I spent a few minutes over coffee this morning and put the transistor matcher into Eagle and laid out a board. These are embedded below.
Have a look at these (Chris too), and if they seem OK, then I'll post the gerber files. You can send them to almost any PCB fab house and get a board made. I use Sierra Circuits "No-Touch", which, for a board like this in qty 1 will run about $50-60.
I put the holes in the corners so you can put standoffs there for feet.
If folks want, I can make a run of these and mail them out. Sierra quotes $42 each plus postage (which is about $17..they only do Fed Ex or UPS), on a 10 day turnaround. 4 days increases the price to $47). So, if I include $6 for priority mail, each board would be about $53.
I separated the test transistors so it is easier to put a small cover over them. I found, as Chris had noted, that any air blowing over them changes the bias slightly. I used a small plastic box to isolate them.
Have a look at these (Chris too), and if they seem OK, then I'll post the gerber files. You can send them to almost any PCB fab house and get a board made. I use Sierra Circuits "No-Touch", which, for a board like this in qty 1 will run about $50-60.
I put the holes in the corners so you can put standoffs there for feet.
If folks want, I can make a run of these and mail them out. Sierra quotes $42 each plus postage (which is about $17..they only do Fed Ex or UPS), on a 10 day turnaround. 4 days increases the price to $47). So, if I include $6 for priority mail, each board would be about $53.
I separated the test transistors so it is easier to put a small cover over them. I found, as Chris had noted, that any air blowing over them changes the bias slightly. I used a small plastic box to isolate them.
Last edited:
Hi maceoc3,
You have a newer release. I haven't done anything in a while now.
Hi Scott,
It looks okay, but missing some features. I had the test positions close to each other so they would be physically touching. You might want to bring them back together. Also, my latest version uses DIP switches instead of the jumpers. It should be easy enough to do that swap.
The main problem is that the LED should be in the hole for the TO-126 transistors. You want the transistor to warm the LED so that it compensates the current. A 3mm LED fits nicely in that hole. I used the normal efficiency RED LEDs. You might want to start there, or maybe experiment with current stability and different colours of LEDs.
Last issue. Test points, you need to take measurements for balance if nothing else. For the collector (drain) 100 R resistors, I used 0.1% resistors from Digikey. I did the same for the base (grid) resistors so that you can measure other parameters and calculate the actual gain at a certain current level. I have calculated beta for some transistors at differing tail currents. Why? 'Cause I could and I was curious. The test points are what those wires poking up are all about. I used real test points on the later one.
Power. I use a bipolar bench supply and clip leads, but a plug might be better once you have used it a bit. It's just safer that way.
-Chris
You have a newer release. I haven't done anything in a while now.
Hi Scott,
It looks okay, but missing some features. I had the test positions close to each other so they would be physically touching. You might want to bring them back together. Also, my latest version uses DIP switches instead of the jumpers. It should be easy enough to do that swap.
The main problem is that the LED should be in the hole for the TO-126 transistors. You want the transistor to warm the LED so that it compensates the current. A 3mm LED fits nicely in that hole. I used the normal efficiency RED LEDs. You might want to start there, or maybe experiment with current stability and different colours of LEDs.
Last issue. Test points, you need to take measurements for balance if nothing else. For the collector (drain) 100 R resistors, I used 0.1% resistors from Digikey. I did the same for the base (grid) resistors so that you can measure other parameters and calculate the actual gain at a certain current level. I have calculated beta for some transistors at differing tail currents. Why? 'Cause I could and I was curious. The test points are what those wires poking up are all about. I used real test points on the later one.
Power. I use a bipolar bench supply and clip leads, but a plug might be better once you have used it a bit. It's just safer that way.
-Chris
Yeah, I noticed the LEDs in the T0126 transistors. I'll have to create a special package for that in Eagle.
Dip Switches are easy peasy.
I'll need to research available test points...
Thanks!
Scott
Dip Switches are easy peasy.
I'll need to research available test points...
Thanks!
Scott
Hi Scott,
I'm all for you doing this. Once you're done the layout, would you consider sharing it with DIYAudio, the actual website? I have been trying to complete this for a while into a package they can do something with. I have to say that it should have been available in the store already, but I haven't kept my end up.
