Here is the first pic of the 300W version of the simple Killer Amp.
The complementary differentials are cascoded allowing the same Ft > 200MHz, Hfe > 450, 65V transistors to be used for inputs diffl's and driver/buffer with 4 pair of matched IRF complementary MOSFETs.
The cascode BJTs are biassed from a divider in the following driver which is modulated with the driver signal out of phase with the input/feedback signal to the differentials. This reduces THD by 2 to 3 times over a standard biassing arrangement.
I stll haven't bought that better camera.
Cheers,
Greg
The complementary differentials are cascoded allowing the same Ft > 200MHz, Hfe > 450, 65V transistors to be used for inputs diffl's and driver/buffer with 4 pair of matched IRF complementary MOSFETs.
The cascode BJTs are biassed from a divider in the following driver which is modulated with the driver signal out of phase with the input/feedback signal to the differentials. This reduces THD by 2 to 3 times over a standard biassing arrangement.
I stll haven't bought that better camera.
Cheers,
Greg
Attachments
Hi Kanwar, Sam 9.
OK,OK I'm shopping for a camera tomorrow! They let me out on day release Saturdays.
Here's the topology.
http://members.dodo.com.au/~gregball/guru_007.htm
Full schematics come with the kit.
Cheers,
Greg
OK,OK I'm shopping for a camera tomorrow! They let me out on day release Saturdays.
Here's the topology.
http://members.dodo.com.au/~gregball/guru_007.htm
Full schematics come with the kit.
Cheers,
Greg
HaHaHa
"You work for small outfits, and can do weird stuff because there is no one there to keep you under control"
I resemble that remark.
Peranders,
THD is 2 to 3 times lower than with a conventional fixed bias cascode. More to come.
The Zout should prove even lower than the smaller 2 o utput pair Killer Amp as the doubled output stage will have 2 x transconductance so 2 x voltage gain, so 2 x NFB, so 2 x lower Zout !
Cheers,
Greg
"You work for small outfits, and can do weird stuff because there is no one there to keep you under control"
I resemble that remark.
Peranders,
THD is 2 to 3 times lower than with a conventional fixed bias cascode. More to come.
The Zout should prove even lower than the smaller 2 o utput pair Killer Amp as the doubled output stage will have 2 x transconductance so 2 x voltage gain, so 2 x NFB, so 2 x lower Zout !
Cheers,
Greg
Hi Greg,
I noticed that your bootstrapped input is also treated by DSelf where he, as you, mentioned a 10 time lower distortion.
(see attached picture and link)
Your cascodes,
I'm surprised that you get 2-3 times lower distortion letting the cascoded transistor working in 180 degree out of phase regarding to the input voltage(if I got it right)... however if set it right you could equalize out the thermal memory distortion, eg. having constant power in your input trany, but that gives only advantage up to couple of hundred Hz where above the thermal distortion is not anymore a major problem even in TO92 typical tranies.
The oposite have been done before bootstrapping cascode trany to the lower tranys emitter giving lower distortion.
What ya think?
Cheers Michael
I noticed that your bootstrapped input is also treated by DSelf where he, as you, mentioned a 10 time lower distortion.
(see attached picture and link)
Your cascodes,
I'm surprised that you get 2-3 times lower distortion letting the cascoded transistor working in 180 degree out of phase regarding to the input voltage(if I got it right)... however if set it right you could equalize out the thermal memory distortion, eg. having constant power in your input trany, but that gives only advantage up to couple of hundred Hz where above the thermal distortion is not anymore a major problem even in TO92 typical tranies.
The oposite have been done before bootstrapping cascode trany to the lower tranys emitter giving lower distortion.
What ya think?
