"That's why you need a base stopper on *each individual device* when paralleling. Try it without them on a FET output stage and it's almost guaranteed to oscillate."
These transistors aren't insanely fast like FETs, the problem is more likely the previous stage not liking the capacitive input of the last stage. QSC parallells transistors like these without individual base resistors without problems.
"I'm showing this because there's no visible difference between the two on output, and clipping behavior is also the same - clean, symmetrical clipping."
It's those pedestals in the VAS waveform when clipping that might be a problem. Have a look at the current through the base stoppers when clipping into an open output
These transistors aren't insanely fast like FETs, the problem is more likely the previous stage not liking the capacitive input of the last stage. QSC parallells transistors like these without individual base resistors without problems.
"I'm showing this because there's no visible difference between the two on output, and clipping behavior is also the same - clean, symmetrical clipping."
It's those pedestals in the VAS waveform when clipping that might be a problem. Have a look at the current through the base stoppers when clipping into an open output
CBS240 said:Replace those slow MJL21193/MJL21194's with a pair of MJL4281A/MJL4302A and see if it is still stable without base resistors........
Hi,
I thought about this, but didn't mention it (fear of ridicule
). I will replace Q3 and Q15 with the faster units. I have a few of these here.I have attached the schematic from post#158 unchanged for perspective.
AndrewT said:
Hi Andrew,
I'm not sure I can get that to work, as the 6-0-6 is a centre tapped transformer and I need the voltage from it for other duties.
Do you see a problem with its rectified and smoothed DC in series with the main supply for the front end?
Attachments
yes,MJL21193 said:
the ripple on the heavy current supplies is imposed on the low current supply and either requires more overhead before the regulators or better PSRR to prevent the ripple appearing at the output of the Voltage amp stage.
A dozen or so turns of 0.6mm wire added to your toroid is needed for each Low Current supply. And one low current bridge rectifier.
megajocke said:
But QSC mounts everything one one PC board, with undersized heat sinks with the collectors at ground potential, and the layouts don't use huge copper flood areas to minimize the resistance. The RMX2450 also uses only 4 in parallel. Those layouts are more "ideal" from a parasitics standpoint, and in those cases stoppers are rarely needed. Go running bus wires off the PCB to remotely mounted devices spread out all over hades (or the huge copper flood areas I use for buses on the PCB) and you usually have to do more to tame it. Usually, if I use more than 4 in parallel, I design the stoppers in. I've had too many amps that oscillated for no apparent reason (other than possible parasitics) that frequency compenation wouldn't help but 1 ohm stoppers cleaned right up.
MJL21193 said:
Hi wg_ski,
Do you really think it's absolutely necessary to have base resistors on all of my output stage devices? I have decided to have one to supply all, not wanting to tear down the work I've done already unless it's needed.
This amps predecessor, the P68 500 watt version had no base stoppers at all, not even on the drivers. I guess Rod doesn't believe in them. Anyhow, that amp ran stable, and it had 4 paralleled outputs per side plus the CFP.
I just got lucky, I think.
Not *absolutely* but many times it fixes things that have no apparent cause. Actually, the causes are well understood, if you have a model for your circuit (specifically, the circuit construction) that is valid up to a couple megahertz. As a general rule, if an audio circuit has a poor RF layout, you'll need to work on stability a little more than with a good RF layout. That usually means a little less open loop gain, and lower power bandwidth. Base stopper in the OPS scrub a little off both.
Why put them in the bases of each output? Because the parallel bank is actually a distributed network, with unequal resistances and inductances to each device. From an RF standpoint, a fixed resistor in series with each devcie helps equalize this.
MJL21193 said:
They shouldn't be. The entire darlington triple should be running off the main supply. If you don't you'll find yourself burning out drivers when you clip heavily. Not to mention that the outputs will "stick" to the rails when saturated. Even in the preferred location, the predrivers will stick to some extent, but those MJE's are much smaller and faster and the sticking isn't as pronounced.
Hi,
1000W into 2r0 requires 63Vpk and 32Apk. Can you get F16A fuses for the supply rails?
The drivers should send about 1Apk to the outputs to maintain your 1000W near continuous capability. The transient peak current could be near 90Apk.
Music should require a continuous driver current well below this so the 440mAdc continuous rating of 0.6mm diam secondary wire is OK.
The current peaks should come from the smoothing caps and the reactive speaker current demand also comes from the smoothing/decoupling caps.
I would suggest using a regulated supply to the front end (upto the pre-divers) and unregulated to the drivers and a big capacity unregulated supply to the output devices.
But, never having built a beast like this who am I to recommend anything?
1000W into 2r0 requires 63Vpk and 32Apk. Can you get F16A fuses for the supply rails?
The drivers should send about 1Apk to the outputs to maintain your 1000W near continuous capability. The transient peak current could be near 90Apk.
Music should require a continuous driver current well below this so the 440mAdc continuous rating of 0.6mm diam secondary wire is OK.
The current peaks should come from the smoothing caps and the reactive speaker current demand also comes from the smoothing/decoupling caps.