With DipTrace, all I did was measure out how far the hole was and stuck the pads for the LED in there. It didn't complain and I made one using the toner transfer method. I wish I could find those files, I would have given those to everyone instead.
-Chris
I'm all for you doing this. Once you're done the layout, would you consider sharing it with DIYAudio, the actual website? I have been trying to complete this for a while into a package they can do something with. I have to say that it should have been available in the store already, but I haven't kept my end up.
With DipTrace, all I did was measure out how far the hole was and stuck the pads for the LED in there. It didn't complain and I made one using the toner transfer method. I wish I could find those files, I would have given those to everyone instead.
-Chris
Yeah sure.
I'll post the eagle files (schematic and board), and the Gerber and drill files for folks who want a professionally fabbed board.
S
I'll post the eagle files (schematic and board), and the Gerber and drill files for folks who want a professionally fabbed board.
S
Hi Scott,
Give yourself credit on the PCB. You did the work.
The project file called it the βalancer using a beta symbol for the "B". I know, corny.
-Chris
Give yourself credit on the PCB. You did the work.
The project file called it the βalancer using a beta symbol for the "B". I know, corny.
-Chris
OK, I redid the matcher with DIP switches, test points, and the LEDs embedded in the TO126 transistors.
I'll double check this again, and then post the Gerber and drill files.
I'll double check this again, and then post the Gerber and drill files.
OK, got it all put back together, and turned it on. Still oscillates. Changed the feedback cap from 82 pF to 33 pF, and it works fine up to some moderate signal level, wherein I get a little oscillation on the positive peak of the output waveform. I can eliminate this by increasing the bypass capacitors C103 and C104 (Which reduces the high frequency gain on the input diff amp stage). I may also be able to do it by increasing C121.
this oscillation is way up in the 3-5 MHz region.
this oscillation is way up in the 3-5 MHz region.
Hi Scott,
How low can you go with the feedback cap? Many amplifiers are fine with 10 pF, or maybe you jumped over the "happy spot".
-Chris
How low can you go with the feedback cap? Many amplifiers are fine with 10 pF, or maybe you jumped over the "happy spot".
-Chris
Funny you should ask. I just ordered some WIMA polypropylene 10 pF caps..
These WIMA caps are really nice (square blocks), and fit the board perfectly.
Last edited:
Hi Scott,
I keep a bunch of low value pF capacitors on hand for this reason. You might be good at 22pF, depending on how high the stray capacitance might be. Your choice of output transistor will also affect this.
-Chris
I keep a bunch of low value pF capacitors on hand for this reason. You might be good at 22pF, depending on how high the stray capacitance might be. Your choice of output transistor will also affect this.
-Chris
Hi Scott,
Yes, I know those work. But they work on the original boards. I'm not surprised that you are having stability problems with a new PCB layout. Keep at it and you will get this licked. It just isn't always easy to do.
Try the 10 pF first before ordering a bunch of stuff. I usually work down slowly through the values skipping every other value until I get close, then I try every one. It's a slow process. There are times when you have to adjust the capacitors across other transistors, like driver transistors as well. If you look at enough schematics, you can see which ones were difficult to stabilize. Since I've never had this problem with these amps, I'll have to look at the schematic with oscillation in mind. Often the answer actually lies in the PCB layout in a case like this.
-Chris
Yes, I know those work. But they work on the original boards. I'm not surprised that you are having stability problems with a new PCB layout. Keep at it and you will get this licked. It just isn't always easy to do.
Try the 10 pF first before ordering a bunch of stuff. I usually work down slowly through the values skipping every other value until I get close, then I try every one. It's a slow process. There are times when you have to adjust the capacitors across other transistors, like driver transistors as well. If you look at enough schematics, you can see which ones were difficult to stabilize. Since I've never had this problem with these amps, I'll have to look at the schematic with oscillation in mind. Often the answer actually lies in the PCB layout in a case like this.
-Chris
I think this is probably related to the use of a two sided board. This means that there is some opportunity for interaction between traces that on the original board were separated.
I developed a new board layout to fix the various issues with the first cut at the board. I included a ground plane between the two layers, so that may reduce these issues somewhat.
