I did ground them, although I am not sure how well. I found that attaching a ground to the negative heat sink actually reduced the oscillation (significantly) on the positive peak, but grounding the other heat sink didn't do much. I think the issue is related to this however. I may try putting the top cover on and see if that changes things.
However, the fact that bypassing the old electrolytics made such a difference tells me that they are not bypassing high frequencies well, and that is probably a core contributor to the problem. It will only take a few minutes to change them out...
However, the fact that bypassing the old electrolytics made such a difference tells me that they are not bypassing high frequencies well, and that is probably a core contributor to the problem. It will only take a few minutes to change them out...
Taking Chris' advice to heart, I did a new board layout.
This time I took care to, as much as possible put all of the components in the same places and orientations as the original. This is no exactly possible because I am using a more modern heat sink, but the result is pretty close. I also put in some polygons where the original had large areas of trace. The one thing I did slightly differently was to use top side copper in lieu of jumper wires.
I found a few spots where my layout was very slightly different (parts swapped in position and/or rotated).
The main thing I found, which MAY have something to do with my oscillation problem, is that the board and the schematic provided by Adcom are not exactly the same. The difference is subtle, but the compensation cap/resistor pairs on the board output C110/R137, and C111/R138 were swapped. On the schematic the resistor goes to the signal line and the cap goes to ground. On the board the cap is on the signal line and the resistor goes to ground.
If you think about this from a en electrical perspective, it should make no difference, but I suspect from a compensation perspective, in the board configuration, the cap will pass only high frequencies to the 10 ohm load to ground, while in the Adcom schematic configuration, signal line is separated from the frequency dependent element (the cap) by the resistor. As a result, I suspect the board configuration has a bit more high frequency suppression than the schematic.
Here is the new board layout.
This time I took care to, as much as possible put all of the components in the same places and orientations as the original. This is no exactly possible because I am using a more modern heat sink, but the result is pretty close. I also put in some polygons where the original had large areas of trace. The one thing I did slightly differently was to use top side copper in lieu of jumper wires.
I found a few spots where my layout was very slightly different (parts swapped in position and/or rotated).
The main thing I found, which MAY have something to do with my oscillation problem, is that the board and the schematic provided by Adcom are not exactly the same. The difference is subtle, but the compensation cap/resistor pairs on the board output C110/R137, and C111/R138 were swapped. On the schematic the resistor goes to the signal line and the cap goes to ground. On the board the cap is on the signal line and the resistor goes to ground.
If you think about this from a en electrical perspective, it should make no difference, but I suspect from a compensation perspective, in the board configuration, the cap will pass only high frequencies to the 10 ohm load to ground, while in the Adcom schematic configuration, signal line is separated from the frequency dependent element (the cap) by the resistor. As a result, I suspect the board configuration has a bit more high frequency suppression than the schematic.
Here is the new board layout.
Got Amp #4 working with the "new" board (not the re-layout, but the first one I made).
It definitely doesn't want any capacitance on the feedback line, and I reduced the capacitance on the two output compensation circuits (C111, and C112) from 220 pF to 100 pF. That fixed the oscillations. It works without any bypass caps on the output board (the ones I added yesterday).
I also found that I had laid out J105 and J104 with the wrong pinouts (I figured this out a while ago). They are numbered 132, instead of 123. I had swapped the wires around in the connectors to allow for this. What I hadn't realized is that I had also laid out J105 and J104 in opposite positions... So, when I plugged them in I had the bias temp compensators swapped. What was worse was that because of this, I had crossed up the wires when I swapped them around. I got that sorted out and the output waveform is perfect, right up to 350 watts.
I am going to fix the J104/J105 issue, and run wight he bard layout immediately above. It is pretty close to the original, and I know the circuit is now correct.
Cheers,
Scott
It definitely doesn't want any capacitance on the feedback line, and I reduced the capacitance on the two output compensation circuits (C111, and C112) from 220 pF to 100 pF. That fixed the oscillations. It works without any bypass caps on the output board (the ones I added yesterday).
