ULD Extreme
Hugh
No you didn't say it should be changed, but you did appear to agree with John's approach. I am just suggesting that the TO92 combined with a fine tuned resistor value is a better approach if we are trying to keep the LTP in closer balance over a wider temperature range.
I also believe that the current mirror transistors should be replaced by the higher HFE BC550C, which complements the front end modification better. The higher voltage ratings of the BC546B are not required here.
Re the emitter follower Q8. Does everybody consider that this is the most suitable device at that location?
Alex
Hugh
No you didn't say it should be changed, but you did appear to agree with John's approach. I am just suggesting that the TO92 combined with a fine tuned resistor value is a better approach if we are trying to keep the LTP in closer balance over a wider temperature range.
I also believe that the current mirror transistors should be replaced by the higher HFE BC550C, which complements the front end modification better. The higher voltage ratings of the BC546B are not required here.
Re the emitter follower Q8. Does everybody consider that this is the most suitable device at that location?
Alex
'but you did appear to agree with John's approach'
Alex,
I see hundreds of statements on this forum that I do not agree with, but I see no benefit in arguing the toss.
I am permitted, on reflection, to agree with John. He is a very talented designer, actually, and his choices are important for me.
I have used both systems, and noticed very little operational difference, particularly when a CM is used. Neither you nor I are wrong about this; it is just a matter of taste.
Your comments are relative, a matter of opinion, but they are not fixed in stone and many inevitably take a different approach. No designer does it the same way, and this is merely part of the rich tapesty.
I do admit to preferences, just as we all do, but I'm happy to do it the SC way, and it really isn't important. Please, lighten up.
Hugh
Alex,
I see hundreds of statements on this forum that I do not agree with, but I see no benefit in arguing the toss.
I am permitted, on reflection, to agree with John. He is a very talented designer, actually, and his choices are important for me.
I have used both systems, and noticed very little operational difference, particularly when a CM is used. Neither you nor I are wrong about this; it is just a matter of taste.
Your comments are relative, a matter of opinion, but they are not fixed in stone and many inevitably take a different approach. No designer does it the same way, and this is merely part of the rich tapesty.
I do admit to preferences, just as we all do, but I'm happy to do it the SC way, and it really isn't important. Please, lighten up.
Hugh
Hi Carl,
Alex's suggestion for high beta BC550Cs for the LTP current mirror is good, I agree. How about a diode or a resistor instead of the collector/base link across Q4?
I also feel that the 150nF cap AFTER the output inductor should be replaced by a 100nF in series with 10R 1W BEFORE the output inductor. This is normal practice with SS, AB Lin type amps. I must thank Arthur (Phoenix) for pointing this out; I'd missed it.
What say we ask here how many people are interested in buying a couple of the revised pcb at cost should you decide to make one? Get an idea of the demand?
Your pcb design is a beauty, it really is, so much superior to the original.
Hugh
Alex's suggestion for high beta BC550Cs for the LTP current mirror is good, I agree. How about a diode or a resistor instead of the collector/base link across Q4?
I also feel that the 150nF cap AFTER the output inductor should be replaced by a 100nF in series with 10R 1W BEFORE the output inductor. This is normal practice with SS, AB Lin type amps. I must thank Arthur (Phoenix) for pointing this out; I'd missed it.
What say we ask here how many people are interested in buying a couple of the revised pcb at cost should you decide to make one? Get an idea of the demand?
Your pcb design is a beauty, it really is, so much superior to the original.
Hugh
That's exactly what I did. There's a few options here, but a nice thing about the BC550C is that it's easy to get. If we try where possible to use parts available from Jaycar/Altronics, and use the existing PCB, then it makes it easier since folks can order everything from one shop.Carl_Huff said:
jp_howard said:That's exactly what I did. There's a few options here, but a nice thing about the BC550C is that it's easy to get. If we try where possible to use parts available from Jaycar/Altronics, and use the existing PCB, then it makes it easier since folks can order everything from one shop.
I'm a bit wary of the transistor brands carried by Jaycar. Those TO-92 'W''s for example. I would strongly recommend where possible to use devices from reputable man's and even better if you can download datasheets. Fairchild and On-Semi (Motorola) both produce very good TO-92 devices and from my experience the superior specs also translate to better sound.
I'm sure that Sandy K and Hugh will both agree.
AKSA said:
I have also done this in my Extreme SCULD100 and can testify that it's worthwhile.
For those interested, I am currently in the process of developing a current mirror oven which will set the temperature of the mirror to 55C. The idea is not only to stabilise the mirror VCE but also to boost the HFE. Because each side of the mirror is only dissipating 3mW the temperature rise is only about 0.3C above ambient. I have bench tested the idea and was able to get a MPSA18's HFE to rise from about 650 to beyond 900.
Obviously there's no point implementing this unless it leasds to better SQ but I should be able to answer that question this weekend.
BTW this will not get around so called memory effects. Even though the temperature rise may be > 30C this only represents a 10% increase in absolute temperature. (ref -273K)
David,
I've put in a diode on one of my amps, and noticed slightly cleaner top end and better decay. It's subtle, but it's there in AB testing.
