That looks really nice!!!
(My current project looks like this: http://i363.photobucket.com/albums/oo80/megajocke/inductor tester small/Splat-test036.jpg)
(My current project looks like this: http://i363.photobucket.com/albums/oo80/megajocke/inductor tester small/Splat-test036.jpg)
jacco vermeulen said:
Thank you jacco,
Not quite a master, he's the guy that has done and learned all this and has trained some apprentices to do his work for him. I'm still doing all the work 😉
Just something to keep me out of trouble.
G.Kleinschmidt said:
Hi Glen,
I'd consider doing some for trade, either in parts or competent help. During this whole project, the board were a very small (and rather enjoyable) part, even with all of the revisions I went through.
Back to the action:
I would like to know why I'm getting worse square wave performance with the Zobel connected. I'm seeing overshoot at 1k double in magnitude to what it is without the Zobel.
I removed the Zobel and just used an L//R with an open air coil - about 12 turns around a 10mm cap as the form. This has cut the overshoot down.
megajocke said:
Hi Mega and thanks,
Don't fret, many of mine have looks just as pretty as yours😀
I have made some changes that seem to improve a few things. I replaced the clamp diodes (equivalent to BAV21) with the UL1003's that I have. Clipping performance is much cleaner. I'm not sure why that is, but who am I to argue. I did this on the board that still has the 47pF Miller cap and HF stability looks better (there was some ringing when driven close to clipping before).
I have been experimenting with different size inductors to find the best fit. I want to eliminate the overshoot without overdoing it. So far I have found that 9 turns on a 12mm form is too little and 15 turns on the same form is too much. The original 12 turns on a 10mm form is the best fit so far.
All of this means that I will need to undo the Zobels at the binding posts. What a pain.
Ok. Zobel is out. I know that many say that the Zobel should be there regardless and that it shouldn't have an impact. Every other amp I've built has needed a Zobel or it would oscillate or ring. Not this one, so I'm going to go with that and just use the L//R to damp the slight overshoot.
I have arived at the best-fit inductor. This paralelled with the 4.7ohm resistor is giving me the results I was looking for. Overshoot is controlled, clipping performance is good and the amp easily handles a capacitive load (I've tested up to 1uF). Amp is stable and the idle current is steady.
A couple of hours down the drain to install the Zobels on the binding posts, plus the time to remove them. All in the name of progress though.
I need to pull the other boards out and make the changes I did on this one. I'll also need to make 5 more inductors and I think I'll mount these, along with the resistors, on small boards next to the binding posts.
Here's a look at the inductor. 20 gauge wire, 14 turns on a 10mm form.
I have arived at the best-fit inductor. This paralelled with the 4.7ohm resistor is giving me the results I was looking for. Overshoot is controlled, clipping performance is good and the amp easily handles a capacitive load (I've tested up to 1uF). Amp is stable and the idle current is steady.
A couple of hours down the drain to install the Zobels on the binding posts, plus the time to remove them. All in the name of progress though.
I need to pull the other boards out and make the changes I did on this one. I'll also need to make 5 more inductors and I think I'll mount these, along with the resistors, on small boards next to the binding posts.
Here's a look at the inductor. 20 gauge wire, 14 turns on a 10mm form.
Attachments
Hi,
you mounted the Zobel (R+C) at the speaker binding posts.
Where was the R//L mounted? Before or after the Zobel?
you mounted the Zobel (R+C) at the speaker binding posts.
Where was the R//L mounted? Before or after the Zobel?
AndrewT said:
Hi Andrew,
There wasn't an L//R at first. I thought I would try just the Zobel.
When the Zobel didn't perform as expected (I should have known, this was apparent before) I tried the inductor/resistor before and after the Zobel, I tried to reduce the C on the Zobel, I tried to increase the C also. The same result - overshoot was more pronounced. The inductor would damp it back a bit but I didn't see the reason to do something undesirable (overshoot from Zobel), only to try to counteract it with another mechanism (inductor).
I know that a better approach is to find the source of the overshoot and correct it, and to determine why the Zobel inflames this situation, but the solution I'm using seems to be adequate. The amp runs stable, the inductor kills the overshoot and doesn't complicate things.