I developed a new board layout to fix the various issues with the first cut at the board. I included a ground plane between the two layers, so that may reduce these issues somewhat.
Very interesting. Lying in bed this mooring early, I was contemplating my oscillation issue. I am pretty sure adding a ground plane between the two layers will solve a lot of issues, but I thought I'd go have a look at the current layout to see if I could identify any issues.
Sure enough, there it was staring me in the face..
Here is a shot of the feedback section of the board, Blue traces are on the bottom,and red are on the top.
Q101, on the right is one of the input Darlingtons, and Q105 is the matching Darlington that forms one of the differential pairs for this amp. Q102 and Q105 are the other diff pair. The feedback signal comes in via the capacitor C105 (You can see the edge of terminal point 7 on the board just above the cap - this is where the feedback wire attaches).
The problem areas are circled. The lit up blue trace is the audio input signal that goes to the base of Q101, and Q102. The feedback signal goes through C105 and into the base of Q105, and Q106 (which is the opposite polarity input for the diff pair, so, since this is a "differential amplifier" the feedback signal is subtracted from the audio input signal..thereby creating "negative feedback"... (I realize this is obvious to many of you, but I wanted to inform some of the readers who might not be quite as familiar with these types of circuits).
You can see in the left side yellow circle that the input signal line actually follows the feedback line, on the other side of the board for some distance. This effectively is coupling the feedback signal to the audio input line, thereby creating "positive feedback" and thus a propensity for oscillation.
It is instructive that the oscillation only occurs at certain signal levels. Presumably the coupling fields are small at small signal levels, and as the output signal grows (remember the output signal is MUCH larger than the input because of the amplifier gain), it eventually couples enough to trigger this oscillation.
Reducing C105 will substantially reduce the level of the feedback signal at this overlap point, but this is also reducing the amount of feedback, and thus probably also increasing the gain and the overall distortion of the amplifier.
There is another issue which is highlighted in the right hand yellow circle..This is not quite as bad, but basically the highlighted blue trace circles around the input transistor Q105, thereby adding more coupling opportunities. At the frequencies where this oscillation is occurring, (3-5 MHz) the bends and kinks in the blue trace will create additional electric fields (known as "fringing capacitance") which will couple to the collector of Q105, and the base side of C105...just adding more of this overall positive feedback coupling.
It is important to note that you get SOME coupling on retraces that cross at right angles, but the overlap region is very small, and the propagation directions are at 90 degrees, so there is relatively little coupling of the actual fields between the two lines. In the car where the traces run parallel, this is exactly NOT the case, and as noted above, in situations where the traces are circulating around each other (the right yellow circle) the effect is similar to parallel traces.
I'll re-layout this board to re-route these lines.. and also keep the new buried ground plane...
Cheers,
Scott
Sure enough, there it was staring me in the face..
Here is a shot of the feedback section of the board, Blue traces are on the bottom,and red are on the top.
Q101, on the right is one of the input Darlingtons, and Q105 is the matching Darlington that forms one of the differential pairs for this amp. Q102 and Q105 are the other diff pair. The feedback signal comes in via the capacitor C105 (You can see the edge of terminal point 7 on the board just above the cap - this is where the feedback wire attaches).
The problem areas are circled. The lit up blue trace is the audio input signal that goes to the base of Q101, and Q102. The feedback signal goes through C105 and into the base of Q105, and Q106 (which is the opposite polarity input for the diff pair, so, since this is a "differential amplifier" the feedback signal is subtracted from the audio input signal..thereby creating "negative feedback"... (I realize this is obvious to many of you, but I wanted to inform some of the readers who might not be quite as familiar with these types of circuits).
You can see in the left side yellow circle that the input signal line actually follows the feedback line, on the other side of the board for some distance. This effectively is coupling the feedback signal to the audio input line, thereby creating "positive feedback" and thus a propensity for oscillation.
It is instructive that the oscillation only occurs at certain signal levels. Presumably the coupling fields are small at small signal levels, and as the output signal grows (remember the output signal is MUCH larger than the input because of the amplifier gain), it eventually couples enough to trigger this oscillation.
