jp_howard said:My supposition was based on what the guy at Jaycar told me, and based on the picture on their web site - both which it turns out are completely wrong. So I just wasted over $100 by buying the kits rather than the chips...
Well not totally wrong
I didn't find the NJL's on Jaycar's website, so they may not have them separately yet.......
jp_howard said:
Now time to sit back and enjoy the "mess".
SC Ultra LD Class AB amp
With bias set at 80mA, a smallish heatsink (6" x 3"), ambient around 24C, after half an hour of quiet playing we saw around 60C and 100mA, and with 0.33R emitter resistors quiescent was still climbing very slowly.
With larger heatsinks, say 50% more, we'd see perhaps 45C, some 20C over ambient, and I suspect it would be stable with 0.22R emitter resistors. IMHO it is unwise to stick with 0.1R emitter resistors; this is asking for trouble. Roender's point that a fifth diode might fix the problem is interesting; at 80mA bias we measured only 61mV across the trimmer, and to my knowledge no diode, not even a Schottky, would have a Vf as low as this, and this was with 0.33R emitter resistors. Thermal runaway is not pretty and very damaging - conservative design is the best way to avoid problems. These compensated devices are expensive.
I used a 100MHz CRO to test for oscillation; none. At 1KHz it easily pulled 100Vpp unloaded. However, I noticed the amp was only 1dB down at 250KHz, and I consider this too high. It needs a LP filter on the front, giving around 6dB down at 200KHz, to avoid EMI problems.
I felt the sound quite OK, but would like to hear it in stereo. Perhaps when Jeremy has it all set up!
A pleasant day examining a competent AB amp from SC. My thanks to Alex, Alfred and all those who have chipped in with good advice - and my thanks to Jeremy for the opportunity to see how another designer does it!
Jeremy, thank you sincerely for setting up my Squeezebox. I'm already into Internet Radio, and thank you for your patient expertise!
Cheers,
Hugh
With bias set at 80mA, a smallish heatsink (6" x 3"), ambient around 24C, after half an hour of quiet playing we saw around 60C and 100mA, and with 0.33R emitter resistors quiescent was still climbing very slowly.
With larger heatsinks, say 50% more, we'd see perhaps 45C, some 20C over ambient, and I suspect it would be stable with 0.22R emitter resistors. IMHO it is unwise to stick with 0.1R emitter resistors; this is asking for trouble. Roender's point that a fifth diode might fix the problem is interesting; at 80mA bias we measured only 61mV across the trimmer, and to my knowledge no diode, not even a Schottky, would have a Vf as low as this, and this was with 0.33R emitter resistors. Thermal runaway is not pretty and very damaging - conservative design is the best way to avoid problems. These compensated devices are expensive.
I used a 100MHz CRO to test for oscillation; none. At 1KHz it easily pulled 100Vpp unloaded. However, I noticed the amp was only 1dB down at 250KHz, and I consider this too high. It needs a LP filter on the front, giving around 6dB down at 200KHz, to avoid EMI problems.
I felt the sound quite OK, but would like to hear it in stereo. Perhaps when Jeremy has it all set up!
A pleasant day examining a competent AB amp from SC. My thanks to Alex, Alfred and all those who have chipped in with good advice - and my thanks to Jeremy for the opportunity to see how another designer does it!
Jeremy, thank you sincerely for setting up my Squeezebox. I'm already into Internet Radio, and thank you for your patient expertise!
Cheers,
Hugh
Andrew, Mihai,
This is getting tiresome..... reducing the bias string current to 2-3mA may generate the required bias (a bit over 2.2Volts) but would require a redesign; the CCS supplying the VAS is around 20mA, from memory (don't have the circuit with me).
In theory you are correct, in the context of this particular amplifier your advice would require a re-design. Working within the confines of the existing circuit, and the pcb, five diodes are not practical, though a trimmer certainly is.
Can we move on?
Hugh
This is getting tiresome..... reducing the bias string current to 2-3mA may generate the required bias (a bit over 2.2Volts) but would require a redesign; the CCS supplying the VAS is around 20mA, from memory (don't have the circuit with me).
In theory you are correct, in the context of this particular amplifier your advice would require a re-design. Working within the confines of the existing circuit, and the pcb, five diodes are not practical, though a trimmer certainly is.
Can we move on?
Hugh
Re: SC Ultra LD Class AB amp
Hi Hugh,
This clearly shows that the OPS is highly thermally undercompensated.
Lets do some calculations. We have an OPS with four Vbe junctions, two NJLs and two drivers, each with 2.1mV/K negative thermal coefficient, and a thermal compensation diode string with four diodes at -1.6mV/K TC each, which give as:
4 x (-2.1mV/K) = -8.1mV/K for the serial Vbes
and
4x(-1.6mV/K) = -6.4mV/K for the serial compensatin diode string
total TC = -8.1mv/K - (-6.4mV/K) = -1.7mV/K, which is almost another diode TC
Without the fifth diode the output stage will be undercompensated with -1.7mV/K
BRg,
Mihai
AKSA said:
Hi Hugh,
This clearly shows that the OPS is highly thermally undercompensated.
Lets do some calculations. We have an OPS with four Vbe junctions, two NJLs and two drivers, each with 2.1mV/K negative thermal coefficient, and a thermal compensation diode string with four diodes at -1.6mV/K TC each, which give as:
4 x (-2.1mV/K) = -8.1mV/K for the serial Vbes
and
4x(-1.6mV/K) = -6.4mV/K for the serial compensatin diode string
total TC = -8.1mv/K - (-6.4mV/K) = -1.7mV/K, which is almost another diode TC
Without the fifth diode the output stage will be undercompensated with -1.7mV/K
BRg,
Mihai
Mihai, I understand the math and had done it earlier, realising the fundamental problem. At no stage did I state that your five diode solution was wrong. It's perfectly clear and correct, BUT, it clearly means the output stage should have six transistors so that five diodes can be used.
