I was working on a small gain stage design a while ago. The output transistor suffered from parasitic oscillation when the compensation cap was inserted. This took we a while to figure out as it couldn't be seen on a 20MHz scope and actually seemed to showed up as nothing more than an inexplicable DC-offset.
It was possible to move your hands over the circuit without touching it to make it go away. The cure in the end was a 47R resistor in series with the base.
Hugh, are you refering to the higher slew rate required for a given full power bandwidth and therefore less lag compensation?
It was possible to move your hands over the circuit without touching it to make it go away. The cure in the end was a 47R resistor in series with the base.
Hugh, are you refering to the higher slew rate required for a given full power bandwidth and therefore less lag compensation?
Dave,
As you raise the rail voltage of a power amplifier, stability becomes more difficult to achieve. You need to increase lag compensation to ensure it. This compromises slew rate, which must increase for more powerful amplifiers pro rata. Since the output stage is much larger - more complementary pairs to achieve the higher current outputs - you normally would increase the stage current through the voltage amplifier anyway. This increased stage current lessens the sonic impact somewhat of increased lag compensation, but not completely.
In cases like this it's wise to adopt a two pronged approach to compensation. An effective approach is to use conventional lag compensation between collector and base of the voltage amplifier, but also to insert a 'Linsley Hood phase compensation cap' (so-called because Linsley Hood has strongly advocated its use) between the collector of the voltage amplifier and the feedback node.
The two together work well, since the phase comp cap enables the amp to tolerate capacitive loads very well, keeping it stable by compensating for the inevitable phase shifts in reactive loads. Such a phase comp cap enables lag compensation to be reduced still further, enhancing sonics.
I really do not know why lag comensation has such draconian effect on sonics. It's a mystery to me, although it is likely related to feedback factor. But it certainly does, sucking the 'lifeforce' from recorded music. Amp topologies should be configured to minimise this insidious but necessary 'fixit'. I am beginning to suspect that we should strive to bring the votlage amplifier gain below unity by the pole frequency using other, less sonically damaging means. I've yet to try brute force methods - such as a simple cap from collector to signal ground - but I'm pretty sure it would need to be second or third order to be of any use, and the resulting phase shift might just worsen stability margin anyway. This is a very complex area, not entirely amenable to math, I suspect. The capacitance values are typically 20-100pF, very low, and comparable with pcb parasitics, and this is the reason board layout is critical in the area of the voltage amplifier.
Cheers,
Hugh
As you raise the rail voltage of a power amplifier, stability becomes more difficult to achieve. You need to increase lag compensation to ensure it. This compromises slew rate, which must increase for more powerful amplifiers pro rata. Since the output stage is much larger - more complementary pairs to achieve the higher current outputs - you normally would increase the stage current through the voltage amplifier anyway. This increased stage current lessens the sonic impact somewhat of increased lag compensation, but not completely.
In cases like this it's wise to adopt a two pronged approach to compensation. An effective approach is to use conventional lag compensation between collector and base of the voltage amplifier, but also to insert a 'Linsley Hood phase compensation cap' (so-called because Linsley Hood has strongly advocated its use) between the collector of the voltage amplifier and the feedback node.
The two together work well, since the phase comp cap enables the amp to tolerate capacitive loads very well, keeping it stable by compensating for the inevitable phase shifts in reactive loads. Such a phase comp cap enables lag compensation to be reduced still further, enhancing sonics.
I really do not know why lag comensation has such draconian effect on sonics. It's a mystery to me, although it is likely related to feedback factor. But it certainly does, sucking the 'lifeforce' from recorded music. Amp topologies should be configured to minimise this insidious but necessary 'fixit'. I am beginning to suspect that we should strive to bring the votlage amplifier gain below unity by the pole frequency using other, less sonically damaging means. I've yet to try brute force methods - such as a simple cap from collector to signal ground - but I'm pretty sure it would need to be second or third order to be of any use, and the resulting phase shift might just worsen stability margin anyway. This is a very complex area, not entirely amenable to math, I suspect. The capacitance values are typically 20-100pF, very low, and comparable with pcb parasitics, and this is the reason board layout is critical in the area of the voltage amplifier.
Cheers,
Hugh
I built one of these amps in the early 90's and had nothing but problems. This amp appeared in Australian Electronics Monthly (AEM) and the board layout used looked fantastic but worked like s**t. All the TO126 transistors where bolted to the heatsink on one side of the board and used long traces. The amp oscillated like a bitch and I eventually scrapped it cause I couldnt get it to work. Pity.
Hey whitetrash,
What problems other than oscillation did you have?
