That is not a good solution.
I think I may have a temperature.
Use a thermometer to check.
Oh, yes, I have a raised temperature.
I'll take the thermometer out, I'm sure if I don't know I can't be ill.
That's what you have done with the diodes.
They revealed a problem.
Take them out.
Problem still there, just you can't see it with the measurements you have.
I think I may have a temperature.
Use a thermometer to check.
Oh, yes, I have a raised temperature.
I'll take the thermometer out, I'm sure if I don't know I can't be ill.
That's what you have done with the diodes.
They revealed a problem.
Take them out.
Problem still there, just you can't see it with the measurements you have.
I obviously don't have your experience with amplifier designs but I do have a lot of experience troubleshooting someone else's blunders. I learned early to eliminate what you can to start with, then get it to operate, then add everything back in one at a time. Too many changes at once and you won't see what broke it and you will be led astray. It's called the KISS theory.
This morning I looked back through the thread to see what we worked through to try and get this working and starting about Post #462 we were dealing with the exact same issues. The amp would oscillate as soon as the bias was raised. Sorry to take you guys through the same issues. I thought I had made some progress but looks like it is either a design flaw or the MJL21193/4 transistors just don't like being part of a big triple. I had issues with them on a slewmaster. I am going to bite the bullet and replace them on one board and see what happens.
Thanks, Terry
Thanks, Terry
this is called circuit wisdom......if we can not explain what components are doing in the circuit, then we are better off without it....
Hi Guys
Sorry for the delay getting back - many real world things to attend to...
The circuit has been amended with some new values and two mods.
As stated before, all of this is out of Self's book, so nothing new: upside-down input stage, EF-VAS, single BJT bias generator, output triple with paralleled outputs, and output inclusive compensation (OIC), otherwise known as transitional miller comp (TMC).
Where some things had been listed as "options", everything has been recast as "required" and some options eliminated. I had not used LTspice until the past few months, so can now look at the circuit in a different way. With specified devices in all locations, not groups of equivalents, the performance is pretty good as asymmetric circuits go for high power.
Easy changes include: new feedback and input leak values, new comp values, removal of the first protection circuit form, removed diodes from the feedback shunt cap, some other R value changes.
Harder changes: reference input cascode divider to ground rather than CCS collector, new protection circuit with limit at 22.5A.
The circuit provides single ppm at 20kHz into 8R, sub ppm at 1kHz. Idle is high, typically twice the Oliver guide. Working into 4R THD is a bit higher. Reducing supply voltages also increases THD at a given power level as is usual.
This layout represent only one way to configure such a circuit. As a flat board parallel to the heat sink and the outputs pointing down, the assembly will fit into 3U.
All the files can be downloaded at my site.
Have fun
Sorry for the delay getting back - many real world things to attend to...
The circuit has been amended with some new values and two mods.
As stated before, all of this is out of Self's book, so nothing new: upside-down input stage, EF-VAS, single BJT bias generator, output triple with paralleled outputs, and output inclusive compensation (OIC), otherwise known as transitional miller comp (TMC).
Where some things had been listed as "options", everything has been recast as "required" and some options eliminated. I had not used LTspice until the past few months, so can now look at the circuit in a different way. With specified devices in all locations, not groups of equivalents, the performance is pretty good as asymmetric circuits go for high power.
Easy changes include: new feedback and input leak values, new comp values, removal of the first protection circuit form, removed diodes from the feedback shunt cap, some other R value changes.
Harder changes: reference input cascode divider to ground rather than CCS collector, new protection circuit with limit at 22.5A.
The circuit provides single ppm at 20kHz into 8R, sub ppm at 1kHz. Idle is high, typically twice the Oliver guide. Working into 4R THD is a bit higher. Reducing supply voltages also increases THD at a given power level as is usual.
This layout represent only one way to configure such a circuit. As a flat board parallel to the heat sink and the outputs pointing down, the assembly will fit into 3U.
All the files can be downloaded at my site.
Have fun
I may need new glasses, but it looks like you don't know how to provide a direct link to your brand new schematic?
IPS looks good (Will definitely work) , maybe a few refinements.
Output stage without any local decoupling or base stoppers will
be either unstable or a outright oscillator.
Old leach with low Ft devices , it could work. Not with modern devices.
Can't believe this was not updated.
Honey badger and slewmaster are live examples of these techniques (100%).
OS
Output stage without any local decoupling or base stoppers will
be either unstable or a outright oscillator.
Old leach with low Ft devices , it could work. Not with modern devices.
Can't believe this was not updated.
Honey badger and slewmaster are live examples of these techniques (100%).
OS
Hi Struth
I have to agree with OS here. I built this circuit per the old schematic, using a board that still4given gave me, and I experienced oscillation in the output stage from power up. I cut in some Re's into the output devices and it was then stable with no signal or load. My gut instinct is that it will also need driver Re's and probably driver/pre-driver collector decoupling to pass a signal under all loads, but I did not get this far in my testing.
