Hi thoru
It's never easy to diagnose a problem "at a distance", but it sounds as though you might have some oscillation.
The two-pole Miller is not usually quite so stable as a one-pole, (it can be, but needs more careful deigsn) so you could check by removing the two-pole tie down resistor to make it a 1-pole and see if that helps.
Another possibility might be due to running the two current sources from one regulator transistor. I've not investigated this, so I cannot guarantee that this will work, but if your problem is due to interactions between the two CCS, you could try separating them. Instead of the current source feedback transistor, remove the transistor and use two diodes instead. Split the base to the CCS to the VAS from the CCS on the input stage by running a separate resistor to the filter capacitor and use a second pair of diodes on the other CCS. By avoiding an active feedback circuit in the CCS it MAY be more stable.
cheers
John
It's never easy to diagnose a problem "at a distance", but it sounds as though you might have some oscillation.
The two-pole Miller is not usually quite so stable as a one-pole, (it can be, but needs more careful deigsn) so you could check by removing the two-pole tie down resistor to make it a 1-pole and see if that helps.
Another possibility might be due to running the two current sources from one regulator transistor. I've not investigated this, so I cannot guarantee that this will work, but if your problem is due to interactions between the two CCS, you could try separating them. Instead of the current source feedback transistor, remove the transistor and use two diodes instead. Split the base to the CCS to the VAS from the CCS on the input stage by running a separate resistor to the filter capacitor and use a second pair of diodes on the other CCS. By avoiding an active feedback circuit in the CCS it MAY be more stable.
cheers
John
Hi,
Thanks for your ideas. I figured out that's it's not an oscillation in the VAS stage. I tried both your recommendations, but they changed nothing.
I measured the input signal and at the connector, it contains a lot of 100kHz noise. This is very effectivily filtered out by the input filter, but there remain very sharp spikes at a rate of 50Hz. They look like delta-pulses on my scope. I think the plops are now inserted by my lab power supply, which has current limiting and it probably sees the spikes as an overcurrent, so it tries to scale down the voltage and immediatly restores the voltage etc.
That probably leaves the question where the sharp 50Hz spikes are coming from and how to get them out. I tried placing a 10 and 150Ohm resistor between signalground and starground, in the hope it was a groundloop problem, but that didn't solve it. The source isn't grounded, so a groundloop can't hardly exist actually.
Any ideas?
Thanks,
Remco Poelstra
Thanks for your ideas. I figured out that's it's not an oscillation in the VAS stage. I tried both your recommendations, but they changed nothing.
I measured the input signal and at the connector, it contains a lot of 100kHz noise. This is very effectivily filtered out by the input filter, but there remain very sharp spikes at a rate of 50Hz. They look like delta-pulses on my scope. I think the plops are now inserted by my lab power supply, which has current limiting and it probably sees the spikes as an overcurrent, so it tries to scale down the voltage and immediatly restores the voltage etc.
That probably leaves the question where the sharp 50Hz spikes are coming from and how to get them out. I tried placing a 10 and 150Ohm resistor between signalground and starground, in the hope it was a groundloop problem, but that didn't solve it. The source isn't grounded, so a groundloop can't hardly exist actually.
Any ideas?
Thanks,
Remco Poelstra
Hi,
Yes, more or less they expose the same problem. I'm currently not using any preamp. I've tried a mixingconsole, CD player and function generator, all directly connected.
The function generator though, gives a much nicer hum, without the high frequency components of the other two. It's also a lot reduced in loudness.
Thanks,
Remco Poelstra
Yes, more or less they expose the same problem. I'm currently not using any preamp. I've tried a mixingconsole, CD player and function generator, all directly connected.
The function generator though, gives a much nicer hum, without the high frequency components of the other two. It's also a lot reduced in loudness.
Thanks,
Remco Poelstra
Hi,
if mixing console AND CD player AND function generator all produce the same or similar problems, then it is extremely unlikely that these sources are causing the problems you are experiencing.
You have also confirmed that low source impedance AND zero signal level is silent. This seems to eliminate the amplifier from some of the causes, but not all!
It appears that there may be a compatibility issue between any signal/amplifier and the temporary PSU.
Sort the PSU and listen for the problem.
if mixing console AND CD player AND function generator all produce the same or similar problems, then it is extremely unlikely that these sources are causing the problems you are experiencing.
You have also confirmed that low source impedance AND zero signal level is silent. This seems to eliminate the amplifier from some of the causes, but not all!
It appears that there may be a compatibility issue between any signal/amplifier and the temporary PSU.
Sort the PSU and listen for the problem.