Presumably you have a scope to tell if it oscillated.
I am interested in other circuits you have built that have good sound. In fact that is the real reason why i posted in first place but seems to have been missed by everyone.
What problems other than oscillation did you have?
Presumably you have a scope to tell if it oscillated.
I am interested in other circuits you have built that have good sound. In fact that is the real reason why i posted in first place but seems to have been missed by everyone.
Seemed to be a layout problem between the FET stage and Q12/14. The traces seemed to be too long and I even tried rebuilding it with a differnent layout of the FET stage.
Another problem was the board size was too big! For a 2 channel amp, the boards plus transformer wouldnt fit into a case. I ended up sawing the board into 2 pieces and mounting it at 90 deg to reduce the footprint.
Seemed a great design, but the way it was implented was not properly followed through. It didnt seem to last long as a kitset either and was discontinued pretty soon.
I eventually built some Leach amps and redesigned the boards (for TO3 trannies onboard) and another version with TO220 trannies.
Another problem was the board size was too big! For a 2 channel amp, the boards plus transformer wouldnt fit into a case. I ended up sawing the board into 2 pieces and mounting it at 90 deg to reduce the footprint.
Seemed a great design, but the way it was implented was not properly followed through. It didnt seem to last long as a kitset either and was discontinued pretty soon.
I eventually built some Leach amps and redesigned the boards (for TO3 trannies onboard) and another version with TO220 trannies.
whitetrash,
From memory there were wat least two versions of this topology from David Tilbrook. The 6000 I built was the high powered version that had two PCB's per channel. I admit it was large but tall not wide, in fact the amp module when assembled I thought was nice and compact (though tall). the lower powered version(s) which had series numbers like 6010 or something did have single PCB's and Tilbrook did admit the layout was compromised, I just relooked at the PCB and the version i built doesnt seen to have overly long traces. i never built the lower power versions so can't comment. Also whilst i will buy PCB's i will never buy a kitset (for audio), combination of stubborn pride and wanting to know where exactly the pedigree of components, i mean just look at those 'lytics on the shelf at DSE and Jaycar, absolute shite.
More importantly how does the Leach sound? Is it really good, not just "Hifi" good. Which version did you build and what PCB's did you use.
From memory there were wat least two versions of this topology from David Tilbrook. The 6000 I built was the high powered version that had two PCB's per channel. I admit it was large but tall not wide, in fact the amp module when assembled I thought was nice and compact (though tall). the lower powered version(s) which had series numbers like 6010 or something did have single PCB's and Tilbrook did admit the layout was compromised, I just relooked at the PCB and the version i built doesnt seen to have overly long traces. i never built the lower power versions so can't comment. Also whilst i will buy PCB's i will never buy a kitset (for audio), combination of stubborn pride and wanting to know where exactly the pedigree of components, i mean just look at those 'lytics on the shelf at DSE and Jaycar, absolute shite.
More importantly how does the Leach sound? Is it really good, not just "Hifi" good. Which version did you build and what PCB's did you use.
It was the single board design that I built, I think it was an upgrade to the series 5000 amp design or something. After I scrapped it I used the mosfets as the outputs in a redesigned Leach amp. Its hard to compare due to all the noise and crap from the ultra. I think the addition of extra mosfets in the output like you added is a great step. Im only running 2 pairs and the dynamics of the Leach seem to be a little missing. It only seems to put out about 60 watts on 55-0-55 volt rails and I think its due to using the output mosfets. Fairly high distortion too as I ran it on a distortion analyzer here at work and got around 0.09%
Hello PMM,
I am currently using the Leach and have been for the last 3 years. Out of all the amps I have built (many!) it has to be the most civilised sounding, and allows you to concentrate on enjoying the music. Having said that, I would describe the sound as laid back, slightly lacking in dynamics, and very sweet and "giving" in the top end. I too have built the Silicon Chip amplifiers and don't like the sound (to my ears all "Self" topology amps sound similar) - I would prefer the Leach by a large margin.
As for the 6000, I have heard this amplifier driving some Duntech Marquis speakers with good source components and it was the best sound I have ever heard. I also had a friend build one and I borrowed one board to try, and compared the sound of one channel of the Leach amp to one channel of the 6000. The 6000 sounded superior in nearly every aspect - MUCH better dynamics, more detail, more solid bass etc, BUT I didn't like the top end - the leach is nicer on the top end. I believe this is due to the MOSFET output in the 6000 (I don't like MOSFET sound). So in summary, I wouldn't advise you to build a Leach if you are happy with the 6000 - I'm sure you will be dissapointed. I'm planning to design a decent PCB for the 6000 and build that in the next few months (the Tilbrook PCBs were extremely annoying to assemble and, as Whitetrash mentioned, had very long traces (even on the folded high power boards) which probably contributed to a lack of stability in some cases. Also I'm planning to design the board to use 2SJ162/2SK1058 MOSFETS.