I would suggest using a regulated supply to the front end (upto the pre-divers) and unregulated to the drivers and a big capacity unregulated supply to the output devices.
But, never having built a beast like this who am I to recommend anything?
wg_ski said:
That's a good point. Still it feels like an easy test adding in common resistors to see if that's enough. I've seen an emitter follower BC547 taking of at a couple of MHz driving just one BD139 or similar transistor so it doesn't feel too remote that isolating driver and output will be enough in this case.
If it isn't enough then the whole thing probably needs to be taken apart to install base resistors like you say.
megajocke said:
Do you mean the base stoppers on Q2 and Q24? They show an increase in current during clipping, but it still is below 1mA. Rms current through Q3 and Q15 jumps from 167uA to 15mA with 1Vpp clipping and no load. 4.8mA to 18mA with a 4 ohm load.
I am not seeing any increase in current through the VAS though, under any circumstances.
I don't know if I'd be comfortable with saturating the output devices, shoot through current through the output stage on clipping recovery might be possible. I don't believe the <1dB extra output that can be gotten is worth the ugly clipping recovery and maybe even potential output stage destruction if it's clipping on a moderately high freqency signal.
All amps with higher input stage voltage I've seen use some sort of clamping to make sure only VAS and diff stage can ever saturate. Without separate rails for input stage there is no need for such circuitry though.
wg_ski suggested 1 ohm stoppers per transistor might be enough, and with this low value (equivalent to even less than a tenth of the value you put in) there should be no need for an extra supply and the problems it can create. The MJLs have about 2V of Vce(sat) so there isn't much to be gained really if the base resistors are kept small. The common resistors can probably be reduced a lot from the 3.3 ohms if the problem is oscillation due to incompatibility between output stage and last driver. If it's interaction between the individual output devices that is the problem then you probably need to follow wg_skis's advice with individual resistors.
All amps with higher input stage voltage I've seen use some sort of clamping to make sure only VAS and diff stage can ever saturate. Without separate rails for input stage there is no need for such circuitry though.
wg_ski suggested 1 ohm stoppers per transistor might be enough, and with this low value (equivalent to even less than a tenth of the value you put in) there should be no need for an extra supply and the problems it can create. The MJLs have about 2V of Vce(sat) so there isn't much to be gained really if the base resistors are kept small. The common resistors can probably be reduced a lot from the 3.3 ohms if the problem is oscillation due to incompatibility between output stage and last driver. If it's interaction between the individual output devices that is the problem then you probably need to follow wg_skis's advice with individual resistors.
megajocke said:
Yes, I'll scrub that idea. Not worth the extra power supply complexity either.
Moving forward, I will try the output stage with just a single base stopper, with the thought that I can always add individual resistors if it doesn't work.
I have ordered some 10W, 36 ohm resistors to make another dummy load, but I've just measured the ones I made and Speaker Workshop shows these as 1.44 ohm resistors, so they are not inductive. I will use some of the resistors to increase the impedance to 2 ohms.
AndrewT said:
I'm not familiar with "F16A". If you mean fast acting 16 amp, then yes, I can get these. Please clarify.
TBH, I was not going to use rail fuses, just ones on the mains input. This amp will be connected and left alone, with not much chance of shorts. I am also constructing heavy duty, stupid proof connectors for the speaker cable, to avoid mishaps.
As I've mentioned above, I've ditched the higher rail front end. Given the end use for this amp, I don't want to add complexity by regulating the front end supply.
This "beast" is not bad...wait until I reveal what other projects I plan to do.
Hi,
I posted in Leach and Krell clone threads that the rails fuses can be about half the peak current into the lowest resistive load for maximum continuous output testing.
Quasi prefers about half the RMS current into the worst load.
So you can try testing with rail fuses between F10A and F16A to see which survive the worst case resistive loads on maximum output.
If you use 4ohm speakers then reduce the fuses to F5A to F8A for your intended use.
Music transients should never blow these fuses and it should require a "fault" for them to rupture.
Check that the output remains near zero Volts DC when only one or other fuse is removed.
I posted in Leach and Krell clone threads that the rails fuses can be about half the peak current into the lowest resistive load for maximum continuous output testing.
Quasi prefers about half the RMS current into the worst load.
So you can try testing with rail fuses between F10A and F16A to see which survive the worst case resistive loads on maximum output.
If you use 4ohm speakers then reduce the fuses to F5A to F8A for your intended use.
Music transients should never blow these fuses and it should require a "fault" for them to rupture.
Check that the output remains near zero Volts DC when only one or other fuse is removed.
AndrewT said:
This makes sense, as each rail is conducting only half of the voltage swing, therefore half of the current. I will reassess my position and most likely make provisions on the power supply board for rail fuses.
I will be doing a Randy Slone speaker protection circuit, that guards against DC offset on the output. I thought that this was a very important thing to do, given the high DC potential here. This circuit includes a turn on delay for the speaker relay, and will halt this if DC is detected at startup.
That's one of the problems with an amp this big - the support circuitry dwarfs the amp circuitry.
With power supplies, soft start, amp protection, speaker protection. It all adds up, but it's better than seeing much hard work go up in smoke.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.