As I have been at pains to point out, the design is set in concrete with just four devices and a physical layout, and unless another MUR120 device is added, and the current through the string reduced, it will forever be undercompensated.
It really needs a Vbe multiplier. My conclusion: these transistors are NOT suitable for designs of less than three pairs.
Ergo, the design as it stands is not adequate. There. I feel better now, I've said it.
Hugh
As I have been at pains to point out, the design is set in concrete with just four devices and a physical layout, and unless another MUR120 device is added, and the current through the string reduced, it will forever be undercompensated.
It really needs a Vbe multiplier. My conclusion: these transistors are NOT suitable for designs of less than three pairs.
Ergo, the design as it stands is not adequate. There. I feel better now, I've said it.
Hugh
Andrew,
Your behaviour is pedantic and silly. It should be obvious that I have contributed, not hindered, this discussion.
I would suggest that you cannot speak for other members. None of us are elected representatives, not even the mods.
Furthermore, if you wish to explain chapter and verse, be my guest. I'm confident all here will enjoy the journey....
Hugh
Your behaviour is pedantic and silly. It should be obvious that I have contributed, not hindered, this discussion.
I would suggest that you cannot speak for other members. None of us are elected representatives, not even the mods.
Furthermore, if you wish to explain chapter and verse, be my guest. I'm confident all here will enjoy the journey....
Hugh
Well I just looked up the Jaycar catalogue, and was impressed to discover that the thermal resistance of each heatsink is specified. The 200mm HS is over twice as resistant as the 150mm which I'm using, so the thermal runaway issue should improve a lot once I get a better HS.jp_howard said:
I've now listened to all of Julia Fischer playing Tchaikovsky (FLAC/96kHz) and I can say that this amp is already a huge amount better than my old one (Rotel RA-920ax). It would have been a lot cheaper too, had I have not been so inefficient in the build.
Whilst I understand that the design is not perfect, I am finding it a good learning project, with a very pleasing result. The kit is nicely put together and well documented.
Are there better kit amps at a similar or cheaper price, which are readily available? If so, I'd love to try one of them out too and compare. I'm sure each build provides its own lessons... Any suggestions much appreciated (I'm in Australia, so locally available products preferred).
Sandy,
The trimmer is in place, set at this stage to 0R, but the next thing Jeremy will do..... I have suggested he make one change at a time, so as not to cloud where the improvements are to be had.
I match all LTP devices carefully too, but prefer not to use a Current Mirror. It's probably more important to match Vbe on the CM at the design current rather than bother with the beta; since the two arms carry necessarily different currents anyway, one supplying bias to both devices; the other, the input side, supplying bias and drive to the VAS.
Cheers,
Hugh
The trimmer is in place, set at this stage to 0R, but the next thing Jeremy will do..... I have suggested he make one change at a time, so as not to cloud where the improvements are to be had.
I match all LTP devices carefully too, but prefer not to use a Current Mirror. It's probably more important to match Vbe on the CM at the design current rather than bother with the beta; since the two arms carry necessarily different currents anyway, one supplying bias to both devices; the other, the input side, supplying bias and drive to the VAS.
Cheers,
Hugh
Re: SC ULD2
I'll also look through my trannies to get all pairs well matched, and stick a bit of thermal transfer tape between each pair to thermally couple them (they're already very close on the PCB), as per your advice.
No, I'm running as per SC's article (except for the trimpot in series with the emitter resistors, and the change to 0R33 emitter resistors) for a while to get a feel for the 'starting point', prior to making any changes. I have however already fitted the trimpot (set to 0R0), so it'll be extremely easy to experiment with this change.sandyK said:
I'll also look through my trannies to get all pairs well matched, and stick a bit of thermal transfer tape between each pair to thermally couple them (they're already very close on the PCB), as per your advice.
SC ULD2
Hugh
I am afraid we will have to agree to disagree, regarding the current mirror device matching. Both vhfman and myself get great results when using matched CM devices with a very high HFE, as well as matched VBE. Many earlier Google results revealed recommendations for devices with an HFE of around 800 in that location.They have however, now disappeared off the radar.
SandyK
Hugh
I am afraid we will have to agree to disagree, regarding the current mirror device matching. Both vhfman and myself get great results when using matched CM devices with a very high HFE, as well as matched VBE. Many earlier Google results revealed recommendations for devices with an HFE of around 800 in that location.They have however, now disappeared off the radar.
SandyK
Sandy,
We don't disagree; I have had very good results using the CM with added diode, something you told me about , and while I don't feel well matched betas are essential, I think within 10% is good practice, and also feel that matched Vbe is good, and I'm confident that we'd have much more to agree about than not.......
Hugh
We don't disagree; I have had very good results using the CM with added diode, something you told me about
, and while I don't feel well matched betas are essential, I think within 10% is good practice, and also feel that matched Vbe is good, and I'm confident that we'd have much more to agree about than not.......Hugh
If I can be so bold, who is up for a revised PCB that incorporates the suggested changes?
Since the design that appeared in Silicon Chip was based on the 'Blameless' work of Doug Self and this version includes features and improvements not in the published article such as 3 output pairs rather than 2, I submit that it is not covered by copyright. If anyone has a claim it would be Doug Self. I've got a first draft ready for review should anyone be interested.
Since the design that appeared in Silicon Chip was based on the 'Blameless' work of Doug Self and this version includes features and improvements not in the published article such as 3 output pairs rather than 2, I submit that it is not covered by copyright. If anyone has a claim it would be Doug Self. I've got a first draft ready for review should anyone be interested.
An externally hosted image should be here but it was not working when we last tested it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.