DVV-
Is Alex available for other similar projects? I'm always looking for resources.
I have a few suggestions for the amp. I have been building around a similar circuit for um, 35 years now so I have some experience with it.
First big one, the cascode on the input is right for an inverting amp but will compromise the performance of a noninverting amp. What I'm doing is cascoding with Supertex DN2535 depletion fets with the gates tied to the sources of the input pair. It makes a big difference. (Simplifys the circuit a lot as well.)
I would also cascode the current source on the input pair the same way. Again improves the CMRR.
I think it needs a DC offset trim. Even under ideal conditions you will probably have 5 mV of offset if the fets are carefully matched and probably more in practice. The LSK489 might be a better option for the input pair. http://www.linearsystems.com/assets/media/file/datasheets/LSK489.pdf . With a 1 Ohm load offset becomes important. I used a trimmer between the sources of the input pair but the same can be done with a trimmer on one of the load resistors, R2 & R3. You could also use a servo.
I'm not sure I understand the function of R50 and R52.
I would add baker clamps on Q6 and Q36. Saturation on those transistors would be bad.
I would also cascode those devices but that may be overkill for this.
Is there a reason why the load fuse is not inside the feedback? Maybe add the relay as well if possible?
One other trick in protection is to add a relay/optocoupler across the bias network to remove the drive/bias when tripped.
I hope these ideas are useful.
Is Alex available for other similar projects? I'm always looking for resources.
I have a few suggestions for the amp. I have been building around a similar circuit for um, 35 years now so I have some experience with it.
First big one, the cascode on the input is right for an inverting amp but will compromise the performance of a noninverting amp. What I'm doing is cascoding with Supertex DN2535 depletion fets with the gates tied to the sources of the input pair. It makes a big difference. (Simplifys the circuit a lot as well.)
I would also cascode the current source on the input pair the same way. Again improves the CMRR.
I think it needs a DC offset trim. Even under ideal conditions you will probably have 5 mV of offset if the fets are carefully matched and probably more in practice. The LSK489 might be a better option for the input pair. http://www.linearsystems.com/assets/media/file/datasheets/LSK489.pdf . With a 1 Ohm load offset becomes important. I used a trimmer between the sources of the input pair but the same can be done with a trimmer on one of the load resistors, R2 & R3. You could also use a servo.
I'm not sure I understand the function of R50 and R52.
I would add baker clamps on Q6 and Q36. Saturation on those transistors would be bad.
I would also cascode those devices but that may be overkill for this.
Is there a reason why the load fuse is not inside the feedback? Maybe add the relay as well if possible?
One other trick in protection is to add a relay/optocoupler across the bias network to remove the drive/bias when tripped.
I hope these ideas are useful.
I believe so, but best you ask him. I'll look up his e-mail address and PM it to you. The last we did something, of my own accord I dedcided it would be most unfair to give him a piece of work that complex (there was more than just the basic amp) and ask him to do it for free. So I dispatched €200 (app. $ 250) to him via Western Union, because that's what I would have had to pay here at local prices. But Romania is in a bad way, so that amount goes way higher over there. Still, fair is fair, although I'm sure he would have done it for free.
Yes, I would assume so.
Never done that before, but it's worth a try, to be sure. However, not now, not on this very particular project, for which Wayne is footing the bill. I slip slide with new ideas at my own cost, so if I get something wrong, nobody's the worse for it.
Besides, the usual note - you folks don't seem to understand that what you have available doesn't even look like what I have available. What's a matter of phoning in an order to you is mission impossible to me. This limits my ideas as well, I don't have parts so I don't think the same wa as you can.
Look at Version 1 and you'll see I did do it that way intially. I might return to that solution, although I know the new one has a consierably better S/N ratio and extreme stability even under very adverse conditions. Also, it's a matter of the power supply noise and ripple base line; I do use separate power lines, and this one regulated and fairly well filtered, so the base line of trouble with the supply is considerably lower than otherwise.