I also found that I had laid out J105 and J104 with the wrong pinouts (I figured this out a while ago). They are numbered 132, instead of 123. I had swapped the wires around in the connectors to allow for this. What I hadn't realized is that I had also laid out J105 and J104 in opposite positions... So, when I plugged them in I had the bias temp compensators swapped. What was worse was that because of this, I had crossed up the wires when I swapped them around. I got that sorted out and the output waveform is perfect, right up to 350 watts.
I am going to fix the J104/J105 issue, and run wight he bard layout immediately above. It is pretty close to the original, and I know the circuit is now correct.
Cheers,
Scott
Hi Scott,
Actually, using the capacitor to ground allows it to shield itself. Too bad they don't mark which lead goes to the outside foil anymore. That is the side that goes to ground.
I'm super happy that you have one board working, whatever revision it is. Let us know how the new board works.
-Chris
Actually, using the capacitor to ground allows it to shield itself. Too bad they don't mark which lead goes to the outside foil anymore. That is the side that goes to ground.
I'm super happy that you have one board working, whatever revision it is. Let us know how the new board works.
-Chris
Interesting. I was pondering this this morning. For really high frequencies, the order probably doesn't matter, although for wrapped foil caps, your shielding concept might be valid (not the case for chip caps or interdigital layered caps.. )
For lower frequencies , the cap will look like an open circuit, and the wave will reflect off it, which argues for putting the resistor on the circuit side, so that you get dissipation of the wave on the inbound path and the outbound path.
In the end though, if the amp is unstable, these sorts of details are not going to help much.
I like the new layout. It has more or less the same component orientations and positions as the original, and it has about 95% the same trace layouts. As noted above, I put the "jumpers" on the top layer.
I think I'll try one using the original heatsinks. That should allow me to get it 99% the same. The big difference is that I am using a CAD layout tool, and the original was obviously laid out with masking and an X-Acto knife..with those organic sweeps and odd shaped traces.
My reasoning for the new heatsinks was that someone making a new board, or an entire new amp from scratch might no have a usable control board..OTOH, that's probably an edge situation, since 99% of folks who might use this will have an existing amp with an electrolyte saturated control board, so they should be able to just pull the heatsinks off, clean them and go...I'll give that a shot.
I was really amazed that changing the control board output bypass caps by 100 pF completely cancelled the oscillation..
Also, it turns out the labeling on my PC layout was poorly positioned, so I di NOT have J105 and J104 switched.. The package drawing just had the label in the wrong place, so the labels looks switched. I DID have the leads wrong though.
Putting the temp comp connectors on the wrong side will not have any performance effect unless the amp changed temperature asymmetrically, in which case if one side is hotter than the other, the compensation will be adjusting the bias on the wrong output bank.. An interesting situation....
Hi Scott,
-Chris 🙂
Think differential pair. Reduce the current on one side, the other side is instantly reduced as well to keep the DC offset centered. Cool - eh? That is unless you meant that you had L and R swapped. That could get interesting.
It's a series circuit. Ic=Ir. In other words, if the current can't flow steady state, the current flow will also be zero in the resistor (= to the capacitor). No reflecting of electrons. It's a physics thing.
-Chris 🙂
Yeah, I never really understood that temp comp circuit.
However, the reflecting thing actually does occur, with non-DC currents. It's a wave impedance thing (which is a different version of physics). Might not really be measurable at lower frequencies where the wavelength is larger than the parts, but a cap is really just an open circuit, and a wave traveling along a transmission line, when it hits an open circuit will simply reflect. It has to do this to satisfy the boundary conditions of the wave equation. Same thing happens with a short circuit. On order to have a zero amplitude wave at the short, there must be a wave propagating in the opposite direction with opposite polarity.. .You can think of this a little like quantum vacuum.. where what we perceive as empty space is really just a virtual emptiness created by matter and anti-matter. (I just finished reading a book called "A Universe from Nothing"...)
Another way to consider this is that at high frequencies the cap is conducting, so current is flowing. At lower frequencies, current must also conduct through the resistor in order to then find the "open circuit" of the capacitor...,, and then the wave simply reverses itself at the cap and cancels the current in the resistor.
Even with DC this happens, but it happens so fast at turn on, than nobody every cares about it. If you could measure the flow at very small time increments, putting a step function voltage on the resistor you would see that voltage step arrive at the cap and then bounce back in opposite phase (look at a Time Domain Reflectometer sometime.. very cool). With a DC step, once the step wave reflects the current drops to zero...