Cheers Michael
Attachments
Michael,
Thanks for the link to Self. I know I have read that particular page before, but had forgotten about it. Just as you say, some amps tie the bias of the diff pair cascodes to the emitters of the diff pair, or to the common point of the emitter resistors. In my understanding, the purpose of this is to try keeping the Vce constant on the diff pair BJTs. That will reduce distorsion cause by the Early effect and non-linearity of Cob. What Self and Greg do, seems to be a quite different thing. What happens here is that the voltage is kept constant over the lower collector resistor, making it act as a CCS. According to the link you posted, Self says that distorsion is reduced by a factor ten due to the higher gain, so I suppose the idea is to increase the OLG this way so the feedback reduces distorsion better.
Thanks for the link to Self. I know I have read that particular page before, but had forgotten about it. Just as you say, some amps tie the bias of the diff pair cascodes to the emitters of the diff pair, or to the common point of the emitter resistors. In my understanding, the purpose of this is to try keeping the Vce constant on the diff pair BJTs. That will reduce distorsion cause by the Early effect and non-linearity of Cob. What Self and Greg do, seems to be a quite different thing. What happens here is that the voltage is kept constant over the lower collector resistor, making it act as a CCS. According to the link you posted, Self says that distorsion is reduced by a factor ten due to the higher gain, so I suppose the idea is to increase the OLG this way so the feedback reduces distorsion better.
While having a shower I sudddenly realized I got things wrong a bit in my previous post. Sensemorale, have showers more often.
What Greg does is, AFAIU, two things.
1) He bootstraps the collector resistor from the VAS, in the same way as Self describes, which gives a higher gain in the input stage.
2) The cascodes work as some kind of local feedback, I think. i was about to write they are Ruch cascodes, but of course they aren't, so I'd better keep shut about them for now. Have to write down the equations or do some sims to grasp this.
What Greg does is, AFAIU, two things.
1) He bootstraps the collector resistor from the VAS, in the same way as Self describes, which gives a higher gain in the input stage.
2) The cascodes work as some kind of local feedback, I think. i was about to write they are Ruch cascodes, but of course they aren't, so I'd better keep shut about them for now. Have to write down the equations or do some sims to grasp this.
Oh no, Christer..
you got my post wrong, it's two diffrent topics I covered in my previous post, I follow you, but again that is not what I talked about.
The bootstrapped circuit as in the attached picture has nothing to do with the cascode Greg uses in his GB300D and why theres lower distortion.
At least I understood Greg that he ment the cascoding alone gave 2-3 times lower distortion, which I make the assumption from that Greg mentioned earlier in his other thread "simple killer amp" where he mention that the Bootstrapping gave him 10 times lower distortion, which is probably more due to the increased FB than lowering the inherent linearity.
So as per my understadning it's the inherent distortion that is lowered by means of cascoding, not through by FB, but the bootstrap may however gain lower distortion by FB because of the increased FB.
So the bootstrap was just a short note (with the attached pic and link to DSelf) alone.
But I may as well have misunderstood the concept discussed herein.
Gregs GB300D again
So reading Greg's first post, he says:
Cheers Michael
EDIT: Christer, I wrote and strugling with my computer for almost 20 min while I discovered you posted again.
EDIT2: According to your latest post, yes! BTW, I have myself forgotten who wrote/invented the cascoding direct to the lower transistor emitter, anybody got the name?
you got my post wrong, it's two diffrent topics I covered in my previous post, I follow you, but again that is not what I talked about.
The bootstrapped circuit as in the attached picture has nothing to do with the cascode Greg uses in his GB300D and why theres lower distortion.
At least I understood Greg that he ment the cascoding alone gave 2-3 times lower distortion, which I make the assumption from that Greg mentioned earlier in his other thread "simple killer amp" where he mention that the Bootstrapping gave him 10 times lower distortion, which is probably more due to the increased FB than lowering the inherent linearity.
So as per my understadning it's the inherent distortion that is lowered by means of cascoding, not through by FB, but the bootstrap may however gain lower distortion by FB because of the increased FB.
So the bootstrap was just a short note (with the attached pic and link to DSelf) alone.
