Hi Woody,
Good questions.
#1 No reason, just neatness. If you'd like to put it under the pcb, where it is undeniably more accessible, go for it!!
#2 C5 and C6 remain deleted from my working amplifier. Stability is rock steady with my TL speakers, the VSonics. I'm confident even with electrostatics, given the lowish loop gain of the amp (around 50dB), no further additions for C5 and C6 are actually required. Forget 'em...... mere humbug, insurance in the early stages of design, nothing more.
#3 I measure the bias current across the 0.1R source resistors. I set 40mV, which corresponds exactly to 400mA. You'd find it's rock steady and quite non-critical to set. The use of three diodes in a string and a much reduced rheostat (read 'pot') all in series makes this adjustment a doddle.
#4 You are very welcome, Woody. This could well become a great DIY forum amp as people discover the wonderful music it makes. The more congenial we keep this and related threads, the more people will be tempted to build, knowing they won't have their heads removed by remote internet diathermy every time they ask a question.....
Cheers,
Hugh
Good questions.
#1 No reason, just neatness. If you'd like to put it under the pcb, where it is undeniably more accessible, go for it!!
#2 C5 and C6 remain deleted from my working amplifier. Stability is rock steady with my TL speakers, the VSonics. I'm confident even with electrostatics, given the lowish loop gain of the amp (around 50dB), no further additions for C5 and C6 are actually required. Forget 'em...... mere humbug, insurance in the early stages of design, nothing more.
#3 I measure the bias current across the 0.1R source resistors. I set 40mV, which corresponds exactly to 400mA. You'd find it's rock steady and quite non-critical to set. The use of three diodes in a string and a much reduced rheostat (read 'pot') all in series makes this adjustment a doddle.
#4 You are very welcome, Woody. This could well become a great DIY forum amp as people discover the wonderful music it makes. The more congenial we keep this and related threads, the more people will be tempted to build, knowing they won't have their heads removed by remote internet diathermy every time they ask a question.....
Cheers,
Hugh
I had a problem yesterday.
Brought down my separated PSU from the amp PCB version. C4, C5 & C6 not fitted.
Coupled it all up.
Used the bulb tester for first start up. The bulb came on. OOPS, why?
Checked voltages, offset, Rails, noise/hum, etc. Could see no fault. Ahh, Vac @ output = ~750mVac
Changed from 60W bulb to 150W bulb. The Zobel smoked !, well I had to try it twice to identify where the smoke was coming from.
Eventually notice the 2.4W of Zobel resistors were very discoloured. I'd guess they dissipated 5W to 10W, they smoked so quickly.
Only after this did I find that I had swapped signal ground and signal hot from the DCB1.
Seems that promoted some form of HF oscillation.
Any ideas?
PS. no damage once I swapped the input leads back the way they should have been.
Brought down my separated PSU from the amp PCB version. C4, C5 & C6 not fitted.
Coupled it all up.
Used the bulb tester for first start up. The bulb came on. OOPS, why?
Checked voltages, offset, Rails, noise/hum, etc. Could see no fault. Ahh, Vac @ output = ~750mVac
Changed from 60W bulb to 150W bulb. The Zobel smoked !, well I had to try it twice to identify where the smoke was coming from.
Eventually notice the 2.4W of Zobel resistors were very discoloured. I'd guess they dissipated 5W to 10W, they smoked so quickly.
Only after this did I find that I had swapped signal ground and signal hot from the DCB1.
Seems that promoted some form of HF oscillation.
Any ideas?
PS. no damage once I swapped the input leads back the way they should have been.
Last edited:
Andrew,
1W resistor in zobel network is only good up to 30kHz, if you go above that then it will be smoked. I did some measurements during the testing on my first version of zobel with 100nF/10R. At 40kHz R19 will dissipate 1.9W and at 50kHz 3W. So if you get any oscillation at high frequency this resistor will not survive and will be smoked really fast.
George
1W resistor in zobel network is only good up to 30kHz, if you go above that then it will be smoked. I did some measurements during the testing on my first version of zobel with 100nF/10R. At 40kHz R19 will dissipate 1.9W and at 50kHz 3W. So if you get any oscillation at high frequency this resistor will not survive and will be smoked really fast.
George
'Not sure whether you suggest changes there or not. Median values of 100nF/8R 1W have been standard in small amps like this for some 40 years. It would be questionable, I think, to use say. a higher rated and possibly inductive R.
The most common frequencies of amplifier oscillation are between 350kHz and about 10Mhz for BJT outputs and up to even 100MHz for Mosfets. I understand that there would be no saving an audio amp oscillating at full output at these frequencies, for which the Zobel network is usually intended.
The most common frequencies of amplifier oscillation are between 350kHz and about 10Mhz for BJT outputs and up to even 100MHz for Mosfets. I understand that there would be no saving an audio amp oscillating at full output at these frequencies, for which the Zobel network is usually intended.
Yes, I agree with Ian. The Zobel resistor is inevitably underrated, because the amp won't survive long periods of oscillation anyway before destroying its output stage - and usually the source resistors too. The Zobel is used to calm the savage breast in normal operation, and since the amp won't normally oscillate at these absurdly high frequencies the Zobel won't ever be taxed.
But if you remove the Zobel, which sits benignly in the amp and appears to do little, you mess with its stability with difficult loads. It has a powerful role in stabilising transient instability with reactive loads.
Cheers,
Hugh
But if you remove the Zobel, which sits benignly in the amp and appears to do little, you mess with its stability with difficult loads. It has a powerful role in stabilising transient instability with reactive loads.
Cheers,
Hugh
Yes - plenty for this app. and a much cheaper, smaller fire if, as in Andrew's case, you accidentally smoke it.
