
Can you post a link pointing to a post covering the changes?
Thanks.
It would seem that I was not following this thread for a while, possibly as the site stopped sending me new post updates.
Why did you lower the gate resistor values to 51 Ohms for both, from the 330 and 220? Was there a sonic or measurable benefit? Lower noise?
Can you post a link pointing to a post covering the changes?
Thanks.
Very interesting question!
While increasing the Miller cap didn't bring any change, the Zobel is the solution.
No oscillation anymore. Great hint, thanks!
BTW:
I also tried to shorten the legs to only few mm, but no change at all. I have also two amps, which are mounted on an L-shape. The guys have no oscillating, also without the Zobel. So the amp seems to have an issue with vertical mounted mosfets. Crazy!
But yeeeesss, I can go on with my amp
Best regards
Stammheim
Maybe the Alfets were picking up radiation from the PCB or components? Those very nearby capacitors?
Does your Zobel resistor get warm? I was oscilloscope probing around an amplifier recently and suddenly the scope showed a weird shape and the Zobel resistor smoked badly! The probe didn't stay on anything but must have touched something that maybe set up feedback maybe through the mains earth? Only guessing. Amp was working before and after with no oscillation really, but there is a persistent 50kHz at low voltage with another amplifier also shows. I established that the Hameg scope is affected badly by being plugged into the same mains electrical socket as the computer, it shows wave form distortion that is not present if plugged in somewhere else, or if the computer if off at the time.
Does your Zobel resistor get warm? I was oscilloscope probing around an amplifier recently and suddenly the scope showed a weird shape and the Zobel resistor smoked badly! The probe didn't stay on anything but must have touched something that maybe set up feedback maybe through the mains earth? Only guessing. Amp was working before and after with no oscillation really, but there is a persistent 50kHz at low voltage with another amplifier also shows. I established that the Hameg scope is affected badly by being plugged into the same mains electrical socket as the computer, it shows wave form distortion that is not present if plugged in somewhere else, or if the computer if off at the time.
Last edited:
Input cut and link
Here's an extract from the lme49830 schematic showing where the Signal Return/Ground (SG) should be cut to separate it from the Power ground.
A new link needs to be added to connect R46 back into the Signal Return.
Before you cut anything check using an audible continuity meter which side of R45 connects to Signal Return and which to Power Ground. Check R46 as well. Once you have identified the two sides add a thin copper strand to make the connection. And cut the trace.
In addition there is no RF attenuation at the input.
I added an 805 NPO ceramic on top of R46, But I did not have the correct value, so I added another NPO on top of that to give a total Crf of 430pF
This combined with the 1k2 of R44 gives an F-3dB of 300kHz.
As usual I added a 47pF NPO ceramic across the input pins of the socket as it enters the enclosure.
The input comes in with 2M2||47pF shunting to SG, then passes through 2u2F MKT (C78) through 1k2 (R44) then shunts via 30k (R46) to Signal Return to go back to the input socket. I would have liked to use 3u3F for C78 but the 0.2" pin pitch and space on the tiny PCB prevented anything bigger and prevented using an MKP.
The NFB upper leg is 30k (R39), lower leg is 1k2 (R43) + 330uF electro (C76) for a gain of 26times (+ 28.3dB)
Here's an extract from the lme49830 schematic showing where the Signal Return/Ground (SG) should be cut to separate it from the Power ground.
A new link needs to be added to connect R46 back into the Signal Return.
Before you cut anything check using an audible continuity meter which side of R45 connects to Signal Return and which to Power Ground. Check R46 as well. Once you have identified the two sides add a thin copper strand to make the connection. And cut the trace.
In addition there is no RF attenuation at the input.
I added an 805 NPO ceramic on top of R46, But I did not have the correct value, so I added another NPO on top of that to give a total Crf of 430pF
This combined with the 1k2 of R44 gives an F-3dB of 300kHz.
As usual I added a 47pF NPO ceramic across the input pins of the socket as it enters the enclosure.
The input comes in with 2M2||47pF shunting to SG, then passes through 2u2F MKT (C78) through 1k2 (R44) then shunts via 30k (R46) to Signal Return to go back to the input socket. I would have liked to use 3u3F for C78 but the 0.2" pin pitch and space on the tiny PCB prevented anything bigger and prevented using an MKP.
The NFB upper leg is 30k (R39), lower leg is 1k2 (R43) + 330uF electro (C76) for a gain of 26times (+ 28.3dB)
Attachments
Last edited:
Another things, I'm thinking about...
Which influence to sound does the quality of the Zobel capacitor have?
If anybody is going to redesign the LME49830 amp... Might make sense to place the capacitors on the sites of the LME49830, so the Mosfet can be placed directly behind the LME49830, which means only few mm traces between chip and Mosfet. This could also make it easier to connect the power connectors on the pcb's front side. The vertical mounted Mosfets are also having directly contact to the radiator, and that makes the pcb very compact. Only an idea
Which influence to sound does the quality of the Zobel capacitor have?
If anybody is going to redesign the LME49830 amp... Might make sense to place the capacitors on the sites of the LME49830, so the Mosfet can be placed directly behind the LME49830, which means only few mm traces between chip and Mosfet. This could also make it easier to connect the power connectors on the pcb's front side. The vertical mounted Mosfets are also having directly contact to the radiator, and that makes the pcb very compact. Only an idea
The typical 8ohms rated amplifier output Zobel is 10r+100nF.
That has an F-3dB @ 160kHz
Normally we see this roll-off from 100kHz to 300kHz.
That puts it outside the audio band. It is inaudible in it's own right.
But where it prevents amplifier misbehaviour then one will hear a difference between a properly behaving amplifier with the correct output Zobel compared to a misbehaving amplifier without the required output Zobel.
One should not "hear" the output Zobel.
A further thought:
if the amplifier still has a low output impedance where the output Zobel is starting to have a significant loading effect, then the amplifier's output is NOT rolled off even though the Zobel is starting to draw current.
The intention of the output Zobel is to prevent the amplifier seeing an open circuit load at VHF and thus becoming less stable if it's gain tries to change.
That has an F-3dB @ 160kHz
Normally we see this roll-off from 100kHz to 300kHz.
That puts it outside the audio band. It is inaudible in it's own right.
But where it prevents amplifier misbehaviour then one will hear a difference between a properly behaving amplifier with the correct output Zobel compared to a misbehaving amplifier without the required output Zobel.
One should not "hear" the output Zobel.
A further thought:
if the amplifier still has a low output impedance where the output Zobel is starting to have a significant loading effect, then the amplifier's output is NOT rolled off even though the Zobel is starting to draw current.
The intention of the output Zobel is to prevent the amplifier seeing an open circuit load at VHF and thus becoming less stable if it's gain tries to change.
Last edited:
After installing the amps I recognized further problems, which I recorded on following picture:
As always all amps are affected. The input signal is ok, fully synchronous.I'm getting annoyed... to be honest. Hope, anybody can help!
An externally hosted image should be here but it was not working when we last tested it.
As always all amps are affected. The input signal is ok, fully synchronous.I'm getting annoyed... to be honest. Hope, anybody can help!
Last edited: