Was wondering about that.
Since your on 15 volts the drivers need more current.
Since they are driving 2 transistors.
with 100 ohm about its about 4.6 ma
get that up too 5.8 or 6.8 with 82 or 68ohms fir drivers
bias will change just re adjust
Since your on 15 volts the drivers need more current.
Since they are driving 2 transistors.
with 100 ohm about its about 4.6 ma
get that up too 5.8 or 6.8 with 82 or 68ohms fir drivers
bias will change just re adjust
It's a 4 ohm load. Amp running at close to clipping.
The other two graphs at the drivers are also with 4 ohm load and a 1 khz sine wave. Sorry should have written that.
driver current. Q14, Q15
Set by resistor R28 , R31
current value is 100 ohms
so with 15 volt rail be around 4.8 ma
current
4.8 be enough for 1 output transistor
but were driving 2
Could raise current to 5.8ma with 82 ohms
but....were driving tip's sooo
not gracious with gain.
So lets shoot for 68 ohms for R28 , R31
current will be around 6.8 ma
Bias will drop with that current change.
So readjust bias. as well
otherwise overall DC seems fined tuned now.
interesting how that hiss went away.
Set by resistor R28 , R31
current value is 100 ohms
so with 15 volt rail be around 4.8 ma
current
4.8 be enough for 1 output transistor
but were driving 2
Could raise current to 5.8ma with 82 ohms
but....were driving tip's sooo
not gracious with gain.
So lets shoot for 68 ohms for R28 , R31
current will be around 6.8 ma
Bias will drop with that current change.
So readjust bias. as well
otherwise overall DC seems fined tuned now.
interesting how that hiss went away.
Last edited:
Ahhh perfect good.
That is why i was comfortable raising current little high for drivers
because i felt in real life 4ohm loads when output transistor gain drops.
we got healthy drivers.
i like high power test nice.
I bet your dummy load resistors love you by now ..LOL
with all this square wave action
you can test amplifiers and make toasted bread
Aha, but how did you calculate the current? There's roughly a 0,6 V drop across each of R28 and R31 at idle. But then drivers are fully "on" the voltage may be about 15 V meaning that; 0,6 V / 100 ohm = 6,6 mA ?
Thanks for the tip though, I think it's time for some more testing today if there's time.
Ok, so no need to raise the current for the drivers? And yeah I got a quite beefy dummy load. 🙂 It's also a good stress test for the amp, if the heat sink can handle this treatment, you know for sure that they're enough for normal use.
I've been thinking about changing the output transistors too actually. You suggested devices with higher fT in a previous post. After some more looking into to it I've found out that TIP31/32C are about 3Mhz, and so are 2N3055 too which seems to be quite popular. 2N6488/91 is the best alternative I've found so far with fT = 5Mhz. But I've only looked for TO-220 devices so far in my local dealer's catalogue. Any suggestions?
yes raise current on drivers.
TIP31/32 or tipper roo's = Texas Instrument Power
Texas probably dont make them anymore, but its widely available
transistor. 40 watt @ 25c 3... ish Mhz
Assuming the lead spacing on the board is for TO-220 package
switch on over to Motorola's or MJE
MJE 15028 / 15029 120 volt suffix
MJE 15030 / 15031 150 volt suffix
15030/15031 probably easier to find.
same thing motorola probably doesnt make them.
but many will use the MJ suffix
On Semiconductor would be one current manufacture.
Keep in mind modern packages are probably PB free.
So they add a G suffix. and can be helpful to add G suffix
when searching.
MJE15030G and MJE15031G
50 watt package and 30 Mhz Ft
gain will be higher too around 40 to 25
when the old tips barely do 30 to 15 gain
driving 4 ohm loads, oh yeah 15 ...maybe for tips
MJE probably hold up to 25 even 30 gain.
so going high power at 4 ohms the rails wont sag
as much. probably weasel a volt or 2 more power
into 4 ohms compared to TIPs. And Ft is 30 Mhz not 3
TIP31/32 or tipper roo's = Texas Instrument Power
Texas probably dont make them anymore, but its widely available
transistor. 40 watt @ 25c 3... ish Mhz
Assuming the lead spacing on the board is for TO-220 package
switch on over to Motorola's or MJE
MJE 15028 / 15029 120 volt suffix
MJE 15030 / 15031 150 volt suffix
15030/15031 probably easier to find.
same thing motorola probably doesnt make them.
but many will use the MJ suffix
On Semiconductor would be one current manufacture.
