PRR said:
Hi PRR,
Yes thanks, I'll probably get his book at some point. I understand what your suggesting about Class B, but I was able to bias the amps that I built without a problem, and after looking at the Universal Tiger again it's clear that the diodes are best on the output devices. I took another look at the CFP configuration that Self discusses and it is different than the Tiger. It has low value resistors in the collector path of the output devices and that is what provides feedback. It does follow that thermal feedback should be from the drivers in that case.
Universal Tiger Simulation
I did a quick SPICE simulation of the Universal Tiger and it seems, surprisingly, that the phase and gain margins are fine, BUT I want to look at it a bit more. I did not simulate lead inductance and I do not, yet, completely trust the device models.
I've been suspicious that the output devices (MJ4502/802) are undersized for power in light of derating, and safe operating area. I'm surprised that Self uses these devices for examples in his book.
I did a quick SPICE simulation of the Universal Tiger and it seems, surprisingly, that the phase and gain margins are fine, BUT I want to look at it a bit more. I did not simulate lead inductance and I do not, yet, completely trust the device models.
I've been suspicious that the output devices (MJ4502/802) are undersized for power in light of derating, and safe operating area. I'm surprised that Self uses these devices for examples in his book.
Re: Universal Tiger Simulation
Ok... this is pretty much as I would have suspected since Dan Meyer was a good electrical engineer.
The problem in the real world was not the thermal issue and not the SOA of the output devices per se... both have/were addressed by me in various rebuilds, and others have done the same.
The problem with the circuit is that it would self destruct without much provocation. Merely plugging in the input while the amp is/was on could cause it to fry instantly.
I still suspect some nasty positive feedback/parasitic path.
My feeling about this is bolstered by the reports that some people have had theirs running flawlessly for a long, long time.
Perhaps there is even some critical variation in gain/bandwidth for one or more stages that is due entirely to random production differences? Or a critical relationship between stages that causes the nice instantaneous fry...
On the other hand, perhaps as I think someone suggested, the failure is due to insufficient SOA for the *driver* stage(s) where if one device fails, the result is catastrophic - or if one device staturates, the thing locks up and fails. (iirc, some driver stages needed a diode properly placed to prevent them from "latching up"?? - <brain mimics swiss cheese>)
Geez, I really would like to *pinpoint* the true cause before I die...
Hmmm...
_-_-bear
PB2 said:
Ok... this is pretty much as I would have suspected since Dan Meyer was a good electrical engineer.
The problem in the real world was not the thermal issue and not the SOA of the output devices per se... both have/were addressed by me in various rebuilds, and others have done the same.
The problem with the circuit is that it would self destruct without much provocation. Merely plugging in the input while the amp is/was on could cause it to fry instantly.
I still suspect some nasty positive feedback/parasitic path.
My feeling about this is bolstered by the reports that some people have had theirs running flawlessly for a long, long time.
Perhaps there is even some critical variation in gain/bandwidth for one or more stages that is due entirely to random production differences? Or a critical relationship between stages that causes the nice instantaneous fry...
On the other hand, perhaps as I think someone suggested, the failure is due to insufficient SOA for the *driver* stage(s) where if one device fails, the result is catastrophic - or if one device staturates, the thing locks up and fails. (iirc, some driver stages needed a diode properly placed to prevent them from "latching up"?? - <brain mimics swiss cheese>)
Geez, I really would like to *pinpoint* the true cause before I die...
Hmmm...
_-_-bear
OUTPUT SOA
I had been meaning to mention a few things about the output device SOA. The MJ802/4502s are nice parts especially in their day, impressive specs 30A, 200W, 100V. The amp was rated for 8 and 4 ohm loads and I don't think there was any mention one way or the other about short circuit ability in Meyer's early articles, protection was added later. The early reviews said that the amp would handle a short simply blowing the fuse. But if we think about the amp idling with 42V rails nominal perhaps 45V when the power line is above nominal, then we can look at the SOA chart for the MJ802 and see that it handles 2A at 40V (80W), and 1A at 50V (50W), and this is *before* temperature derating and assumes a Tc of 25deg C which is not even a hot summer day. I've seen situations where the heat sinks were too hot to keep a finger on them with these amps. The rail fuse was 2A IIRC and I believe that they'll survive a short circuit when cold, not when hot. As good as the output devices look from a glance at their specs, they're way behind in terms of SOA. Here are the figures for some newer parts:
Part #____ 40V _____ 50V
MJ15003 6A/240W 5A/250W
MJ4281 5A/200W 4A/200W 1 second spec
MJW21196 5A/200W 4A/200W 1 second spec
2N3773 3.5/140W 3A/150W DC
2N3773 5A/200W 4A/200W .5 second spec
These are all well above the 80W and 50W figures for the older devices. I'd double up on outputs or use newer devices if rebuilding an old tiger amp.