Reducing C105 will substantially reduce the level of the feedback signal at this overlap point, but this is also reducing the amount of feedback, and thus probably also increasing the gain and the overall distortion of the amplifier.
There is another issue which is highlighted in the right hand yellow circle..This is not quite as bad, but basically the highlighted blue trace circles around the input transistor Q105, thereby adding more coupling opportunities. At the frequencies where this oscillation is occurring, (3-5 MHz) the bends and kinks in the blue trace will create additional electric fields (known as "fringing capacitance") which will couple to the collector of Q105, and the base side of C105...just adding more of this overall positive feedback coupling.
It is important to note that you get SOME coupling on retraces that cross at right angles, but the overlap region is very small, and the propagation directions are at 90 degrees, so there is relatively little coupling of the actual fields between the two lines. In the car where the traces run parallel, this is exactly NOT the case, and as noted above, in situations where the traces are circulating around each other (the right yellow circle) the effect is similar to parallel traces.
I'll re-layout this board to re-route these lines.. and also keep the new buried ground plane...
Cheers,
Scott
Hi Scott,
I'm sorry you had to respin the PCB. I do worry about the ground plane adding some capacitance in places that might cause enough phase shift to create different problems. Maybe increase the distance between the feedback trace and other things. I'm thinking of the high impedance section and leakage currents. Also, jumpers are not always bad things when you are laying out an amplifier if they eliminate wandering traces. Just something to keep in mind.
-Chris
I'm sorry you had to respin the PCB. I do worry about the ground plane adding some capacitance in places that might cause enough phase shift to create different problems. Maybe increase the distance between the feedback trace and other things. I'm thinking of the high impedance section and leakage currents. Also, jumpers are not always bad things when you are laying out an amplifier if they eliminate wandering traces. Just something to keep in mind.
-Chris
Yeah, I was thinking about that too. I was able to re-layout the board in the feedback area so that the top and bottom traces are only crossing at 90 degrees, and there are no complicated wandering traces, so I think I'll run with that.
The reduced feedback cap helped a lot, but I think that's really just a band-aid fix. Better to figure out why the oscillation is occurring and fix it at the root cause.
Here are the two different layouts side by side:
New one:
Old one:
The reduced feedback cap helped a lot, but I think that's really just a band-aid fix. Better to figure out why the oscillation is occurring and fix it at the root cause.
Here are the two different layouts side by side:
New one:
Old one:
Last edited:
Interestingly, I pulled up the long wandering input signal trace, and put a jumper across the board (up off the board). Made no difference at all.
Removing the 33 pF cap entirely helped a little.
I noticed, however, that the ground signals on the output boards have little bursts of oscillation, where the ground at the supply caps does not..not a hint of it..
I am wondering if the bypass caps on the output boards are perhaps showing some loss at high frequencies. If I put a .01 uF cap across the 47 uF electrolytic cap on the negative side board, the negative peak oscillation goes away...so that's telling (I did the negative side since that's where most of the oscillation is...).
I have some new replacement caps for those boards, so I'll try that next.
Removing the 33 pF cap entirely helped a little.
I noticed, however, that the ground signals on the output boards have little bursts of oscillation, where the ground at the supply caps does not..not a hint of it..
I am wondering if the bypass caps on the output boards are perhaps showing some loss at high frequencies. If I put a .01 uF cap across the 47 uF electrolytic cap on the negative side board, the negative peak oscillation goes away...so that's telling (I did the negative side since that's where most of the oscillation is...).
I have some new replacement caps for those boards, so I'll try that next.
Hi Scott,
That looks better in that area. Let's see how that goes. I bet you feel better about it.
Normally you don't need to bypass those supply capacitors, but it never hurts to do so. If you do need to bypass them, I'd be looking for another reason.
Question for you. Did you ground the heat sinks during your tests? Failure to do that will often allow oscillations to occur.
-Chris
That looks better in that area. Let's see how that goes. I bet you feel better about it.
Normally you don't need to bypass those supply capacitors, but it never hurts to do so. If you do need to bypass them, I'd be looking for another reason.
Question for you. Did you ground the heat sinks during your tests? Failure to do that will often allow oscillations to occur.
-Chris