Keep in mind modern packages are probably PB free.
So they add a G suffix. and can be helpful to add G suffix
when searching.
MJE15030G and MJE15031G
50 watt package and 30 Mhz Ft
gain will be higher too around 40 to 25
when the old tips barely do 30 to 15 gain
driving 4 ohm loads, oh yeah 15 ...maybe for tips
MJE probably hold up to 25 even 30 gain.
so going high power at 4 ohms the rails wont sag
as much. probably weasel a volt or 2 more power
into 4 ohms compared to TIPs. And Ft is 30 Mhz not 3
After returning from my weekend trip I felt for some more testing. Raised the current by adding another 150 ohm resistor to R28 and R31 (total resistance then equal to 60 ohm). Took all the DC voltage readings with no input or load connected. Had to readjust the bias voltage of course. Then tested the amp and compared it the the other board with R28 and R31 as they were before.
Didn't see any big difference. Output exactly the same for both card, 25W rms into 4 ohm. And square wave the response look also the same. Not saying that you're wrong about raising the current for the drivers. Just curious what exactly the effect will be?
Attached is the DC readings of the amp without any input or load connected. Only differences is that the voltages at the base of the output transistors are now approx 0,2 V higher (lower on negative side).
Didn't see any big difference. Output exactly the same for both card, 25W rms into 4 ohm. And square wave the response look also the same. Not saying that you're wrong about raising the current for the drivers. Just curious what exactly the effect will be?
Attached is the DC readings of the amp without any input or load connected. Only differences is that the voltages at the base of the output transistors are now approx 0,2 V higher (lower on negative side).
Good tip, thanks. Just read the data sheet and they look very interesting. Not very overpriced either it seems like, a few buck per device or so. If they are avalable at my local dealer I'll buy some tomorrow and test.
Got some MJE15030/31G today from my local supplier today, so I changed the output transistor on one of the board to see if there would be any difference. And what I could see from my test - no change at all really.
Stability was about the same as before (tried to lower the compensating cap C8 and it behave the same as before). The power into 4 ohm load was the same as before. Also the slew rate was the same.
I did notice one disturbing thing, though. When I was testing the amps side by side, I noticed that the board with MJE15030/31 transistors had a slight ringing on the positive side of the wave form. This ringin only occured then output level into the 4 ohm load was close to clipping. Maybe about 20 W or so, and it disapeared then the amp was driven into clipping again. At low load there wasn't any problems. Below is channel 1 TIP31/32 board and channel 2 MJE15030/31 devices, 1 khz sine wave into a 4 ohm load.
Tried a couple of different things, adjusting bias. Throwing some capacitor at between base and collectors at driver transistors. But suddenly it disappeared without any action. Can recreate this problem now which really bothers me, don't have I clue what caused it. Hard to believe that it was the MJE15030 devices, or that is interference somewhere in my lab (since it only affected on board and only at high power).
Checked all the DC voltages again (at idle, no input, no load) and can't see any strange, see attachment. Think I may switch back to the TIP31/32 devices if still behaves like this. Or listen and see if there's any different. If I had the possibility to accurately measure the distortion (working an it) now would be a very interesting time to do it. The MJE15030/31 should perform better I still think reading the data sheet.
Stability was about the same as before (tried to lower the compensating cap C8 and it behave the same as before). The power into 4 ohm load was the same as before. Also the slew rate was the same.
I did notice one disturbing thing, though. When I was testing the amps side by side, I noticed that the board with MJE15030/31 transistors had a slight ringing on the positive side of the wave form. This ringin only occured then output level into the 4 ohm load was close to clipping. Maybe about 20 W or so, and it disapeared then the amp was driven into clipping again. At low load there wasn't any problems. Below is channel 1 TIP31/32 board and channel 2 MJE15030/31 devices, 1 khz sine wave into a 4 ohm load.