Do you have any ideas about why the overshoot would increase with the Zobel?
PS: Sorry for all the big pics. 🙂
I have similar problems:-overshoot that I can't cure.
Big Pics?, no they're only 95k to 170k and some are optional.
Big Pics?, no they're only 95k to 170k and some are optional.
MJL21193 said:
That should not happen. Where are you connecting the CRO probe ground clip?
Excessive squarewave overshoot generally means too little phase margin. First thing to do is to pull the 47pF and try a ~82pF miller comp cap.
Cheers,
Glen
Hi, MJL, very nice work you have done so far (Very DIY)
Since I,m just a little behind you (about to stuff my boards)
with a very similar amp (4 OP EF stage),
this is interesting .
scratching his head as to why his amp's performance degraded
with the zoble.
Could it be where you are placing your zobel..? (after the wire,
at the output jacks) and does the zoble react well driving just
a resistive load (you are using your heatsinked load I saw
in an earlier post??) As you see above I,ve allowed for
change by making a little "daughtercard" just for the output
network.(I use 2.3uH/4.7- L/R & 10/.1uf - R/C)
Here is a thread about zoble's ....http://www.diyaudio.com/forums/showthread/t-120748.html
It seems the famous ESP P3a doesn't like/need the zoble either.
Well, in a few days I'll be at the building/debugging stage
as well .hope you solve this little glitch in a otherwise
beautiful project.
OS
Since I,m just a little behind you (about to stuff my boards)
with a very similar amp (4 OP EF stage),
this is interesting .
An externally hosted image should be here but it was not working when we last tested it.
Prof. Leach ,in a early design phase of his amp ,was also
scratching his head as to why his amp's performance degraded
with the zoble.
Could it be where you are placing your zobel..? (after the wire,
at the output jacks) and does the zoble react well driving just
a resistive load (you are using your heatsinked load I saw
in an earlier post??) As you see above I,ve allowed for
change by making a little "daughtercard" just for the output
network.(I use 2.3uH/4.7- L/R & 10/.1uf - R/C)
Here is a thread about zoble's ....http://www.diyaudio.com/forums/showthread/t-120748.html
It seems the famous ESP P3a doesn't like/need the zoble either.
Well, in a few days I'll be at the building/debugging stage
as well .hope you solve this little glitch in a otherwise
beautiful project.
OS
Hi John,
I may be talking out of turn again, but overshoot and undershoot I have traced to the combination of miller cap and NFB cap.
These must be tuned until the overshoot and undershoot smooths out and you find the typical 1 st order response (slightly rounded).
We had a similar problem on our first few prototypes.
We have successfully dampen the effect by putting a low value resistor in series with the miller cap, whether this is common practice I do not know but it worked for us.
Kind regards
Nico
I may be talking out of turn again, but overshoot and undershoot I have traced to the combination of miller cap and NFB cap.
These must be tuned until the overshoot and undershoot smooths out and you find the typical 1 st order response (slightly rounded).
We had a similar problem on our first few prototypes.
We have successfully dampen the effect by putting a low value resistor in series with the miller cap, whether this is common practice I do not know but it worked for us.
Kind regards
Nico
AndrewT said:
Have you tried adjusting the L//R? It's doing the trick for me. Self shows how reducing the series resistance or increasing the inductance of the coil will tame overshoot.
G.Kleinschmidt said:
The ground clip for these tests was on the wire out from the binding post to the dummy load. I observed the same behavior while testing the amps in their "raw" state, with the probe gnd clip on the main board ground or at the load.
I was operating with 100pF for Cdom and still had this overshoot. Do you think 2-pole compensation would help?
I'm going to test the amp with an inductive load to see how it reacts. If it maintains stability, I think I can safely forget about the Zobel.
ostripper said:
Thanks OS,
I tried every conceivable combination and location for the Zobel. Before the output inductor, after it. Close to the board, back at the binding posts. Changed cap values, all to no avail.
Like i mentioned above, I will do a test with an inductive load and see how the amp holds up. It it goes crazy (it didn't when driving a real 2-way speaker with a fairly complex passive crossover) I'll go back to the drawing board and try to track down the cure. It wasn't a problem in the original Patchwork, maybe due to the slower devices used.