Cheers,
S
Hi Scott,
Yes, that is the charging current through the capacitor. Once that has stopped, that's all she wrote for current flow. Comparing audio frequencies to high RF is simple. Audio frequencies are like DC with a tiny bit of leakage by comparison.
If you try this without the capacitor, you still see brief current flow due to stay capacitance. Looking at it as a feed line isn't going to work as the length of the traces are so short compared to the wavelengths (corrected distance for PCB traces). It never even comes close to being a transmission line.
-Chris
Yes, that is the charging current through the capacitor. Once that has stopped, that's all she wrote for current flow. Comparing audio frequencies to high RF is simple. Audio frequencies are like DC with a tiny bit of leakage by comparison.
If you try this without the capacitor, you still see brief current flow due to stay capacitance. Looking at it as a feed line isn't going to work as the length of the traces are so short compared to the wavelengths (corrected distance for PCB traces). It never even comes close to being a transmission line.
-Chris
Right, I realize that. I guess even at 3-5 MHz, where I have seen these oscillations, the system is still pretty much lumped parameter.. My early career and grad school were all about microwave devices.
I modified the board layout to use the old heatsinks. This definitely makes the layout cleaner.
At this stage all of the red traces except the ones down at the bottom of the image represent the jumpers on the original layout - more or less.
This layout follows the original pretty closely, although I was uncomfortable with the very thin traces they squashed together down around the trim pot on the original, so I widened this and went for the two topside traces to spread things out a little.
I'll make a run of these and see how they work...
Scott
At this stage all of the red traces except the ones down at the bottom of the image represent the jumpers on the original layout - more or less.
This layout follows the original pretty closely, although I was uncomfortable with the very thin traces they squashed together down around the trim pot on the original, so I widened this and went for the two topside traces to spread things out a little.
I'll make a run of these and see how they work...
Scott
Hi Scott,
Nice work! This one should work well. It does look neat and the squashed have been fixed. I really hope this is the one!
-Chris
Nice work! This one should work well. It does look neat and the squashed have been fixed. I really hope this is the one!
-Chris
Chris;
How would you like me to share this?
I can post the Gerber and Excellon files, or send them to you. Your choice.
Scott
OK, so I just ordered Rev 2 of the Adcom Control Board. This is more or less the one I posted above, only I fattened up one of the ground traces, and fixed a couple of ragged traces.
Should be here late next week, so I'll build it up then and report back. I am pretty confident that this one will work out of the box.
On this round I'll be changing the lower value caps from the ceramic and mica units to WIMA polypropylene caps, and running a new set of matched Darlingtons.
Cheers,
S
Should be here late next week, so I'll build it up then and report back. I am pretty confident that this one will work out of the box.
On this round I'll be changing the lower value caps from the ceramic and mica units to WIMA polypropylene caps, and running a new set of matched Darlingtons.
Cheers,
S
Scott,
Are you still thinking about having some transistor matching boards produced? Or posting the files for that purpose? I'm anxious to get one put together and learning how to do this.
maceo
Are you still thinking about having some transistor matching boards produced? Or posting the files for that purpose? I'm anxious to get one put together and learning how to do this.
maceo
Hi Scott,
Both, if you want to run a group buy, or just post the files. Either is fine by me. I'll give the files to the store guys. Many folks would rather just buy a board when they want to. I don't know if they will release it as a kit, or just board. It would be good as a kit with all the parts to avoid some problems.
Everyone, inexpensive Chinese header strips aren't any good. Tried and ended up with a few $$ worth of garbage. They will work, but the leads catch on the edges in the cup (entry way). Same for the precision resistors. I bit the bullet and used Dale 0.1% resistors for the collector loads and the base loads. Standard 1% resistors will do for the current setting resistors. You need them to have a low temperature co-efficient. 100ppm is a lot better than 250 ppm, or worse. Unspecified. Getting the 50 ppm parts isn't normally expensive and you may find that the less expensive resistors might actually be 50 ppm.
Same thing goes for the DIP switches for setting current. They pass the tail current, as the resistors do (buy 1/2 to 0.6 watt parts, some 1 watt parts are the same size). Get switches that will last.