Thoru said:
you have any other supply you can try?
you might be right it can be your Lab Supply vs. Amp interaction
that is the cause.
--------------------------
What you should check, my ideas:
1. Any of those changes you have done - to 'Reduce Hum'
Slone amplifier is a system that fits together.
When modifying, there may be thing in this system
that does not match the original set up.
2. Check components in power supply rail,
from the input stage and backwards to your power source.
Sometimes such a simple thing as one Electrolytic bypass capacitor
connected with wrong polarity, + and - ... can cause strange effects.
If they for some reason do not EXPLODE
3. Build a power supply for your amp and try if this works.
then we know it has to be Lab Supply mismatch with your amp.
----------------------------
Regards
lineup
This may not be the problem here but once I had perhaps an analogous experience. The DC input impedance of the Sloane is 10k-12K. This is fine for nearly any audio preamp. However, I discovered there are some devices that have trouble with this low an impedance and remember it declines as frequency rises. Two devices I had trouble with were a computer sound card and a battery operated sine generator. It was very frustrating until I eventually realized that the horrible THD+N measurements were the signal source.
I put a pot in between and the problems went away when they were set so that the effective input impedance was about 30k. This isn't the ideal solution but it points to building simple unity gain opamp buffer to insert between signal source and amp. Usually this is one of the basic things preamps do as a matter of course. The other thing this suggests is that the 11.4 could me modified for a higher input impedance, say 47k (a traditional value) -- this involves sustituing at least three resistors and a cap or two. I've only done this in LTSpice so I'm not going to provide a "how to" as it may not work out in real life with out some experimentation.
I put a pot in between and the problems went away when they were set so that the effective input impedance was about 30k. This isn't the ideal solution but it points to building simple unity gain opamp buffer to insert between signal source and amp. Usually this is one of the basic things preamps do as a matter of course. The other thing this suggests is that the 11.4 could me modified for a higher input impedance, say 47k (a traditional value) -- this involves sustituing at least three resistors and a cap or two. I've only done this in LTSpice so I'm not going to provide a "how to" as it may not work out in real life with out some experimentation.
Hi,
I've finally build a nice PSU for the amp. It's a basic unregulated design, just like Slone suggests. This solves the problem of added noise when I connect the mixing console or CD player. When I connect the function generator, I still have some noise/hum, but that could also be due to a bad design of that devices (it was a rather cheap one). I even dared to connect the CD player to the console and it was a pleasure to here the amp play.
But what I do have now is a constant amount of hum. It's has a very soft character and it's also there when I short the input to signal ground, so I suspect it's coming from the PSU. It's immediatly gone when I pull the AC plug.
Concerning the anti-hum modifications I did to the amp:
I seperated the ground track in two: one for the decoupling caps and one for the signal, feedback and Q5's ground.
I inserted a 47Ohm resistor in both supply lines between R20, R25 and R22,R28. This to form a lowpass filter with the decoupling caps.
So this new PSU helped me in one way, but degraded performance in another. Anyone any idea on how I can get rid of that final bit of hum?
Thanks in advance,
Remco Poelstra
I've finally build a nice PSU for the amp. It's a basic unregulated design, just like Slone suggests. This solves the problem of added noise when I connect the mixing console or CD player. When I connect the function generator, I still have some noise/hum, but that could also be due to a bad design of that devices (it was a rather cheap one). I even dared to connect the CD player to the console and it was a pleasure to here the amp play
.But what I do have now is a constant amount of hum. It's has a very soft character and it's also there when I short the input to signal ground, so I suspect it's coming from the PSU. It's immediatly gone when I pull the AC plug.
Concerning the anti-hum modifications I did to the amp:
I seperated the ground track in two: one for the decoupling caps and one for the signal, feedback and Q5's ground.
I inserted a 47Ohm resistor in both supply lines between R20, R25 and R22,R28. This to form a lowpass filter with the decoupling caps.
So this new PSU helped me in one way, but degraded performance in another. Anyone any idea on how I can get rid of that final bit of hum?
Thanks in advance,
Remco Poelstra
Hi,
Currently it's a mono design. In the future I want it to be stereo, but mono seems allready difficult enough for now
I think it's fine to split off this thread (I started it as a seperate one originally), but I do think it's the amp which is wrong here now. My PSU is as simple as it can get and it's the recommended design by both Slone and Self, and my wiring is also as it's suggested by these two. It actaully turned out that Slone's PCB is not design according to his own ideas.