Another option is to use a very standard CCS, transistor with zener at base, but instead od simply having a resistor to the ground, throw in a FET between base and ground, with a little resistor separateuing them. That kicks the CMRR up by like 15 dB over the resistor only version.
I will have to make up my mind about that by tomorrow night.
But it does have it, always has had it, look towards the bottom end of the picture, between the PSU reg and the protection circuits. I always put it there, so it's not fooling around with the input stage as a part of it, but is a separate circuit.
The oldest Otala trick ever.
What they do is to provide a constant impedance load of the VAS, and to determine its gain. You want more gain, increase their value, you want less, decrease their value. They let you adjust the VAS current to whatever you think is best without letting the gain go wild on you. DAMN useful, trust me.Harman/Kardon has been using that trick like forever, and it works like a charm. Helps make the VAS completely independent of whatever is going on at the output.
That's a thought.
So would I, if I needed greater voltage gains than I do. As you say, that would be art for art's sake in this case.
The reason is that I am not sure it will even be there. Looking back, it seems to me that the usual two fuses in the PSU lines should be quite enough.
And I do use two parallel relays, each rated at 16A continuous and 24A impulse, for overheat and excess DC protection.
I have seen such circuits, but have never worked with them. As I said, I hate experimenting with other people's money, so I will need to first get my bearings with such circuits before I actually use them.
And, I always say this, in my view, H/K Citation 24 power amp I have is the most neutral amp I have ever come across, and it is an AC design, with output relays. Not the last word in circuit design, but obviously a classic done well enough to beat a lot of the newbes.
Thank you for taking the time and trouble to make the list. Invaluable feedback for me, I thrive on it because that's the fastest way to learn I now of.
That would open the feedback loop and test your amp's absolute stability every time. You'd probably have to bypass it with a resistor, which would limit everything but would keep the NFB loop appropriately closed.
Actually, there quite a few takes on this.
Thorsten sent me a proposal of his which is a bit unusual, which does include two relays in the NFB loop, where one relay is a small, say 2A, but very high quality relay, and the other one is a macho 20+A relay. Obviously, the idea is to actually bypass but not eliminate the necessary big relay by including a high quality one. However, this one is not making it into this amp, I have never built a circuit like that, and before I go ahead and use it, I need to get hands-on experience with that circuit,
Plus a few other ideas, some of which seem great, easier to implement, so I'll be looking at that for a day or two, with a very possible, even probable, Current Factory Mk III coming out.
I think it might make Damian (1audio) a little happier.
Thorsten sent me a proposal of his which is a bit unusual, which does include two relays in the NFB loop, where one relay is a small, say 2A, but very high quality relay, and the other one is a macho 20+A relay. Obviously, the idea is to actually bypass but not eliminate the necessary big relay by including a high quality one. However, this one is not making it into this amp, I have never built a circuit like that, and before I go ahead and use it, I need to get hands-on experience with that circuit,
Plus a few other ideas, some of which seem great, easier to implement, so I'll be looking at that for a day or two, with a very possible, even probable, Current Factory Mk III coming out.
I think it might make Damian (1audio) a little happier.
That sounds backwards to me... Contacts open, higher resistance, R2/R1 ratio different, gain increases. At the limit, it's a comparator...
Without giving it a lot of thought I would use the extra contacts on the relay to engage a shunt resistor that lowers the gain to unity. In truth, a good silver high current contact should be pretty linear and not benefit from feedback. Not so with a fuse, which must be thermally non-linear to work.
My only issue with fusing in the supplies and not the speaker is twofold- First, what happens when the plus supply opens but the minus supply is still live? Thinking through and verifying all the failure cases can prevent a smoking speaker to go along with your smoking transistors.
Second, the supply fuses may be too big intrinsically to protect your speaker. For example something like a Lowther may fry its voice coil at 1A continuous and a 1A slow blow fuse is the smallest you can use successfully in the rail. Another set of problems to think through.
My only issue with fusing in the supplies and not the speaker is twofold- First, what happens when the plus supply opens but the minus supply is still live? Thinking through and verifying all the failure cases can prevent a smoking speaker to go along with your smoking transistors.