I really hope you can make provisions for at least base stoppers in your projects (BL20!!) because even if by some miracle you can stabilise it with your combination of devices, I can absolutely guarantee that the majority of DIYers, like Terry and myself, will not be able to make it work without them.
Can you explain why you have removed the anti parallel diodes around the feedback shunt cap? I have successfully used this arrangement in other circuits I've designed/built, because they provide useful protection of the low voltage cap under fault conditions. The problem is that the leakage current of common 1n4148 diodes is too high, drawing base current from the LTP -IN and upsetting the DC balance as a result. I have used 1N914B and FDH300 diodes from Fairchild with very low leakage current and had absolutely no problem with offset.
I like your overload protection circuit; very economical in parts count; I'll have to look at it in more detail.
I have to agree with OS here. I built this circuit per the old schematic, using a board that still4given gave me, and I experienced oscillation in the output stage from power up. I cut in some Re's into the output devices and it was then stable with no signal or load. My gut instinct is that it will also need driver Re's and probably driver/pre-driver collector decoupling to pass a signal under all loads, but I did not get this far in my testing.
I really hope you can make provisions for at least base stoppers in your projects (BL20!!) because even if by some miracle you can stabilise it with your combination of devices, I can absolutely guarantee that the majority of DIYers, like Terry and myself, will not be able to make it work without them.
Can you explain why you have removed the anti parallel diodes around the feedback shunt cap? I have successfully used this arrangement in other circuits I've designed/built, because they provide useful protection of the low voltage cap under fault conditions. The problem is that the leakage current of common 1n4148 diodes is too high, drawing base current from the LTP -IN and upsetting the DC balance as a result. I have used 1N914B and FDH300 diodes from Fairchild with very low leakage current and had absolutely no problem with offset.
I like your overload protection circuit; very economical in parts count; I'll have to look at it in more detail.
OK these are the changes I see that I would need to make to get to the new circuit on the boards I have.
Change
C1 from 100p to 270p
C4 from 4m7 to 10m
C8 from 220p to 330p
C12 from 22p to 52p
C19 from 22p to 150p
R2 from 47k5 to 10k
R11 from 10R to 100R
R16 from 47k5 to 10k
R27 (trimmer) from 1k to 200R
R32 from 1k to 100R
R52 from 100k to 22k
Q1, 2, 3 & 5 from 2n5401 to BC560
Q4 & 6 from 2N5551 to BC550
Cut traces and reroute R52/C10 to ground.
Cut in R41, R42 (348R) between Q1/Q29 and Q2/Q28
Remove D1-4
Remove and jumper R29, R30
Reworking the protection circuit is too difficult on my boards so I'm not going to attempt it.
The other things are doable but my question is, has this circuit been built and tested or just tested in in Ltspice?
It seemed unstable before but now the base stoppers have been removed from the drivers too. I don't mind making the changes if the circuit has been proved but not if I'm just still chasing my tail.
Change
C1 from 100p to 270p
C4 from 4m7 to 10m
C8 from 220p to 330p
C12 from 22p to 52p
C19 from 22p to 150p
R2 from 47k5 to 10k
R11 from 10R to 100R
R16 from 47k5 to 10k
R27 (trimmer) from 1k to 200R
R32 from 1k to 100R
R52 from 100k to 22k
Q1, 2, 3 & 5 from 2n5401 to BC560
Q4 & 6 from 2N5551 to BC550
Cut traces and reroute R52/C10 to ground.
Cut in R41, R42 (348R) between Q1/Q29 and Q2/Q28
Remove D1-4
Remove and jumper R29, R30
Reworking the protection circuit is too difficult on my boards so I'm not going to attempt it.
The other things are doable but my question is, has this circuit been built and tested or just tested in in Ltspice?
It seemed unstable before but now the base stoppers have been removed from the drivers too. I don't mind making the changes if the circuit has been proved but not if I'm just still chasing my tail.
Hi Guys
Everyone has their own experience.
In the amps I've built, I've never used base-stops in the outputs and had no problems with parallel pairs. Similarly it seemed like base-stops for the predriver would be good to have especially with output-referenced protection circuits, but spice showed it added distortion and experience has shown they are not always needed. Since I removed the old protection circuit, the main purpose for the base-stops to be there was gone.
In a similar vein, the clamp for the VAS contributes distortion and causes wied clipping. From other designs I've been working on, the increased Re value works better, so that is done here. Besides, the new protection circuit clamps both the VAS and the current source. Asymmetric R values are required in the asymmetric Thompson-Lin circuit, but symmetric values can be used in symmetric circuits with push-pull VAS.
The diodes across the feedback cap should make no difference to normal function according to Self. I had never used them until I read his book, even though I saw his articles in '95 or so. They are not used in amps designed by people I respect, nor do they use base-stops for the outputs. The idea of base-stops is very sensible but I've never needed them.