As a final (controversial) statement, I think the Tilbrook amps (5000 and 6000) are the only MOSFET amps I have heard that I could enjoy music through.
I am currently using the Leach and have been for the last 3 years. Out of all the amps I have built (many!) it has to be the most civilised sounding, and allows you to concentrate on enjoying the music. Having said that, I would describe the sound as laid back, slightly lacking in dynamics, and very sweet and "giving" in the top end. I too have built the Silicon Chip amplifiers and don't like the sound (to my ears all "Self" topology amps sound similar) - I would prefer the Leach by a large margin.
As for the 6000, I have heard this amplifier driving some Duntech Marquis speakers with good source components and it was the best sound I have ever heard. I also had a friend build one and I borrowed one board to try, and compared the sound of one channel of the Leach amp to one channel of the 6000. The 6000 sounded superior in nearly every aspect - MUCH better dynamics, more detail, more solid bass etc, BUT I didn't like the top end - the leach is nicer on the top end. I believe this is due to the MOSFET output in the 6000 (I don't like MOSFET sound). So in summary, I wouldn't advise you to build a Leach if you are happy with the 6000 - I'm sure you will be dissapointed. I'm planning to design a decent PCB for the 6000 and build that in the next few months (the Tilbrook PCBs were extremely annoying to assemble and, as Whitetrash mentioned, had very long traces (even on the folded high power boards) which probably contributed to a lack of stability in some cases. Also I'm planning to design the board to use 2SJ162/2SK1058 MOSFETS.
As a final (controversial) statement, I think the Tilbrook amps (5000 and 6000) are the only MOSFET amps I have heard that I could enjoy music through.
Using the old Hitachi mosfets in the Leach is my mistake I think. The amps dont like to play loud and sound a little out of there depth on transients. The hitachis have quite a high Ron and using a more modern mosfet would make a big difference. I actually just checked the price of the 2sk1058/2sj162 from Farnell, Aus and they are wanting AU$40/$49 a piece! Tech-diy.com have them for US$27 for 2 matched pairs (4 fets total) With some decent trannies in the Leach I bet it would sing.
For any board redesign for the 6000, this would be a major job. But using decent Fets in the output would also make it a killer.
For any board redesign for the 6000, this would be a major job. But using decent Fets in the output would also make it a killer.
Hi whitetrash,
Yes I think using MOSFETs on the Leach is definitely a mistake. The VAS and driver stages of the Leach are setup to suit bipolar outputs and would probably not drive the capacitance of source follower connected MOSFETs well (depending on where you connected them). Also, the Leach is a "low feedback" design and doesn't have enough open loop gain to allow sufficient feedback to compensate for the non-linearities of MOSFETs at the output (= much higher distortion). And then there's the lack of efficiency as you said caused by the high R-on. Besides, why would you want to spoil a good amp setup for bipolars by using MOSFETs? ;-)
BTW, I used MJL21193/4 bipolars for the output of the Leach - but I doubt that they sound very different to the original MJ15003/4s specified. Why not try the Leach with these (they're pretty cheap - around AU$7 each) - I'm sure you will like it.
The newer Hitachi MOSFETs I mentioned have fairly similar characteristics to the old TO3 ones but have lower g-s capacitance, which should be of benefit. But the main reason for using them is that the TO3s are obsolete and mounting the TO3Ps is so much easier!
Yes the board redesign will be a challenge. Especially in a fully differential design where you want to try and keep the bandwidth of the traces for each side equal...
BTW, Jaycar have the K1058/J162s for about AU$17 each, a bit better than Farnell!
As an aside, a 6000 with bipolar outputs appeals to me....but I doubt whether it could be stabilised with sluggish outputs. Hmmmm....
Yes I think using MOSFETs on the Leach is definitely a mistake. The VAS and driver stages of the Leach are setup to suit bipolar outputs and would probably not drive the capacitance of source follower connected MOSFETs well (depending on where you connected them). Also, the Leach is a "low feedback" design and doesn't have enough open loop gain to allow sufficient feedback to compensate for the non-linearities of MOSFETs at the output (= much higher distortion). And then there's the lack of efficiency as you said caused by the high R-on. Besides, why would you want to spoil a good amp setup for bipolars by using MOSFETs? ;-)
BTW, I used MJL21193/4 bipolars for the output of the Leach - but I doubt that they sound very different to the original MJ15003/4s specified. Why not try the Leach with these (they're pretty cheap - around AU$7 each) - I'm sure you will like it.
