MJL21193 said:
Ok, I tried it but somehow it slipped my mind that it's not needed here. Remember way back, I raised the VAS current (as you suggested, to maximize the gain of the MJE15031 I was using at that time) and noticed clamping action. The weak link was the EF, and your suggestion to put a 5K resistor collector to ground nicely limits current there during high frequency clipping to under 20mA.
Of course, this is in the simulation and I haven't checked this on the prototype.
Its looking good John.
We have found here where I work over the last several years that optimising component placement to allow for an unbroken contigous ground plane works best. (We do control boards for generators amongst other things so have high current, low signal, analogue, digital on one board no RF 🙂 thank god, and have found layout becoming critical for signal integraty, so my tuppence worth has been based on experiences etc we have encounted and changes in our working practices. We also use audio circuits for test equipement etc and are implementing full grounds on these.)
Another thing a full ground does is reduce mutual inductive coupling between signals.
Any way I hope I've helped in some way, as there only seems to be you and Pete left on this thread, which is a shame because it is an informative thread. (Though I can only comment on layout)(I've been bad with broncitis hence my abcence).
We have found here where I work over the last several years that optimising component placement to allow for an unbroken contigous ground plane works best. (We do control boards for generators amongst other things so have high current, low signal, analogue, digital on one board no RF 🙂 thank god, and have found layout becoming critical for signal integraty, so my tuppence worth has been based on experiences etc we have encounted and changes in our working practices. We also use audio circuits for test equipement etc and are implementing full grounds on these.)
Another thing a full ground does is reduce mutual inductive coupling between signals.
Any way I hope I've helped in some way, as there only seems to be you and Pete left on this thread, which is a shame because it is an informative thread. (Though I can only comment on layout)(I've been bad with broncitis hence my abcence).
re th e discussion on the VAS current limiting problem. Another way to tackle this and avoid all the issues around introducing distortion due to partial conduction of th e clamp circuit (NB gets worse as temperature increases as well) is to use a cascode. Between th e cascode emitter and the amplifier transistor collector, insert a current limiting resistor.
This does have th e disadvantage of losing a few volts swing, but in some cases this may be a better tradeoff than trying to optimise the current limit circuit.
This does have th e disadvantage of losing a few volts swing, but in some cases this may be a better tradeoff than trying to optimise the current limit circuit.
marce said:
Hi Marc,
Hope you're feeling better.🙂
Yes, you're help is much appreciated. Not a very compelling subject, I guess. Just another AB amp, nothing special.
I am happy with this layout now. Barring a small change here or there, I'll go with this.
I have some different ideas on a few things that I'd like to use with this amp. The previously mentioned clipping/short circuit detector is one, heat management is another that I want to explore.
This thread is long from dead.😉
Bonsai said:
Hi Bonsai,
Do you mean a cascode on the VAS? This was mentioned (I think) earlier by LazyCat. I did not implement it because the simulation shows high frequency clamping here without VAS protection. This came about after I raised Vas operating current.
In your opinion, is this a real world thing or only in the simulation?
MJL21193 said:
Hi John, The 5K resistor in the EF collector to ground path only protects the EF, not the VAS itself. The EF will provide more than enough base current to the VAS to fry it with the output shorted. It is why we chose something like a 1.5A VAS rather than a .1 to .2 A rated VAS transistor. However, even with a 1.5A VAS you have to consider SOA, and temp derating, to be sure that it does not fail under shorted/overload conditions. I believe that we got different results under short circuit conditions, probably due to output transistor model differences. The VAS has a more difficult time with output devices that show beta droop above say 10 A or so.
Pete B.
Hi Pete,
Well, I'm counting on the clipping detector to take the short circuit condition out of the picture (if it reacts quickly enough). This leaves the high frequency clipping.
What I am seeing in the simulation though is clamping - the VAS does not draw excess current when driven well into clipping at 20KHz. This was not the case before I raised VAS current, or am I giving credit to the wrong fix? Did I do something else at that time to alter this behavior?
What I'm asking is this a real condition or is the simulator getting it wrong?
Well, I'm counting on the clipping detector to take the short circuit condition out of the picture (if it reacts quickly enough). This leaves the high frequency clipping.
What I am seeing in the simulation though is clamping - the VAS does not draw excess current when driven well into clipping at 20KHz. This was not the case before I raised VAS current, or am I giving credit to the wrong fix? Did I do something else at that time to alter this behavior?