Thanks,
Remco Poelstra
Currently it's a mono design. In the future I want it to be stereo, but mono seems allready difficult enough for now
I think it's fine to split off this thread (I started it as a seperate one originally), but I do think it's the amp which is wrong here now. My PSU is as simple as it can get and it's the recommended design by both Slone and Self, and my wiring is also as it's suggested by these two. It actaully turned out that Slone's PCB is not design according to his own ideas.
Thanks,
Remco Poelstra
Hello tcpip,
let me chime in late here due to a couple of layout and 'manufacturing' problems I see.
I don't know Slone's book, but I have read Doug Self's.
He writes a lot more on his findings about distortion sources in amplifers than on his website. Especially does he sort them in the order of relevance on the sound and noise/distortion performance. Ground scheme and inductive/capacitive coupling problems come first in the 'blameless' layout, stage topology is second, exact component choice is third - or so to speak.
I understand that you plan on adjusting the board layout after you got all the sizey components. That's always a great idea. However, you will be surprized how many adjustments you will likely have to make. From my experience with eagle, let me collect some thoughts about your most recent layout. Those come from experience, so take them as an inspiration, although I don't earn my money with board layouts.
Here's how you should start: determine your board manufacturing process, e.g. the precision you can achieve. Adjust Eagle's design rule set accordingly. If unsure, choose a manufacturer's design rule file, such as the EAGLE DRU file on this page and modify it to your needs (lots of good info there, too).
You can then rely on Eagle's electrical (in the schematic) and design (in the board editor) rule checks, which are mandatory for a board to be considered ready. If the design rules are valid and Eagle sees design errors, then there are design errors!
Name and classify your net list. Give important signals a self-explaining name, such as V+, V-, PGND, SGND, IN, OUT, NFB, etc. You probably did that. Then classify them. That you probably didn't. Read up on it in the Eagle help and tutorial. Most important classes will be power, grounds, small signal, large signal, etc. This will automagically make Eagle respect wire widths, creepages between copper of different classes, and such. You can then relax from thinking through the safety and functionality issues everytime you change something; just do a DRC. This would also help you get rid of the copper pour under F1, which is essentially a time bomb - while still ensuring sufficiently wide traces (as you define in the classes).
I also recommend determining the power dissipation in every single component, especially the resistors. Most of them could be 0204, bent narrowly (or with their legs as standoffs as in good old amplifiers), as you say you need the small board size. Some others probably need to be bigger than your 0207, even in the small signal stage (e.g. current source, perhaps, didn't check myself right now).
The input capacitor package C1 you chose looks way too large, as do the emitter resistors R18/R24. A pin spacing of 22.5mm for a 10uF/63V or smaller is sufficient. Try to find MPC packaged 'metal tape' resistors. Smaller footprint, higher quality. Eagle has an 'mpc-71' package in it's library 'resistor-power'. But even the wire-wound resistors you chose are available in way smaller packages for 5W power dissipation, i.e. 0617 or similar in the 'rcl' library.
Your mounting holes are spaced reasonably, the fuses and C1 are not. You can't mount the power rails safely on the very board edge - it's just too unsafe. Keep at least 2.5 to 5 mm between copper and board edge (e.g. by aligning the component so that the placement symbol barely touches the board edge). What's R38 doing underneath C1?
The heatsinks you chose are free-mounted, non-standing. That's not good. They are very small, yet very disadvantageous to mount. I recommend suspended sinks, e.g. with mounting holes on the board. The schematic can be laid out with all relevant transistors aligned on one or two small common heatsinks, which improves real estate occupation and temperature stability.
Think about your connectors. You chose a fast-on standoff for speaker output, but everything else shall be soldered? I can recommend the same type of connector for every power connection. 'ST6,3' from 'con-rib' is good for this, they are also available as 90degree horizontal standoffs. Soldering will work (though I don't recommend it), but your holes look too narrow from over here for power wiring.
As pinkmouse already recommended early, group the components by functional blocks. You've done this in the big picture, but the small signal stage and drivers are spread over more than half of the whole board area. The problem here is radiation and coupling sensitivity (which you can also read about on Self's site). Make this as small as possible and keep a little luxurious real estate to high current traces. As others mention, yours intertwine just to much. This blameless would be to blame.
How I would solve it? Using the autorouter!
Just 'ripup' any wires you like to improve upon and click the 'Select' button to choose the wires that should be tried. The autorouter can't route your board, but it can do the work of finding possible routes and showing up bottlenecks and alternatives. Don't see the autorouter as a defunct magic board maker (that always fails), it works as a tool to look into specific trace route problems. Because it respects our design rules! If it doesn't route the way you want, then it would violate your design rules to do so (creepage, clearance, trace width, net class). It is good practice to shuffle components and reiterate with the autorouter until everything fits, only to then rip everything up one by one and finish the traces by hand. Correct and beautiful.