The newer Hitachi MOSFETs I mentioned have fairly similar characteristics to the old TO3 ones but have lower g-s capacitance, which should be of benefit. But the main reason for using them is that the TO3s are obsolete and mounting the TO3Ps is so much easier!
Yes the board redesign will be a challenge. Especially in a fully differential design where you want to try and keep the bandwidth of the traces for each side equal...
BTW, Jaycar have the K1058/J162s for about AU$17 each, a bit better than Farnell!
As an aside, a 6000 with bipolar outputs appeals to me....but I doubt whether it could be stabilised with sluggish outputs. Hmmmm....
Hi Owdeo,
would you really trust semi's from Jaycar? Most of the ones I have seen are "clones" or dont have manufactureres markings or markings I dont recognise. I normally steer well clear of DSE or Jaycar 'lytics & semis for audio stuff.
The Leach you built, where did you source the PCB's? (just out of interest), I have a fully working Silicon Chip ULD amp which I built dual mono, however as I said earlier I think it sounds shite. as the power supply rails are 52 volt and the semis probably suitable for a Leach I may be able to knock one up fairly quickly and cheaply (if I could get PCB's easy).
The Naim amp I am considering also would be mostly fresh start cos power supply rails are lower and different parts needed.
would you really trust semi's from Jaycar? Most of the ones I have seen are "clones" or dont have manufactureres markings or markings I dont recognise. I normally steer well clear of DSE or Jaycar 'lytics & semis for audio stuff.
The Leach you built, where did you source the PCB's? (just out of interest), I have a fully working Silicon Chip ULD amp which I built dual mono, however as I said earlier I think it sounds shite. as the power supply rails are 52 volt and the semis probably suitable for a Leach I may be able to knock one up fairly quickly and cheaply (if I could get PCB's easy).
The Naim amp I am considering also would be mostly fresh start cos power supply rails are lower and different parts needed.
Hi PMM,
Like you say, I wouldn't normally trust anything from Jaycar, but in the case of these MOSFETs I have seen them and I'm as sure as I can be without testing them that they are genuine Hitachis. I used to work at Jaycar in my student days and I know that they source their semis from Adilam, so the quality depends on what they buy. I agree - wouldn't touch their 'lytics with a barge pole
I downloaded the gerber files from Leach's site and had the boards made up in 2oz copper by a local PCB maker. I had a whole panel made at the time as several friends also made Leach amps (and all love them). I have a couple spare but I have plans for them (active speakers) otherwise I would offer to let you have a pair.
I would highly recommend getting rid of those Silly-Chip boards and replacing them with Leachs - you can reuse the driver and output devices too. I've never tried a Naim clone, but as it is a "Self topology" I expect it would sound similar to the Silly-chip ULD (though it sounds as if Hugh of AKSA has different opinions on that subject).
Mikek - that is a mild statement compared to some of your other posts. Surely you could extend it to: "using MOSFETs in ANY amp is perverse" ;-)
Like you say, I wouldn't normally trust anything from Jaycar, but in the case of these MOSFETs I have seen them and I'm as sure as I can be without testing them that they are genuine Hitachis. I used to work at Jaycar in my student days and I know that they source their semis from Adilam, so the quality depends on what they buy. I agree - wouldn't touch their 'lytics with a barge pole
I downloaded the gerber files from Leach's site and had the boards made up in 2oz copper by a local PCB maker. I had a whole panel made at the time as several friends also made Leach amps (and all love them). I have a couple spare but I have plans for them (active speakers) otherwise I would offer to let you have a pair.
I would highly recommend getting rid of those Silly-Chip boards and replacing them with Leachs - you can reuse the driver and output devices too. I've never tried a Naim clone, but as it is a "Self topology" I expect it would sound similar to the Silly-chip ULD (though it sounds as if Hugh of AKSA has different opinions on that subject).
Mikek - that is a mild statement compared to some of your other posts. Surely you could extend it to: "using MOSFETs in ANY amp is perverse" ;-)
owdeo said:
Taken in context, my statements elsewhere are reasonably mild......No.....horses for courses....etc.....MOSFETs have their uses....in audio....200W+
An excellent example of such here:
http://www.qscaudio.com/support/library/papers/pl90wp.pdf
and here:
http://www.qscaudio.com/support/library/papers/proaud2.pdf
more good stuff:
http://www.qscaudio.com/support/library/papers/puzzle.pdf
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.