But I may as well have misunderstood the concept discussed herein.
Gregs GB300D again
So reading Greg's first post, he says:
So by the contrary this is not reducing the early effect, though it depends on C6.
Cheers Michael
EDIT: Christer, I wrote and strugling with my computer for almost 20 min while I discovered you posted again.
EDIT2: According to your latest post, yes!
BTW, I have myself forgotten who wrote/invented the cascoding direct to the lower transistor emitter, anybody got the name?Ultima Thule said:Oh no, Christer..
you got my post wrong, it's two diffrent topics I covered in my previous post, I follow you, but again that is not what I talked about.
I realisze that now. Sorry Michael. It is probably because I misunderstoot you that I mixed things up a bit. Not your fault, though.
I agree with you that the cascodes do not linearize the diff pair transistors as ordinary cascodes would, but I think I did say that too. However, I have to think more about this to try understanding what is really happening here. Maybe you already do understand how it works, and if so, I misunderstood that too. Anyway, it's quite an intricate and interesting circuit.
EDIT:
I hadn't seen your edits when I wrote this post.
I am not sure what your "yes" in EDIT 2 refers too, but it sounds like something positive.
Nah, thats simple.. at least for a "visual mathematician" as my IQ test report revealed for me for couple of days!
(Well, I made Tickle's test just for fun on the net in Mars this year, and was now for couple of days suddenly offerd to view the report for free..)
Well, I'm curious to what Gregs comments will be.
Greg, would you mind telling me what is the corner frequency of R10, R11 and C6 node as we don't know your component values?
Cheers Michael
EDIT.. again:
(Well, I made Tickle's test just for fun on the net in Mars this year, and was now for couple of days suddenly offerd to view the report for free..)
Well, I'm curious to what Gregs comments will be.
Greg, would you mind telling me what is the corner frequency of R10, R11 and C6 node as we don't know your component values?
Cheers Michael
EDIT.. again:
..something like that!
Wow.
Hi all. Sorry about the absence - hectic family w'end.
Hi Michael (U.T.) and Christer,
Regarding the bootstrap effect, in fact, in sim the reduction in THD was from around 0.25% to 0.005% so 50 times, I believe through a combination of raised gain and improved linearity of the diffl stage. Here's why. The output stage is required to swing +/-50V and has a gain at 8 ohms of 25, so, with no V gain in the driver EF, +/-2V is needed at the collector of stage 1. Without bootstrap C's the amp is -
a1 x a2 x a3 = 40 x 1 x 25
and the signal handling differentially across Q1/2 and Q3/4 bases is only +/- 75mV (2 x 26mV + minimal 23mV of Re for matching currents) so, at a gain of 40 times is 66% modulated on Ic generating fairly high THD as it approaches it's signal handling limit. Enter the CC emitter follower driver with Hfe =500 stage and bootstrap C's
C2,C3. These boost the voltage swing on the top of R3, bottom of R5 in an attempt to follow the other end, makingh it's value ----->
infinity. Measured modulation of collector current went, from 66% to 0.18% in sim. The gain obviously went through the roof but Ic modulation went down dramatically improving innate linearity as well. Double benefit! Sim delivered 0.005% THD.
Now to the cascode biassing. I tried a conventional biassing but then, when doing the EF driver I needed 3 x 1W resistors in series for the load R's R11,R19,R12 so, if they were to dissipate equally they would be of equal value and the two nodes between them would be roughly at 1/3 of each supply - ideal for a bias point, so I ran the sim and the THD halved. I fiddled with other ratios and got a further reduction. Wonderful - a ready made bias point.
However, in doing the bode plots, the cascodes were not benefitting from a low source impedance so stage bandwidth was suffering above 1MHz. So I decided to bypass each node to ground with a small C of 100pF -1000pF. There was an optimum restoring bandwidth to over 7MHz with min THD and I am using polystyrene in the kit for C6,7.