Okay, enough for now. I hope this clarifies, rather than to set you off.
Cheers,
Sebastian.
let me chime in late here due to a couple of layout and 'manufacturing' problems I see.
I don't know Slone's book, but I have read Doug Self's.
He writes a lot more on his findings about distortion sources in amplifers than on his website. Especially does he sort them in the order of relevance on the sound and noise/distortion performance. Ground scheme and inductive/capacitive coupling problems come first in the 'blameless' layout, stage topology is second, exact component choice is third - or so to speak.
I understand that you plan on adjusting the board layout after you got all the sizey components. That's always a great idea. However, you will be surprized how many adjustments you will likely have to make. From my experience with eagle, let me collect some thoughts about your most recent layout. Those come from experience, so take them as an inspiration, although I don't earn my money with board layouts.
Here's how you should start: determine your board manufacturing process, e.g. the precision you can achieve. Adjust Eagle's design rule set accordingly. If unsure, choose a manufacturer's design rule file, such as the EAGLE DRU file on this page and modify it to your needs (lots of good info there, too).
You can then rely on Eagle's electrical (in the schematic) and design (in the board editor) rule checks, which are mandatory for a board to be considered ready. If the design rules are valid and Eagle sees design errors, then there are design errors!
Name and classify your net list. Give important signals a self-explaining name, such as V+, V-, PGND, SGND, IN, OUT, NFB, etc. You probably did that. Then classify them. That you probably didn't. Read up on it in the Eagle help and tutorial. Most important classes will be power, grounds, small signal, large signal, etc. This will automagically make Eagle respect wire widths, creepages between copper of different classes, and such. You can then relax from thinking through the safety and functionality issues everytime you change something; just do a DRC. This would also help you get rid of the copper pour under F1, which is essentially a time bomb - while still ensuring sufficiently wide traces (as you define in the classes).
I also recommend determining the power dissipation in every single component, especially the resistors. Most of them could be 0204, bent narrowly (or with their legs as standoffs as in good old amplifiers), as you say you need the small board size. Some others probably need to be bigger than your 0207, even in the small signal stage (e.g. current source, perhaps, didn't check myself right now).
The input capacitor package C1 you chose looks way too large, as do the emitter resistors R18/R24. A pin spacing of 22.5mm for a 10uF/63V or smaller is sufficient. Try to find MPC packaged 'metal tape' resistors. Smaller footprint, higher quality. Eagle has an 'mpc-71' package in it's library 'resistor-power'. But even the wire-wound resistors you chose are available in way smaller packages for 5W power dissipation, i.e. 0617 or similar in the 'rcl' library.
Your mounting holes are spaced reasonably, the fuses and C1 are not. You can't mount the power rails safely on the very board edge - it's just too unsafe. Keep at least 2.5 to 5 mm between copper and board edge (e.g. by aligning the component so that the placement symbol barely touches the board edge). What's R38 doing underneath C1?
The heatsinks you chose are free-mounted, non-standing. That's not good. They are very small, yet very disadvantageous to mount. I recommend suspended sinks, e.g. with mounting holes on the board. The schematic can be laid out with all relevant transistors aligned on one or two small common heatsinks, which improves real estate occupation and temperature stability.
Think about your connectors. You chose a fast-on standoff for speaker output, but everything else shall be soldered? I can recommend the same type of connector for every power connection. 'ST6,3' from 'con-rib' is good for this, they are also available as 90degree horizontal standoffs. Soldering will work (though I don't recommend it), but your holes look too narrow from over here for power wiring.
As pinkmouse already recommended early, group the components by functional blocks. You've done this in the big picture, but the small signal stage and drivers are spread over more than half of the whole board area. The problem here is radiation and coupling sensitivity (which you can also read about on Self's site). Make this as small as possible and keep a little luxurious real estate to high current traces. As others mention, yours intertwine just to much. This blameless would be to blame.
How I would solve it? Using the autorouter!
Just 'ripup' any wires you like to improve upon and click the 'Select' button to choose the wires that should be tried. The autorouter can't route your board, but it can do the work of finding possible routes and showing up bottlenecks and alternatives. Don't see the autorouter as a defunct magic board maker (that always fails), it works as a tool to look into specific trace route problems. Because it respects our design rules! If it doesn't route the way you want, then it would violate your design rules to do so (creepage, clearance, trace width, net class). It is good practice to shuffle components and reiterate with the autorouter until everything fits, only to then rip everything up one by one and finish the traces by hand. Correct and beautiful.