The other circuits I've been building that I borrowed ideas from for this one all operate at 100V rails and are fully symmetric. I don't actually like the Thompson-Lin, but it is amenable to many add-ons to boost gain and potentially reduce THD. I did prototype this using the output stage of one of my other amps, which has eight pairs of outputs. Everything works fine. I prefer fully symmetric circuits as they just seem more congenial and hifi, and easier to compensate.
By accident one of the circuits I ended up with uses an EF4. It has amazing low THD even into 1R. EF4s and EF5s were common in the '80s and '90s. Some were hifi, some were just brute force PA amps.
Continuously clipping the amp and blowing tweeters, other speakers and the amp is certainly not part of an enjoyable musical experience, so it is not what I do with my amps. An amp for myself is working into a known load, and I assume that when a hobbyist builds something for themselves that they know what load it will be too. So there seems little reason to bog the circuit down with bandaids to make it work into unknown loads.
Also, it is of limited economic benefit to use supply rails that are only adequate to achieve the power desired. THD will be much lower with higher rails and there is the added benefit of having more transient energy available from the filters. I had a few designs where I wanted the output stage voltage to be as low as would still provide the very low wattage I needed for my system, so I could run it class-A. But the THD would always be lower if I raised the rails a bit, or a bit more. The idle dissipation could be dialled down as rails were increased, which was an interesting verification of observations from others. Leach's first low-TIM amp had adequate rails, which he increased to be ample and reduce THD at the same time.
Have fun
Everyone has their own experience.
In the amps I've built, I've never used base-stops in the outputs and had no problems with parallel pairs. Similarly it seemed like base-stops for the predriver would be good to have especially with output-referenced protection circuits, but spice showed it added distortion and experience has shown they are not always needed. Since I removed the old protection circuit, the main purpose for the base-stops to be there was gone.
In a similar vein, the clamp for the VAS contributes distortion and causes wied clipping. From other designs I've been working on, the increased Re value works better, so that is done here. Besides, the new protection circuit clamps both the VAS and the current source. Asymmetric R values are required in the asymmetric Thompson-Lin circuit, but symmetric values can be used in symmetric circuits with push-pull VAS.
The diodes across the feedback cap should make no difference to normal function according to Self. I had never used them until I read his book, even though I saw his articles in '95 or so. They are not used in amps designed by people I respect, nor do they use base-stops for the outputs. The idea of base-stops is very sensible but I've never needed them.
The other circuits I've been building that I borrowed ideas from for this one all operate at 100V rails and are fully symmetric. I don't actually like the Thompson-Lin, but it is amenable to many add-ons to boost gain and potentially reduce THD. I did prototype this using the output stage of one of my other amps, which has eight pairs of outputs. Everything works fine. I prefer fully symmetric circuits as they just seem more congenial and hifi, and easier to compensate.
By accident one of the circuits I ended up with uses an EF4. It has amazing low THD even into 1R. EF4s and EF5s were common in the '80s and '90s. Some were hifi, some were just brute force PA amps.
Continuously clipping the amp and blowing tweeters, other speakers and the amp is certainly not part of an enjoyable musical experience, so it is not what I do with my amps. An amp for myself is working into a known load, and I assume that when a hobbyist builds something for themselves that they know what load it will be too. So there seems little reason to bog the circuit down with bandaids to make it work into unknown loads.
Also, it is of limited economic benefit to use supply rails that are only adequate to achieve the power desired. THD will be much lower with higher rails and there is the added benefit of having more transient energy available from the filters. I had a few designs where I wanted the output stage voltage to be as low as would still provide the very low wattage I needed for my system, so I could run it class-A. But the THD would always be lower if I raised the rails a bit, or a bit more. The idle dissipation could be dialled down as rails were increased, which was an interesting verification of observations from others. Leach's first low-TIM amp had adequate rails, which he increased to be ample and reduce THD at the same time.
Have fun
Hi Kevin
A lot of what you say resonates with me, and you come across as someone experienced and knowledgeable in these matters.
But I cannot reconcile what you say about the base stops, because my experiences with the EF3 topology is that they are absolutely required for stable operation, for your LTT4 circuit and layout, and others that I've built. I have tried and failed on numerous occasions to build an EF3 without base stoppers, using all different combinations of fast and slow drivers and outputs.
I would love for somebody to prove me wrong by showing me your (built and tested) EF3 circuit but until then I'm going to remain suspicious.
A lot of what you say resonates with me, and you come across as someone experienced and knowledgeable in these matters.
But I cannot reconcile what you say about the base stops, because my experiences with the EF3 topology is that they are absolutely required for stable operation, for your LTT4 circuit and layout, and others that I've built. I have tried and failed on numerous occasions to build an EF3 without base stoppers, using all different combinations of fast and slow drivers and outputs.
I would love for somebody to prove me wrong by showing me your (built and tested) EF3 circuit but until then I'm going to remain suspicious.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.