Tried a couple of different things, adjusting bias. Throwing some capacitor at between base and collectors at driver transistors. But suddenly it disappeared without any action. Can recreate this problem now which really bothers me, don't have I clue what caused it. Hard to believe that it was the MJE15030 devices, or that is interference somewhere in my lab (since it only affected on board and only at high power).
Checked all the DC voltages again (at idle, no input, no load) and can't see any strange, see attachment. Think I may switch back to the TIP31/32 devices if still behaves like this. Or listen and see if there's any different. If I had the possibility to accurately measure the distortion (working an it) now would be a very interesting time to do it. The MJE15030/31 should perform better I still think reading the data sheet.
I just noticed that the current balancing emitter resisters, like R32, are very low in value. I would triple them.
The power trannies aren't going to share very well unless you hand matched them. In any case, with single transistors 200m-ohm is normal, so for twins double that.
May not cure your burst oscillations, though.
So many problems...challenging.
The power trannies aren't going to share very well unless you hand matched them. In any case, with single transistors 200m-ohm is normal, so for twins double that.
May not cure your burst oscillations, though.
So many problems...challenging.
Last edited:
Spent a lot time trying to figure from were these oscillations comes from today. Maybe a had dry solder joint the other day, cause the oscillations I saw yesterday is real for sure. At some output level, just before clipping these oscillations occur at the positive side of the wave form. After probing around I've found that they are most significant at input of the VAS (base of Q10). This is how the oscillation looks at the output with 1 kz sine wave into a 4 ohm load;
And this is how it looks like at base of Q10, same conditions;
So I thought that i could get rid of it by changing the current in the input stage. Had no effect at the oscillations/ringing, only lowered the slew rate when increasing it (strange?). Tried different values for degen resistor R5 & R6 (again), and same result as before. Lower values makes the whole amp unstable. Higher values did have some effect at the ringings, but the slew rate was lowered too much.
The only thing that helped was to increase compensating capacitor C8 to about 147 pf. Then I got rid of the ringings, at least nothing visible at the sine waves. Square wave response looks a bit more unsteady. Don't know if this is will actually cause any problems though as it is only close to clipping. Compared it with the board with TIP31/32 devices, and it actually looks like a better amp now. Output power is same as before, but the slew rate looks like it has slightly improved with the MJE15030/31 (and bigger compensating cap.) Below is channel one MJE15030/31 with 147 pf compensating cap and channel 2; TIP31/32 with 100 pf compensating cap. Both with 4 ohms loads, 20khz square wave and output level somewhere before clipping.
However I'd really like to know why this is happening? Is it the higher gain of the output devices that causes the oscillations at the VAS?
Any clue, @WhiteDragon?
And this is how it looks like at base of Q10, same conditions;
So I thought that i could get rid of it by changing the current in the input stage. Had no effect at the oscillations/ringing, only lowered the slew rate when increasing it (strange?). Tried different values for degen resistor R5 & R6 (again), and same result as before. Lower values makes the whole amp unstable. Higher values did have some effect at the ringings, but the slew rate was lowered too much.
The only thing that helped was to increase compensating capacitor C8 to about 147 pf. Then I got rid of the ringings, at least nothing visible at the sine waves. Square wave response looks a bit more unsteady. Don't know if this is will actually cause any problems though as it is only close to clipping. Compared it with the board with TIP31/32 devices, and it actually looks like a better amp now. Output power is same as before, but the slew rate looks like it has slightly improved with the MJE15030/31 (and bigger compensating cap.) Below is channel one MJE15030/31 with 147 pf compensating cap and channel 2; TIP31/32 with 100 pf compensating cap. Both with 4 ohms loads, 20khz square wave and output level somewhere before clipping.
However I'd really like to know why this is happening? Is it the higher gain of the output devices that causes the oscillations at the VAS?
Any clue, @WhiteDragon?