-Chris
Both, if you want to run a group buy, or just post the files. Either is fine by me. I'll give the files to the store guys. Many folks would rather just buy a board when they want to. I don't know if they will release it as a kit, or just board. It would be good as a kit with all the parts to avoid some problems.
Everyone, inexpensive Chinese header strips aren't any good. Tried and ended up with a few $$ worth of garbage. They will work, but the leads catch on the edges in the cup (entry way). Same for the precision resistors. I bit the bullet and used Dale 0.1% resistors for the collector loads and the base loads. Standard 1% resistors will do for the current setting resistors. You need them to have a low temperature co-efficient. 100ppm is a lot better than 250 ppm, or worse. Unspecified. Getting the 50 ppm parts isn't normally expensive and you may find that the less expensive resistors might actually be 50 ppm.
Same thing goes for the DIP switches for setting current. They pass the tail current, as the resistors do (buy 1/2 to 0.6 watt parts, some 1 watt parts are the same size). Get switches that will last.
-Chris
I ordered a couple of boards from my design, so I can assure that they work. Chris suggested I make them available through the website, but I am not sure what that means (see a few posts above).
The boards at 2X cost $70 each (incl. tax and postage). They are much cheaper in quantity, but you have to buy a couple to determine if they work in the first place. So, assuming this one works properly, I'll sell it to you at my cost, which will be $70 plus postage - probably about $75.
Scott
Hi Scott,
I'm suggesting that we give the files to the store and allow them to worry about it. In the mean time, members who can print and make a board can do so. This is for the transistor matching jig.
For your new Adcom PCB, that is something you have complete control over. It's totally your design. Even just posting the board files should be enough, but you could run a group buy as well for some minimum quantity of PCBs. I don't know that the subject would be of general interest to the membership like the beta matcher might be. That is what the store is concerned about as they do have to invest in the project to produce hardware to sell. I still have to do some write-ups on how to use it, and why someone should consider using it. I did that once already, but darned if I can find them.
-Chris
I'm suggesting that we give the files to the store and allow them to worry about it. In the mean time, members who can print and make a board can do so. This is for the transistor matching jig.
For your new Adcom PCB, that is something you have complete control over. It's totally your design. Even just posting the board files should be enough, but you could run a group buy as well for some minimum quantity of PCBs. I don't know that the subject would be of general interest to the membership like the beta matcher might be. That is what the store is concerned about as they do have to invest in the project to produce hardware to sell. I still have to do some write-ups on how to use it, and why someone should consider using it. I did that once already, but darned if I can find them.
-Chris
OK. I'll post the files after I test out the board and can be sure it works properly. Then I assume you can pull them down and get them to the forum store folks.
As for the Adcom board..I'll probably see if there is interest in a group run. I have discussed with another member about possibly making and selling fully populated new boards (essentially a drop-in replacement). Not sure how many amps Adcom sold, or what the market might be for this. I'd probably have to sell them for about $200 ea. to make it worth doing. I must say though, unlike a board that has been pickled in electrolyte (that sort of has vague offset control), that board is rock solid in terms offset...I am hoping the new layout solves the oscillation issue.
Hi Scott,
Yes, that works for all of us.
I can't thank you enough for creating the transistor matching jig. I simply didn't have enough time to recreate the layout again.
Your Adcom board will probably save a few amplifiers out there. You should think about boards only, and fully stuffed and tested PCBs. You can test them using a few diodes to fake emitter base junctions of the outputs and a transistor to use as the temperature sensor. The thermal switch junctions only need to be tested for continuity, so a short. You can just use LEDs to check that function. So make a few plug in cables to quickly connect a module and check it out. Peace of mind, and enough value to charge a couple hundred for each one.
-Chris
Yes, that works for all of us.
I can't thank you enough for creating the transistor matching jig. I simply didn't have enough time to recreate the layout again.
Your Adcom board will probably save a few amplifiers out there. You should think about boards only, and fully stuffed and tested PCBs. You can test them using a few diodes to fake emitter base junctions of the outputs and a transistor to use as the temperature sensor. The thermal switch junctions only need to be tested for continuity, so a short. You can just use LEDs to check that function. So make a few plug in cables to quickly connect a module and check it out. Peace of mind, and enough value to charge a couple hundred for each one.
-Chris