"Better performance" during what conditions?
When it's freezing outside?
When it's raining/hailstones?
When you're listening to the stereo?
Better at getting from the beginning to the end in the shortest possible time?
No, to all of those questions.
When it's freezing outside?
When it's raining/hailstones?
When you're listening to the stereo?
Better at getting from the beginning to the end in the shortest possible time?
No, to all of those questions.
Exactly, no to all those conditions."Better performance" during what conditions?
When it's freezing outside?
When it's raining/hailstones?
When you're listening to the stereo?
Better at getting from the beginning to the end in the shortest possible time?
No, to all of those questions.
Better performance in music reproduction. 😎
David,
"Manyung Galleries, Mt Eliza Victoria -- also happened to be my first exhibition in Victoria."
Is this you, Sire? Might come along.......
Hugh
"Manyung Galleries, Mt Eliza Victoria -- also happened to be my first exhibition in Victoria."
Is this you, Sire? Might come along.......
Hugh
I just start at about 10 ohm + 10 nF in total. So for 2 pairs use 20 ohm + 5 nF, and so on.
Thanks
No, you need to buy a V16 car to match your amp😉
Best wishes
David
Hi David,
Is there a rule of thumb for working out the zobel values, specifically the value of C? You see a lot of amps with a single zobel of 8R + 100n. Is there a good reason for this or is just because that is what a lot of others use? Would be good to have some reasoning rather than "this looks nice".
Paul
PS. I'll have to go for 3 pairs in my amps. 😉
Dr. Cherry gives guidance in his paper discussing the Thiele Network.
If I remember correctly, he suggests F-3dB <300kHz.
If I remember correctly, he suggests F-3dB <300kHz.
Dr. Cherry gives guidance in his paper discussing the Thiele Network.
If I remember correctly, he suggests F-3dB <300kHz.
Thank you, Andrew.
Going to have a search for that paper... Need to understand this properly rather than just inserting values.
There are many links to his paper on this Forum.
But google will probably be quicker.
I think it was repeated in an electronics today international (ETI) article.
But google will probably be quicker.
I think it was repeated in an electronics today international (ETI) article.
Hi David,
Is there a rule of thumb for working out the zobel values, specifically the value of C? You see a lot of amps with a single zobel of 8R + 100n. Is there a good reason for this or is just because that is what a lot of others use?...
Self recommended 10R + 100nF so I assume most people just copy that, as they do most of his work, usually without credit.
He based this on a few tests but a careful look at his book doesn't reveal that it was very extensive or systematic.
It is desirable to minimize the extra load that the Zobel network places on the amp but obviously the Zobel has to put some load on the amp so it's a bit of a trade-off.
The analytical way to do it is to look at the feedback around the loop(s) as you alter the Zobel. The load of the Zobel should keep the Return Ratio pretty stable even when the amp works into an external open circuit, (and other plausible loads, of course.)
I start at 10R because it seems reasonable to have a load about the same as the nominal load of the amp. I only reconsidered this when Andrew pointed out the chip amp with the Zobel R much smaller than the nominal load.
Makes sense if that is what a chip needs to stabilize it but my inclination for a DIY amp would be to alter other parts of the compensation so that such a shunt load is not needed.
I start at 10 nF because I aim to maximize the ULGF so I want a smaller time constant than typical of Self.
This puts the load on at around 1.5 MHz, close to my ULGF.
I use a Bode maximum feedback compensation scheme so the Return Ratio plot is more complicated than a simple dominant pole scheme.
This makes it important to control the exact point when the load comes on.
The lower value loads the Zobel resistor(s) less, especially if anyone wants to do extreme tests on the amp like a 100 kHz square wave or whatever.
Does that seem reasonable?
Best wishes
David
Last edited:
Self recommended 10R + 100nF so I assume most people just copy that, as they do most of his work, usually without credit.
He based this on a few tests but a careful look at his book doesn't reveal that it was very extensive or systematic.
My limited knowledge is based on AndrewT's recommendations in the past. Haven't really studied Self's explanations in his book. Spend more time reading Bob's book and the linear audio publications. Will provide some feedback on your articles soon. 🙂
It is desirable to minimize the extra load that the Zobel network places on the amp but obviously the Zobel has to put some load on the amp so it's a bit of a trade-off.
The analytical way to do it is to look at the feedback around the loop(s) as you alter the Zobel. The load of the Zobel should keep the Return Ratio pretty stable even when the amp works into an external open circuit, (and other plausible loads, of course.)
I start at 10R because it seems reasonable to have a load about the same as the nominal load of the amp. I only reconsidered this when Andrew pointed out the chip amp with the Zobel R much smaller than the nominal load.
Makes sense if that is what a chip needs to stabilize it but my inclination for a DIY amp would be to alter other parts of the compensation so that such a shunt load is not needed.
Makes sense to me. Worryingly my leach amps spent years without any zobel. Didn't realise its significance until relatively recently.
I start at 10 nF because I aim to maximize the ULGF so I want a smaller time constant than typical of Self.
This puts the load on at around 1.5 MHz, close to my ULGF.
And this is the essential info. Figured that it wasn't a case of one size fits all.
Do you go above or below the ULGF?
I use a Bode maximum feedback compensation scheme so the Return Ratio plot is more complicated than a simple dominant pole scheme.
This makes it important to control the exact point when the load comes on.
The lower value loads the Zobel resistor(s) less, especially if anyone wants to do extreme tests on the amp like a 100 kHz square wave or whatever.