I just noticed that the current balancing emitter resisters, like R32, are very low in value. I would triple them.
The power trannies aren't going to share very well unless you hand matched them. In any case, with single transistors 200m-ohm is normal, so for twins double that.
May not cure your burst oscillations, though.
So many problems...challenging.
Haven't tried yet, but thanks for the tip anyway.
Yes, a quite challanging project this ended up. Trying to solve the problems one by. Very educational though.
Correct its not the Vas
But very close the VBE multiplier dangling above it.
C10 at 47u is incredible high.
You wont see much ringing on output.
But your detective work is good by looking at the differential output
If C10 is wayyyyy to large that is what youll see in the differential output to Vas
Ringing on one side.
C10 should be very small to catch what is barfing into
it from output stage.
Lower C10 to 10n or 100n
Also why i raised the current to the drivers.
their having a hard time driving the outputs
with feedback no change will be visible on output.
in the differential youll see that flat side hopefully curve out a bit.
if it is the Vas throwing a fit, current for beta enhancement transistor
might still be too high. Change R20 330ohm to 470 ohm
R19 at 1k = pointless. Jumper it
Cdom C8 should never have to be higher than 100p.
I was trying to get it back down to 68 or 82p
First try should be C10 highly highly likely that large value.
causing that ring.
Raising the current of the input anything past 2ma will make the amp
instantly unstable. Leave it at exactly 2ma.
We have a current mirror. That offers incredible improvement to slew rate.
Its also driving the next stage with beta enhancement. which reduces the
load to the input dramatically. So dumping current into a input stage.
only helps with simple circuits. The extra current kinda helps slew rate
when there is no mirror, and driving difficult transistors. this case
Q10 is extremely easy to drive. Increasing current to the input stage with
a mirror makes it oscillate. Ideal would be lower at 1ma but I already know that
is wayy to low for this topology. 2ma is exactly what is needed.
most CFP with mirror input and beta enhancment.
Degen usually ends up being around 172 or 182 ohms
which is just a .1% or .25% tolerance E192
if just 1% or E96 series just make it 182 ohms.
But very close the VBE multiplier dangling above it.
C10 at 47u is incredible high.
You wont see much ringing on output.
But your detective work is good by looking at the differential output
If C10 is wayyyyy to large that is what youll see in the differential output to Vas
Ringing on one side.
C10 should be very small to catch what is barfing into
it from output stage.
Lower C10 to 10n or 100n
Also why i raised the current to the drivers.
their having a hard time driving the outputs
with feedback no change will be visible on output.
in the differential youll see that flat side hopefully curve out a bit.
if it is the Vas throwing a fit, current for beta enhancement transistor
might still be too high. Change R20 330ohm to 470 ohm
R19 at 1k = pointless. Jumper it
Cdom C8 should never have to be higher than 100p.
I was trying to get it back down to 68 or 82p
First try should be C10 highly highly likely that large value.
causing that ring.
Raising the current of the input anything past 2ma will make the amp
instantly unstable. Leave it at exactly 2ma.
We have a current mirror. That offers incredible improvement to slew rate.
Its also driving the next stage with beta enhancement. which reduces the
load to the input dramatically. So dumping current into a input stage.
only helps with simple circuits. The extra current kinda helps slew rate
when there is no mirror, and driving difficult transistors. this case
Q10 is extremely easy to drive. Increasing current to the input stage with
a mirror makes it oscillate. Ideal would be lower at 1ma but I already know that
is wayy to low for this topology. 2ma is exactly what is needed.
most CFP with mirror input and beta enhancment.
Degen usually ends up being around 172 or 182 ohms
which is just a .1% or .25% tolerance E192
if just 1% or E96 series just make it 182 ohms.
Last edited:
Ahhh totally missed that and is the fun with CFB
DC is very picky at every level.
I raised the current for the drivers assuming still driving TIPs
the MJE are very easy to drive.
If driver current to low, rings, too high can also ring.
But usually only a issue with transistors with very good gain.
if driver current too high.
C10 at 47u I still see a problem 10n or 100n likely kill that ring.