What I'm asking is this a real condition or is the simulator getting it wrong?
it won't work quickly enough.MJL21193 said:
How long will it take to trigger the relay activation circuit?
How long will it take for the relay to drop out?
How long will it take for the arc across the relay contacts to extinguish?
How long will the semiconductor junctions survive during your clip or short circuit?
How do these timings compare?
We already know that close rated fuses are too slow to protect the semiconductors.
Your relay circuit is better able to prevent long term overheating.
MJL21193, it is a practical solution. Best way to test it on your simulation is to drive your amp into severe current limiting and check the VAS current. The VAS current should be no more than the cascode base bias voltage (I'd use 3.9V) divided by the VAS amplifier transistor emitter degeneration resistor + the resistor I mentioned earlier.
BTW, your new layout looks a lot better - similar to my layout with small signal stuff at the opposite end of the board compared to the power side.
Good luck with your amp.
BTW, your new layout looks a lot better - similar to my layout with small signal stuff at the opposite end of the board compared to the power side.
Good luck with your amp.
AndrewT said:
Ah Andrew,
On your morning rounds, spreading optimism and good cheer.🙂
My idea is only as slow as the components involved. The comparator sees a difference at the LTP, it turns on the transistor, this in turn opens the relay. This should be very quick, no? Certainly faster than a fuse. This scheme is only useful for clipping and short circuit protection, not long term overheating (for that I'll use the the thermal cut-outs).
As mentioned before, a properly sized relay with a high DC rating will not maintain an arc.
There is only one way to answer this in the real world - I'll mock up the circuit and try it. I have some old output transistors that I could swap with the MJL3281A/1302A, that I won't miss.
Hi,
I think you should separate the two functions.
Clipping detection and input signal mute can all be done electronically, at the input. It can be nearly instantaneous i.e.<<1mS, maybe <10uS.
If the mute is a relay that closes to short out or attenuate the input signal then any transistor effect across the input signal is avoided, although a little capacitance remains.
The short circuit protection can be cutting the supply rails, or shorting the supply rails to ground or opening the output line, or electronic current limiting. The only version likely to be fast enough to reliably save the semiconductors is electronic current limiting. However, the heat generated during this limiting action is severe, requiring a second stage of protection. This could be supply rail cut-off or output line opening. Supply rail cut-off would also protect the speakers from a failed output stage that shorts to supply rail.
Your temperature sensing on the main heatsink can also be of some use here.
All of the above can be latched, or delayed return to normal, or instant return to normal. You have to decide for each senario which of these is more appropriate.
I think you should separate the two functions.
Clipping detection and input signal mute can all be done electronically, at the input. It can be nearly instantaneous i.e.<<1mS, maybe <10uS.
If the mute is a relay that closes to short out or attenuate the input signal then any transistor effect across the input signal is avoided, although a little capacitance remains.
The short circuit protection can be cutting the supply rails, or shorting the supply rails to ground or opening the output line, or electronic current limiting. The only version likely to be fast enough to reliably save the semiconductors is electronic current limiting. However, the heat generated during this limiting action is severe, requiring a second stage of protection. This could be supply rail cut-off or output line opening. Supply rail cut-off would also protect the speakers from a failed output stage that shorts to supply rail.
Your temperature sensing on the main heatsink can also be of some use here.
All of the above can be latched, or delayed return to normal, or instant return to normal. You have to decide for each senario which of these is more appropriate.
How about instead of cutting output to the speakers, it operates similar to a traditional short circuit protection by pulling the base of the drivers to ground? It could use much smaller, faster relays.
I will try this in the simulation and see how it works.
I will try this in the simulation and see how it works.
According to simulation, it works well. It takes about 12uS for it to trigger (as seen on the virtual scope). The relays work well, as there is no partial conduction prior to their closing, therefore no increase in distortion.
A third relay could make use of the same circuit to cut AC to the rails power supply. This would be latching, and would require the power to be cycled.
A third relay could make use of the same circuit to cut AC to the rails power supply. This would be latching, and would require the power to be cycled.
Hi John,
May I leave further comments on your layout?
I have missed a lot in your thread. I see that you have moved supply and rails to one side, looks fine....exept....