I made small boards to mount my output inductors on. They line up with the binding post spacing, so it was quickly mounted with eyelet connectors soldered to the board. If I decide to go with the Zobel, I'll modify the amp modules for it. Hindsight on that is I should have had it on the board from the get-go.
Nico Ras said:
You never talk out of turn in my book, Nico. 🙂
I will do some experiments and try out your suggestions. The resistor in series with the Miller cap is interesting and different. Easy enough to do also.
Thanks.
MJL21193 said:
Hi John.
When you tried the amp with the 100pF Cdom were you running with the original value LTP emitter degeneration resistors or the higher value ones recalculated for the 47pF Cdom?
To rule out the possibility of insufficent phase margin you need to test the amp with the ~100pF Cdom and the 47pF LTP degeneration resistors.
Cheers,
Glen
G.Kleinschmidt said:
Sorry for the delay Glen,
All of the modules now have 75ohm emitter resistors, so when I changed to 100pF, I didn't change those.
I'll be doing further investigation but in the meantime the output inductor has effectively killed the overshoot. I'm seeing excellent stability driving a capacitive load with an almost perfect squarewave before the inductor while driving the 8ohm dummy and a 1uF cap.
I have just tried an inductive load with a ~.5mH coil in series with the dummy load and I saw no signs of instability. 10kHz squarewave was nearly sawtoothed and here's a look at the 1k squarewave:
Here is the coil:
Probe was connected after the inductor (at the dummy load) on the above pic. This pretty much settles it. I don't need a Zobel here
I have a small problem. The isolation resistor between signal ground and common ground on one of my modules has gone open. This happened to me before, in my big sub amp thread. I blamed it on cheap Sure resistors from Ebay over there, but I have seen advice regarding this exact problem since then. I can't find the thread now, but I think the gist of it was to put opposing diodes in parallel with the resistor. Is this the best way of dealing with this? Or could I use a higher wattage 10 ohm resistor?
I don't know why it happens, but I think it's on initial power up.
I don't know why it happens, but I think it's on initial power up.
Is there a solid connection shorting it somewhere else (some kind of star?) or are these resistors the only ground path from input jacks to central ground?
is it R25
gone open? what does this mean?
either it is Infinitely high resistance = no contact
or shorted = 0 Ohm
I do not use such 10 Ohms.
For me Ground is 0 Volt. And there is only one Ground.
Not several.
So I would use use a jumper and short that resistor out.
To begin with.
But this is me .. 😉
Possibly this schematic:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1636622&stamp=1224513872
Possibly this layout:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1636638&stamp=1224515295
R25, 10R is in the far right lower corner
gone open? what does this mean?
either it is Infinitely high resistance = no contact
or shorted = 0 Ohm
I do not use such 10 Ohms.
For me Ground is 0 Volt. And there is only one Ground.
Not several.
So I would use use a jumper and short that resistor out.
To begin with.
But this is me .. 😉
Possibly this schematic:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1636622&stamp=1224513872
Possibly this layout:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1636638&stamp=1224515295
R25, 10R is in the far right lower corner
megajocke said:
Hi Mega,
Yes, the resistor is the only link with the common ground. There is one on each module.
lineup said:
Hi Lineup,
Open = infinite R (as in open circuit).
I use this technique to kill ground loops and it works extremely well. The input signal ground is effectively isolated from the common ground. Also, no dirty current will flow into the signal ground, as it will take the path of least resistance (straight to the star ground) and not up through the 10R resistor.
I have some amps that are operating like this without a problem. My big sub amp is still fine, being used ever day. Turned on and off every day too.
It would just be nice to have an effective way of ensuring this doesn't happen again.
MJL21193 said:
Ok, I've done that and will do it on the other modules. I used 2 1N4148's and just tacked them on the copper side.
Sure would like to know the cause though.
Are your input jacks isolated from chassis ground?
Do your pre amp and power amp both have 3 prong plugs?
What are you using for a pre amp?
Do your pre amp and power amp both have 3 prong plugs?
What are you using for a pre amp?
- Status
- Not open for further replies.