Second, the supply fuses may be too big intrinsically to protect your speaker. For example something like a Lowther may fry its voice coil at 1A continuous and a 1A slow blow fuse is the smallest you can use successfully in the rail. Another set of problems to think through.
At a very simple hand waving level, consider if the load draws high current and the intrinsic output impedance is high - then the voltage across the load will drop below what it should be, because of the voltage drop across that output impedance. The voltage negative feedback, attempting to correct this error, will then force the amplifier voltage at the output to be higher, to be at the correct level - meaning, from the point of view of the load, that the effective output impedance is much lower.
The output impedance is reduced as a function of the amount of feedback available. If the open loop output impedance is 10 Ohms for example and a 10 Ohm load is attached the output voltage with be reduced by 1/2. If 20 dB of feedback is available then the reduces output going back to the feedback junction would no longer balance against the input so the forward gain is increased internally until the feedback balances. With 20 dB of feedback it means there is 10X of available gain to balance the system so the drive to the output transistors can be increased as much as 10 X. However you run out of the extra drive when the effective internal impedance becomes 1 Ohm. Its a 10X reduction in internal impedance. However it also suggests don't attempt a power amp starting with a 10 Ohm open loop impedance.
I hope this helps. Don't use it this way for a final exam. . .
I hope this helps. Don't use it this way for a final exam. . .
Still has the same problems that the fusing on the rails has.
I may be overly paranoid, comes from making commercial high end amps with the possibility of having to pay for way overpriced drivers that got fried. If you build 1000 expensive boxes you start to see how they fail. Fortunately the care in design held. It seems the failure rate is really low with many old Spectral amps still in service 25 years later.
Damian, that Current Factory amp is a one-off, a 100% Wayne effort.
If it was for me, I'd still use a solid state protection circuit. I've spent many a year on playing with it and I think I've got it down about as good as I know how. Admittedly, I've never tried with a 1 Ohm load circuit, only the usual fare, nominal 8 or 4 Ohm load impedance, capable of going down to 2 Ohms without a hitch in relatively short peaks.
Prehaps it's wrong, but I still believe a solid state switch is far faster and more reliable than any fuse. Perhaps that's because I've seen an awful lot of burnt out outputs stages with their fuses intact.
You work for Spectral?
If it was for me, I'd still use a solid state protection circuit. I've spent many a year on playing with it and I think I've got it down about as good as I know how. Admittedly, I've never tried with a 1 Ohm load circuit, only the usual fare, nominal 8 or 4 Ohm load impedance, capable of going down to 2 Ohms without a hitch in relatively short peaks.
Prehaps it's wrong, but I still believe a solid state switch is far faster and more reliable than any fuse. Perhaps that's because I've seen an awful lot of burnt out outputs stages with their fuses intact.
You work for Spectral?
Last edited:
@ Demian,
I'm not adverse to fuses , down to silver or gold types , at low -z continuous drive, parasitic loses show up fast and are clearly heard. The relay maybe the best way to avoid fuse rolling ..
@DVV,
What you talking about , this could very well be the best one yet , bearing out baby centurion ..
I'm not adverse to fuses , down to silver or gold types , at low -z continuous drive, parasitic loses show up fast and are clearly heard. The relay maybe the best way to avoid fuse rolling ..
@DVV,
What you talking about , this could very well be the best one yet , bearing out baby centurion ..
@ Demian,
I'm not adverse to fuses , down to silver or gold types , at low -z continuous drive, parasitic loses show up fast and are clearly heard. The relay maybe the best way to avoid fuse rolling ..
@DVV,
What you talking about , this could very well be the best one yet , bearing out baby centurion ..
a.wayne, ask dvv to send you a picture of prototype or any POWER amplifier he ever made. That one should be interesting...
OK, you should post yours up too, this way we can compare ......
No he can't. He's insistent on it because he makes other people's designs for money, so he's always on the lookout for something new.
As you may have noticed, I do not communicate with him, I don't have time to waste.
- Status
- Not open for further replies.