Far as driver current.
Driving TIP driver resistors should be 68 ohms
Driving MJE could probably go back up to 82 or 100 ohms.
for R28 , R31
and yes degen be more aggressive with MJE
but dont wanna kill slew to much , but be stable.
As mentioned usually around 172 , 182 ohms
but could at least try 150.
I think the MJE are fast enough to finally make C10 being to large a issue.
we jumped from 3 Mhz to 30 Mhz
@bengtssk
Are the main heatsinks grounded, if not you might like to try that, sorting out these electro-mechanical paths should be one of the first things on the debugging list, also, putting the whole amp inside a box may have an even further effect on how it may behave due to the bottom plate (if made of some conductive material) underneath the PCB is forming a ground plane, at least have the enclosures bottom plate mounted, and make sure it and the heatsinks are all galvanically fully connected to each other and the electrical ground.
It could also be an idea to elaborate whether grounding of the smaller heatsinks have any effect, alternatively try an RC zobel from the smaller heatsinks to ground via a resistor, try a wide range something like 4,7 to 47 Ohm to begin with (the physical capacitor in series with the grounding resistor forming the zobel network can be omitted because the supposedly isolated driver transistor and the heatsink is already forming a virtual capacitor), just make sure in beforehand that the transistor have no galvanic connection with the heatsink by ohm measuring with a multimeter, make also sure the probe going to the heatsink have proper contact with it, that is because the black colored el-ox covering the heatsinks is highly resistive so you must scratch off a bit of the black el-ox from the heatsink in order to get to the conducting metal the heatsink is made of.
Are the main heatsinks grounded, if not you might like to try that, sorting out these electro-mechanical paths should be one of the first things on the debugging list, also, putting the whole amp inside a box may have an even further effect on how it may behave due to the bottom plate (if made of some conductive material) underneath the PCB is forming a ground plane, at least have the enclosures bottom plate mounted, and make sure it and the heatsinks are all galvanically fully connected to each other and the electrical ground.
It could also be an idea to elaborate whether grounding of the smaller heatsinks have any effect, alternatively try an RC zobel from the smaller heatsinks to ground via a resistor, try a wide range something like 4,7 to 47 Ohm to begin with (the physical capacitor in series with the grounding resistor forming the zobel network can be omitted because the supposedly isolated driver transistor and the heatsink is already forming a virtual capacitor), just make sure in beforehand that the transistor have no galvanic connection with the heatsink by ohm measuring with a multimeter, make also sure the probe going to the heatsink have proper contact with it, that is because the black colored el-ox covering the heatsinks is highly resistive so you must scratch off a bit of the black el-ox from the heatsink in order to get to the conducting metal the heatsink is made of.
Interesting thought, @Ultima Thule . I mounted the transistors at the main heat sinks isolated from the heat sink on purpose, so I eventually could ground the heat sinks if wanted to. At the moment the "amp" is till on the workbench without any enclosure, but then i finally (I hope) finish and put it in a enclosure they will probably be ground to the chassis. Putting some jumpers at them to ground doesn't seem to have any effect at the moment.
Maybe I'll try that with the smaller heat sinks tomorrow.
Maybe I'll try that with the smaller heat sinks tomorrow.
Don't know why I missed this. Was mentioned by @patrick101 earlier. Replaced C10 with 100n. Decreased driver current by changing R28 and R31 back to 100 ohm again, misunderstanding by me. Also increased the the degen resistors R5 & R6 to 188 ohm (120 + 68), but I was still not able to lower compensating cap C8 more than 147 pf. At 147 the amp had no ringing at all, and DC offset was about 1 mV at output. With C8 = 100 pf there was this ringings again...
However what surprises me is that the frequency response seems to have improved with larger value of C8 (!). There most be something seriously wrong with the VAS I'm suspecting now. Will have to try to change those resistors values tomorrow. I agree that C8 should be decreased some how, just how to found out how to.
Here's with C8 = 100 pf (channel 1), the board with TIP's in comparison (channel 2). Notice some ringing at channel 1 too;
With C8=147, it looks pretty good actually;