Since it does not protect against high amplitude oscillation, there is no point to over-rating unless the amplifier is also increased in rating to say, 150, 350W etc. where it may be appropriate by scaling, to use 2W, 5W etc.
Since it does not protect against high amplitude oscillation, there is no point to over-rating unless the amplifier is also increased in rating to say, 150, 350W etc. where it may be appropriate by scaling, to use 2W, 5W etc.
Last edited:
if one's mistake forces the amp into HF oscillation then the Zobel will pass significant AC current.
While this is happening and with the calming effect of that damping resistor the amplifier may survive.
If the Zobel, either resistor or capacitor become open circuit due to overcurrent then that calming effect is lost and the amp oscillation could become much worse and the amp then could destroy itself.
A well behaved amp benefits from having an output network that provides the load the amp needs to see to remain stable, but that output Network must not be accidentally removed by a less than destructive oscillation or even a deliberate attempt to test the HF response at high signal level, eg. power bandwidth test.
While this is happening and with the calming effect of that damping resistor the amplifier may survive.
If the Zobel, either resistor or capacitor become open circuit due to overcurrent then that calming effect is lost and the amp oscillation could become much worse and the amp then could destroy itself.
A well behaved amp benefits from having an output network that provides the load the amp needs to see to remain stable, but that output Network must not be accidentally removed by a less than destructive oscillation or even a deliberate attempt to test the HF response at high signal level, eg. power bandwidth test.
How do we know what zobel impedance is required to ensure stability of the amp?
- keantoken
it seems to be trial and error.
Many builders adopt 47nF to 220nF and 1r0 to 22r0.
That is a mighty big range.
I tend towards a total resistance of the various Network paths ~ = design lowest load impedance.
And choose ~100nF to 150nF as the total Network capacitance to ensure the Network load only works at highish frequencies.
Many builders adopt 47nF to 220nF and 1r0 to 22r0.
That is a mighty big range.
I tend towards a total resistance of the various Network paths ~ = design lowest load impedance.
And choose ~100nF to 150nF as the total Network capacitance to ensure the Network load only works at highish frequencies.
I am working on coming up with a more scientific way of choosing Zobel values based on real-world measurements that will leave no uncertainty about potential oscillation. It should require only an oscilloscope and a sinewave generator, preferably working at 50MHz or above, but I think 20MHz will do most of the time for power amps.
Until then, I've discovered in simulation that a 1uF+1R snubber on the rails may greatly aid in the stability of some power amplifiers. This is because at potential oscillation frequencies, say 10MHz and above, poweramps generally have poor PSRR, and phase anomalies in the rails response can have a surprising effect on phase margin.
- keantoken
Until then, I've discovered in simulation that a 1uF+1R snubber on the rails may greatly aid in the stability of some power amplifiers. This is because at potential oscillation frequencies, say 10MHz and above, poweramps generally have poor PSRR, and phase anomalies in the rails response can have a surprising effect on phase margin.
- keantoken
There's no fancy derivation but the resistance is around the lowest nominal value the amp is expected to work into, offering a similar load at HF. The capacitance has been experimentally determined and verified many times by designers and manufacturers alike as a nominal 100-150nF - One size fits all, since the expected job is the same in all full- range domestic audio amplifiers.
Several authors, including Self, will support that. Cordell does it exactly, split in 2 half sections in a Pi configuration, with the coil network. (p202)
It's not necessarily precise. nor is that a simple pi filter. given that you don't need a steep slope. Digging around EDN papers, there's a lot of confusion over even what it does, so watch what you read, some forum opinions are really out there - somewhere.
.
Thanks Ian,
Well put. It is chosen either empirically, or worse, by convention. Neither is it critical.
Often a flameproof resistor is used, and the correct dielectric for the cap is self-healing polypropylene, even X2 rating. The resistor should be mounted 3mm off the board so that in the event of immolation it does not damage the pcb. There is also an argument for a high wattage carbon resistor here; as metal films use a helical, and hence inductive, conductor.
Cheers,
Hugh
Well put. It is chosen either empirically, or worse, by convention. Neither is it critical.
Often a flameproof resistor is used, and the correct dielectric for the cap is self-healing polypropylene, even X2 rating. The resistor should be mounted 3mm off the board so that in the event of immolation it does not damage the pcb. There is also an argument for a high wattage carbon resistor here; as metal films use a helical, and hence inductive, conductor.
Cheers,
Hugh
In the H/K amp I took apart they use a flameproof resistor, and the PCB under it is cut out, probably for the reason you describe.
I keep wondering about the stray inductance of film resistors. It would probably be very small. While carbon resistors don't have a coil, the conductive core has a larger diameter and thus more skin effect. Film resistors, because of the thin resistive layer, won't have as much skin effect. So I wonder. It may even be that multiple low-wattage carbon films (think Litz wire) would be better than a single high-wattage carbon film in terms of skin effect and power dissipation, as with lower skin effect the dissipation won't be confined only to the outer layers of the resistor. However it could be that skin effect is a non-issue for resistors.
Carbon resistors would handle intermittent bursts of oscillation better, because they have better surge handling ability than films.
- keantoken
I keep wondering about the stray inductance of film resistors. It would probably be very small. While carbon resistors don't have a coil, the conductive core has a larger diameter and thus more skin effect. Film resistors, because of the thin resistive layer, won't have as much skin effect. So I wonder. It may even be that multiple low-wattage carbon films (think Litz wire) would be better than a single high-wattage carbon film in terms of skin effect and power dissipation, as with lower skin effect the dissipation won't be confined only to the outer layers of the resistor. However it could be that skin effect is a non-issue for resistors.
Carbon resistors would handle intermittent bursts of oscillation better, because they have better surge handling ability than films.
- keantoken
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.