Okay, enough for now. I hope this clarifies, rather than to set you off.
Cheers,
Sebastian.
Thanks Sek,
I might try and get Eagle to work for me now. I have given up on it once already.
How about a Wiki on Eagle:-
getting set up.
creating schematic
altering existing
creating first layout.
altering layout.
etc.
I'm sure yoy could think of other chapters.
Must be loads of contributors to write a chapter each. Then the little tips can be tagged on.
I might try and get Eagle to work for me now. I have given up on it once already.
How about a Wiki on Eagle:-
getting set up.
creating schematic
altering existing
creating first layout.
altering layout.
etc.
I'm sure yoy could think of other chapters.
Must be loads of contributors to write a chapter each. Then the little tips can be tagged on.
Hi,
I've finally found the hum problem. While modifying the PCB for the lowpass filter on the supply lines, I forgot a trace. So Q4 was fed with an unfiltered supply, which caused the hum. Not that all problems are solved now, I've a nice supply-dependend 3,5MHz oscillation now. So it's probably a nice time to re-read the book regarding oscillations.
Thanks for all support,
Remco Poelstra
I've finally found the hum problem. While modifying the PCB for the lowpass filter on the supply lines, I forgot a trace. So Q4 was fed with an unfiltered supply, which caused the hum. Not that all problems are solved now, I've a nice supply-dependend 3,5MHz oscillation now. So it's probably a nice time to re-read the book regarding oscillations
.Thanks for all support,
Remco Poelstra
Thanks for the patient and detailed set of inputs. As with any other post which has lots of substance, it'll take me some time to absorb all of it.sek said:
You're absolutely right about me not using net classes. I have not become that proficient in the usage of these ECAD tools. I cannot respond to all your points, but a few I will respond to. Since your comments were about a version of PCB layout which I had posted on 12 Jan, more than a month has passed, and I had been making small changes since then. I've uploaded the latest v9 of the PCB layout below. Your larger issues have not been addressed even in this version, but some of the smaller ones have been.
I've had long chats with the guy who will make my boards. He's made boards for me in the past. My minimum track thickness and inter-track clearance settings in Eagle are based on what he's said he's comfortable with... in fact, I've kept a safety margin beyond his limits.
I was using the DRC facility on the board quite regularly, but didn't know about the electrical rules check in the schematic. I'll try that and see.
I am curious to know why people are uneasy about copper pours. I have very little knowledge in this area, so can you please explain in a bit more detail what problems occur due to those blocks?
I'm using 0207 as the baseline resistor shape, actually.
I'll check those.
The C1 is for a 31.5mm MKP cap with 27.5mm lead pitch, which is what I seem to find for 10uF MKP/MKT caps from some brands. The emitter resistors are standard 5W power resistors, and their body outline and lead pitch match exactly with the inexpensive resistors I've bought before. The 10W resistors I have bought are even larger.
Give me some part numbers? I'd love to have a smaller C1.
Where do I find them?
I don't get them here. I'll have to import them probably. Do you have any part numbers from Digikey or similar?
I'm hard pressed for space; I'll see what I can do. Point noted, and thanks.
Well, at a crunch, I was intending to fit the resistor below the board. But anyway it's history now... you'll see it's been moved in the new layout.What's R38 doing underneath C1?
I went to my local retailer, and found only these heatsinks of reasonable size. Any larger heatsinks were very large, and even those didn't have soldering points to mount on the board. I'll live with these for the time being, I think.
Not really. What wasn't obvious in the earlier board (but is obvious now) is that I deliberately chose the solder pad positions to allow 5mm lead-pitch Phoenix terminal blocks to be mounted. That way, if someone wants to solder, he can, and if I want to fit terminal blocks, I can.
Yes, I discovered these after I'd defined my own Faston parts for my library.
Actually, they're 1.3mm drill size, so they should be fine I thought.
Yes, I agree that there may be concern about this. But I've decided that the current layout is worth building one iteration with, just to let me learn what happens. This is work in progress, and I'm certainly going to keep working on it.
Does the free version of Eagle support autorouting?
Interesting. I'll try it someday, if I can buy a version of Eagle which has the autorouter.It is good practice to shuffle components and reiterate with the autorouter until everything fits, only to then rip everything up one by one and finish the traces by hand. Correct and beautiful.
I really appreciate your patience. I've learned a lot.Okay, enough for now. I hope this clarifies, rather than to set you off.
And this is the new layout:
An externally hosted image should be here but it was not working when we last tested it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.