Pardon me if I am reading your picture wrong but it seems like speaker return is going through the ground plane and passing low level stages on its way, why not speaker return directly to supply ground and from there a reference to a ground plane?
May I leave further comments on your layout?
I have missed a lot in your thread. I see that you have moved supply and rails to one side, looks fine....exept....
Pardon me if I am reading your picture wrong but it seems like speaker return is going through the ground plane and passing low level stages on its way, why not speaker return directly to supply ground and from there a reference to a ground plane?
Attachments
4fun said:
Hi 4fun,
Your comments are always welcome.
I fretted over that for a while, but really it doesn't cross any small signal components. As you can see in the pic below, there's not much between the speaker ground and the main ground. There are a handful of the vbe parts and the EF and VAS.
I thought if it turned out to be a problem, I'd re-route it back to the second ground pad.
Attachments
Do you think it will be a problem? Other amps that I have built I have always connected the speaker return to the main chassis ground, not to the board directly. Here I'm being told that's wrong, the speaker return should be close to the output devices. To avoid wasted space on the board and long traces, I have circuitry near the output devices. I though the best place to tap the output trace was near the middle, but If it will make a difference, I could move it nearer the PS caps.
What are the adverse effects of this return current?
What are the adverse effects of this return current?
I concur with 4fun, I would have the speaker return as near the caps as possible as this is the "star point". The current at low frquencys will spread out to use most if not all of the plane as it travels towards the caps, if its next to the caps its no where to travel. A lot of designs (commercial) have the caps of board, but hardwire speaker returns etc directly to the caps. (audiolab 8000a and S are a good example of this, and I belive somewhere on the forums are some pics of the insides, 8000A tweeks ).
VAS Protection
Hi John,
I can't see your simulation so it's hard to say what's going on. The strange HF clipping behavior was solved by the current limit resistor on the EF in front of the VAS.
When I run a modified sim to get more reasonable results into a short circuit, I get as high as 400 mA in the VAS. Note also that I lowered the base stoppers to 2 ohms since I might prefer that value, or 5 ohms depending on stability. Even with higher stoppers I got 200 mA which is excessive. I added a 2N5087 as a clamp right at the VAS using the 10 ohm emitter resistor as the sense, and it clamps at about 72 mA. This makes sense since one Vbe drop or .7V on 10 ohms is 70 mA.
It did not raise the distortion at all. All small signal transistors I tried worked fine, 2N3906 for low capacitance, it's not critical. The clamp transistor is not stressed at all, since the EF is current limited to about 10 mA, and the VAS Vbe junction limits the clamp Vce to under a few volts. The EF shields the diff pair from the added capacitance of the clamp.
This is a good fix, the EF and VAS are now protected.
Pete B.
MJL21193 said:
MJL21193 said:
Hi John,
I can't see your simulation so it's hard to say what's going on. The strange HF clipping behavior was solved by the current limit resistor on the EF in front of the VAS.
When I run a modified sim to get more reasonable results into a short circuit, I get as high as 400 mA in the VAS. Note also that I lowered the base stoppers to 2 ohms since I might prefer that value, or 5 ohms depending on stability. Even with higher stoppers I got 200 mA which is excessive. I added a 2N5087 as a clamp right at the VAS using the 10 ohm emitter resistor as the sense, and it clamps at about 72 mA. This makes sense since one Vbe drop or .7V on 10 ohms is 70 mA.
It did not raise the distortion at all. All small signal transistors I tried worked fine, 2N3906 for low capacitance, it's not critical. The clamp transistor is not stressed at all, since the EF is current limited to about 10 mA, and the VAS Vbe junction limits the clamp Vce to under a few volts. The EF shields the diff pair from the added capacitance of the clamp.
This is a good fix, the EF and VAS are now protected.
Pete B.
MJL21193 said:
Relays on the output are usually used to protect speakers from turn on thumps, or DC if the output stage fails and someone puts in oversized fuses that do not blow. You can certainly add short circuit detection but once again, I would make it current based not clipping based.
You might want to have a look at the Hafler DH-500 output relay circuit, which provides DC offset protection. The schematics are on the web.
There are many simpler ways to provide short circuit protection, the traditional sense across the output emitters is most common obviously, so why choose a design with relays on the drivers? Mechanical devices have much lower reliability that properly designed semiconductor based circuits.
Pete B.
- Status
- Not open for further replies.