R11 x C6 = 3.2uS ( note : R10 is FB resistor in input leg)
Thanks Leolabs,
Meditation..... Ommmmmm
It had to be special - it is to me! Solves many issue, simply and elegantly. Sexy, hey (Mikeks)?
Cheers,
Greg
Hi all. Sorry about the absence - hectic family w'end.
Hi Michael (U.T.) and Christer,
Regarding the bootstrap effect, in fact, in sim the reduction in THD was from around 0.25% to 0.005% so 50 times, I believe through a combination of raised gain and improved linearity of the diffl stage. Here's why. The output stage is required to swing +/-50V and has a gain at 8 ohms of 25, so, with no V gain in the driver EF, +/-2V is needed at the collector of stage 1. Without bootstrap C's the amp is -
a1 x a2 x a3 = 40 x 1 x 25
and the signal handling differentially across Q1/2 and Q3/4 bases is only +/- 75mV (2 x 26mV + minimal 23mV of Re for matching currents) so, at a gain of 40 times is 66% modulated on Ic generating fairly high THD as it approaches it's signal handling limit. Enter the CC emitter follower driver with Hfe =500 stage and bootstrap C's
C2,C3. These boost the voltage swing on the top of R3, bottom of R5 in an attempt to follow the other end, makingh it's value ----->
infinity. Measured modulation of collector current went, from 66% to 0.18% in sim. The gain obviously went through the roof but Ic modulation went down dramatically improving innate linearity as well. Double benefit! Sim delivered 0.005% THD.
Now to the cascode biassing. I tried a conventional biassing but then, when doing the EF driver I needed 3 x 1W resistors in series for the load R's R11,R19,R12 so, if they were to dissipate equally they would be of equal value and the two nodes between them would be roughly at 1/3 of each supply - ideal for a bias point, so I ran the sim and the THD halved. I fiddled with other ratios and got a further reduction. Wonderful - a ready made bias point.
However, in doing the bode plots, the cascodes were not benefitting from a low source impedance so stage bandwidth was suffering above 1MHz. So I decided to bypass each node to ground with a small C of 100pF -1000pF. There was an optimum restoring bandwidth to over 7MHz with min THD and I am using polystyrene in the kit for C6,7.
R11 x C6 = 3.2uS ( note : R10 is FB resistor in input leg)
Thanks Leolabs,
Meditation..... Ommmmmm
It had to be special - it is to me! Solves many issue, simply and elegantly. Sexy, hey (Mikeks)?
Cheers,
Greg
Hi Mastertech,
Dunno, he said that symmetrical amps were 'sexy' so I thought I'll quote him. Next I'll package it with my PS board using 'nude' capacitors and he'll really get excited!
The amp is to be available as a kit within a week and so I try to do my testing on what might be a typical minimum heatsink for the job, say, 0.5deg C/W. I have a few lying around I can clamp it to. They never become dissipated.
I note you're searching for heatsink options - Conrad represents quality and value. You just need to adjust the ratings for 50C rise rather than 80C. If, however you're after extruded maybe Anthony Holton in Tassy can direct you. Most are made in Taiwan and on their various websites. There are also some very quiet fans around as well so maybe a combo...
Cheers,
Greg
Dunno, he said that symmetrical amps were 'sexy' so I thought I'll quote him. Next I'll package it with my PS board using 'nude' capacitors and he'll really get excited!
The amp is to be available as a kit within a week and so I try to do my testing on what might be a typical minimum heatsink for the job, say, 0.5deg C/W. I have a few lying around I can clamp it to. They never become dissipated.
I note you're searching for heatsink options - Conrad represents quality and value. You just need to adjust the ratings for 50C rise rather than 80C. If, however you're after extruded maybe Anthony Holton in Tassy can direct you. Most are made in Taiwan and on their various websites. There are also some very quiet fans around as well so maybe a combo...
Cheers,
Greg
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.