I'm still on dominant pole compensation myself so hopefully working out the correct values will be easier. Walking before running and all that. Tonight's studying is now sorted.
Does that seem reasonable?
Yes. Thank you for the reply.
Best wishes
David
ASTX, hope you don't mind me asking these questions in your thread. Don't mean to crash your thread at all. If any offence is being caused, I apologise, and will stop.
Last edited:
Will provide some feedback on your articles soon. 🙂
Feedback on articles about feedback. How appropriate, I look forward to it.
And this is the essential info. Figured that it wasn't a case of one size fits all.
Do you go above or below the ULGF?
I start centred on the ULGF. I used the phrase "close to my ULGF" only to indicate that it does not need to be perfect but I think it should not be too far off.
The Zobel time constant can probably be adjusted above or below, to tweak the Return Ratio a little but it seems clumsy and I have only done limited experiments with this so far.
Best wishes
David
...
ASTX, hope you don't mind me asking these questions in your thread. Don't mean to crash your thread at all. If any offence is being caused, I apologise, and will stop.
You and all others are always welcome! The current topic is "to zobel or not to zobel" so every contribution is very welcome as there is always something for you and me to learn about.🙂
My practical experience is sorta like Self.Self recommended 10R + 100nF so I assume most people just copy that, as they do most of his work, usually without credit.
He based this on a few tests but a careful look at his book doesn't reveal that it was very extensive or systematic.
....
I start at 10 nF because I aim to maximize the ULGF so I want a smaller time constant than typical of Self.
I've come a cropper trying less than 100n for the 'conventional' Zobel. My experiments were for precisely your reasons. 10R seems to be OK.
____________
What IS useful, from sims, some Jurassic analysis & loadsa 'real life' is a smaller resistor in parallel with the Inductor ... say 4R7
Little thought if I may ...
If power supply is 2x70V, that gives about 3.9mA through R2, which gives about 1.3V max on R28, which gives roughly 0.5V max on R22/R66. That is just enoug to fully drive VAS, but what if the power supply collapse during high load (mains voltage also can vary)? I think You should consider decreasing R2 to about 10k.
If power supply is 2x70V, that gives about 3.9mA through R2, which gives about 1.3V max on R28, which gives roughly 0.5V max on R22/R66. That is just enoug to fully drive VAS, but what if the power supply collapse during high load (mains voltage also can vary)? I think You should consider decreasing R2 to about 10k.
And what about fast rising edge of signal? You may need twice more current, to feed R28 and suckout Cbc of Q7 and Q34. So better check voltage on D1 during square vave test (with oscilloscope, or in simulation). Of course You use very low Cbc transistors, so this is not likely to happen, but...
My another idea is to place R28 between base and emiter Q7/Q34. That reduces curent needed by this resistor, and makes it more like current source (near constant voltage over it).
My another idea is to place R28 between base and emiter Q7/Q34. That reduces curent needed by this resistor, and makes it more like current source (near constant voltage over it).
If power supply is 2x70V, that gives about 3.9mA through R2, which gives about 1.3V max on R28, which gives roughly 0.5V max on R22/R66. That is just enough to fully drive VAS, but what if the power supply collapse during high load (mains voltage also can vary)? I think You should consider decreasing R2 to about 10k.
Dear Wojtek,
Nice to hear from you! Thanks for reporting!
R2 15k at nominal voltage (+/-71V) ~ 230mW ~ using 1W resistor
R2 10k at nominal voltage (+/-71V) ~ 340mW ~ using 1W resistor
Should help down to +/- 50V rails. Will adapt it for next tests upcoming days. My fault: have changed R28 from 470 to 330 during the latest experimental tests (higher LTP current for more slew rate steam and to get higher ft range for BC550C).
So you mean I should split R28 into R28a and R28b?And what about fast rising edge of signal? You may need twice more current, to feed R28 and suckout Cbc of Q7 and Q34. So better check voltage on D1 during square vave test (with oscilloscope, or in simulation). Of course You use very low Cbc transistors, so this is not likely to happen, but...
My another idea is to place R28 between base and emiter Q7/Q34. That reduces curent needed by this resistor, and makes it more like current source (near constant voltage over it).
R28a from Q7 base to emitter/R22
R28b from Q34 base to emitter/R66
Yes, 680Ohm each, and if so, R2 can stay at 15k. Of course, after all changes you need check all in simulation - square wave test (check voltage at D1), and all THD test (Im not sure how THD will be affected).
For now Im too lazy to do this myself.
For now Im too lazy to do this myself.
Please specify the EXACT circuit and/or post where it apppears that you are referring to.Yes, 680Ohm each, and if so, R2 can stay at 15k.
I've downloaded at least 4 different versions and some have different component numbering.
Please specify the EXACT circuit and/or post where it apppears that you are referring to.
I've downloaded at least 4 different versions and some have different component numbering.
See latest asc file from post #868 😉
Some "wires" to ips and output load have been cutted due to cap test. So you need to reconnect them ...
Yes, 680Ohm each, and if so, R2 can stay at 15k. Of course, after all changes you need check all in simulation - square wave test (check voltage at D1), and all THD test (Im not sure how THD will be affected).
For now Im too lazy to do this myself.
Thx! Will start simulating and soldering it tomorrow!
- Home
- Amplifiers
- Solid State
- 2stageEF high performance class AB power amp / 200W8R / 400W4R