The Vf of a 1N4148 is around 480mV at these tiny currents, typically 12uA total for two beta 500 devices in the current mirrors. 390mV is less than this, of course, but it could be argued that the nulling of even order distortion in the LTP occurs when current balance is perfect, rather than the DC voltage balance. When you examine the voltage balance at equal currents with a diode bridge, 3mA apiece, you get 110mV difference in DC balance in over 55V, which is around 0.2%.
The currents flowing under simulation reveal that the AC output of each side of the LTP is very different however. The attachment below shows the two collector outputs for 6mA stage current for 2.1Vp input at 1KHz. Pink is the active output, blue the passive output, and yellow the constant 6mA stage current. If you simulate with a 72K resistor from collector to base of Q4, you do get equal signal amplitudes at the two outputs. But the waveforms are very different because of the current drive to the VAS.
I believe the distortion of the LTP is nulled only when the currents balance, the voltage across the devices is accurate within 0.2%, AND the AC outputs balance.
Cheers,
Hugh
I've put in a diode on one of my amps, and noticed slightly cleaner top end and better decay. It's subtle, but it's there in AB testing.
The Vf of a 1N4148 is around 480mV at these tiny currents, typically 12uA total for two beta 500 devices in the current mirrors. 390mV is less than this, of course, but it could be argued that the nulling of even order distortion in the LTP occurs when current balance is perfect, rather than the DC voltage balance. When you examine the voltage balance at equal currents with a diode bridge, 3mA apiece, you get 110mV difference in DC balance in over 55V, which is around 0.2%.
The currents flowing under simulation reveal that the AC output of each side of the LTP is very different however. The attachment below shows the two collector outputs for 6mA stage current for 2.1Vp input at 1KHz. Pink is the active output, blue the passive output, and yellow the constant 6mA stage current. If you simulate with a 72K resistor from collector to base of Q4, you do get equal signal amplitudes at the two outputs. But the waveforms are very different because of the current drive to the VAS.
I believe the distortion of the LTP is nulled only when the currents balance, the voltage across the devices is accurate within 0.2%, AND the AC outputs balance.
Cheers,
Hugh
Attachments
VHF man said:
Dave, interested. I've been thinking the same thing but wouldn't it be equally as good for the input transistors. I think I read years ago that JC does something to stop the flow of air around the transistors to keep the transistor temperature constant.
I was thinking the transitors should be buried in the main heatsink and internal air flow restricted. The main heatsink would need to be reduced in size the increase temp a little. Poor capacitors.
Alex's amp probably already runs at greater than 55C.
regards
don't remove the 150nF across the speaker terminals, just reduce it to 47nF. Now add a resistor to that output capacitor. Try 5r.
Add an R+C Zobel at the power amp output. Try 8r to 10r + 47nF.
Try reducing the damping resistor across the inductor. Maybe 3r to 5r
At audio frequencies the output Thiele looks like a low impedance inductor feeding an easy speaker load.
At RF frequencies the Thiele looks like the 5r across the speaker terminals + 5r across the inductor ~=10r and the Zobel looks like 10r to ground.
The total VHF load is thus ~5r even if the speaker appears as an infinite impedance load.
Now looking back into the amplifier with the speaker or the speaker cables injecting back emf into the NFB loop.
The terminal mounted Zobel appears as a 5r to VHF and the inductor resistor appears as a 5r followed by the 10r to ground and the //12k going into the NFB and the //impedance of the output stage.
This results in much less interference entering via the speaker terminals and it stabilises the amplifier at HF.
Add an R+C Zobel at the power amp output. Try 8r to 10r + 47nF.
Try reducing the damping resistor across the inductor. Maybe 3r to 5r
At audio frequencies the output Thiele looks like a low impedance inductor feeding an easy speaker load.
At RF frequencies the Thiele looks like the 5r across the speaker terminals + 5r across the inductor ~=10r and the Zobel looks like 10r to ground.
The total VHF load is thus ~5r even if the speaker appears as an infinite impedance load.
Now looking back into the amplifier with the speaker or the speaker cables injecting back emf into the NFB loop.
The terminal mounted Zobel appears as a 5r to VHF and the inductor resistor appears as a 5r followed by the 10r to ground and the //12k going into the NFB and the //impedance of the output stage.
This results in much less interference entering via the speaker terminals and it stabilises the amplifier at HF.
jp_howard said:
JP,
I would recommend a diode because it has a lower dynamic (ac) resistance. However, as Hugh has indicated, the ability of the mirror to balance the tail currents is limited by the VAS loading which increases with frequency - particularly in this design because the 100pf capacitor provides a dominant pole for stability. This feedback around the VAS is a good thing for reducing VAS distortion and because the output impedance is approximatd by XC/Beta + re the VAS impedance can be made low enough to reduce some of the output stage crossover distortion. The aim of the game is to optimise the gain distribution for the best sound quality outcome. This may not simply align with the lowest measured THD and yes this explamation is simplified so there are also other factors involved.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.