I had been meaning to mention a few things about the output device SOA. The MJ802/4502s are nice parts especially in their day, impressive specs 30A, 200W, 100V. The amp was rated for 8 and 4 ohm loads and I don't think there was any mention one way or the other about short circuit ability in Meyer's early articles, protection was added later. The early reviews said that the amp would handle a short simply blowing the fuse. But if we think about the amp idling with 42V rails nominal perhaps 45V when the power line is above nominal, then we can look at the SOA chart for the MJ802 and see that it handles 2A at 40V (80W), and 1A at 50V (50W), and this is *before* temperature derating and assumes a Tc of 25deg C which is not even a hot summer day. I've seen situations where the heat sinks were too hot to keep a finger on them with these amps. The rail fuse was 2A IIRC and I believe that they'll survive a short circuit when cold, not when hot. As good as the output devices look from a glance at their specs, they're way behind in terms of SOA. Here are the figures for some newer parts:
Part #____ 40V _____ 50V
MJ15003 6A/240W 5A/250W
MJ4281 5A/200W 4A/200W 1 second spec
MJW21196 5A/200W 4A/200W 1 second spec
2N3773 3.5/140W 3A/150W DC
2N3773 5A/200W 4A/200W .5 second spec
These are all well above the 80W and 50W figures for the older devices. I'd double up on outputs or use newer devices if rebuilding an old tiger amp.
Re: Re: Universal Tiger Simulation
The simulations are not perfect the output stage seems to stick and have *much* more crossover distortion than other designs. I suspect a problem with the device models and I don't know if I'll have time to debug it.
As far as blowing up goes, the closed loop bandwidth extends at least partly into the AM radio band. Plugging in an input with it powered on, will have the cable act as an antenna for any local radio station. Most power amps would survive but I think this output stage might latch up, burn up those undersized resistors and or perhaps the drivers, which lead to melt down and smoke.
It could also be a grounding problem where the right load and input conditions might have lead to oscillation. I'd probably add current limiting to the power supply, then experiment to see if it oscillates if I had one today to work on.
Simulations also showed that this CFP with gain output stage easily provides very high output current probably due to the drivers having higher Vce to work with in driving the output stage. The old Citation only did 56W into 4 ohms, the Tiger did over 100W into 4 ohms IIRC. I do remember seeing this on the bench with the Universal Tiger. This further supports the need to double up output devices or use newer ones.
Yes, it's been a curiosity of mine to figure out what was going on with these amps for many years.
bear said:
Ok... this is pretty much as I would have suspected since Dan Meyer was a good electrical engineer.
The problem in the real world was not the thermal issue and not the SOA of the output devices per se... both have/were addressed by me in various rebuilds, and others have done the same.
The problem with the circuit is that it would self destruct without much provocation. Merely plugging in the input while the amp is/was on could cause it to fry instantly.
I still suspect some nasty positive feedback/parasitic path.
My feeling about this is bolstered by the reports that some people have had theirs running flawlessly for a long, long time.
Perhaps there is even some critical variation in gain/bandwidth for one or more stages that is due entirely to random production differences? Or a critical relationship between stages that causes the nice instantaneous fry...
On the other hand, perhaps as I think someone suggested, the failure is due to insufficient SOA for the *driver* stage(s) where if one device fails, the result is catastrophic - or if one device staturates, the thing locks up and fails. (iirc, some driver stages needed a diode properly placed to prevent them from "latching up"?? - <brain mimics swiss cheese>)
Geez, I really would like to *pinpoint* the true cause before I die...
The simulations are not perfect the output stage seems to stick and have *much* more crossover distortion than other designs. I suspect a problem with the device models and I don't know if I'll have time to debug it.
As far as blowing up goes, the closed loop bandwidth extends at least partly into the AM radio band. Plugging in an input with it powered on, will have the cable act as an antenna for any local radio station. Most power amps would survive but I think this output stage might latch up, burn up those undersized resistors and or perhaps the drivers, which lead to melt down and smoke.
It could also be a grounding problem where the right load and input conditions might have lead to oscillation. I'd probably add current limiting to the power supply, then experiment to see if it oscillates if I had one today to work on.
Simulations also showed that this CFP with gain output stage easily provides very high output current probably due to the drivers having higher Vce to work with in driving the output stage. The old Citation only did 56W into 4 ohms, the Tiger did over 100W into 4 ohms IIRC. I do remember seeing this on the bench with the Universal Tiger. This further supports the need to double up output devices or use newer ones.
Yes, it's been a curiosity of mine to figure out what was going on with these amps for many years.
When you say "stick" regarding the output stage - you mean latch up? Or?
Also, having one around to work on is tough, since the blow up mode is *instant*. No time to see a parasitic! Blam! gone.
They seem to work fine, and then without notice (no need to be hot) blow up.
I will have to look later this weekend to see what I did the *last* time I had a crack at them... I seem to recall using higher power output devices, maybe even MJ15003s, etc... ? Not sure now.
Someone mentioned the SOA of the drivers being suspect too...
Current limiting might be an option for testing... but kinda voids the point of the amp in application...
Hmmm...
_-_-bear
Also, having one around to work on is tough, since the blow up mode is *instant*. No time to see a parasitic! Blam! gone.
They seem to work fine, and then without notice (no need to be hot) blow up.
I will have to look later this weekend to see what I did the *last* time I had a crack at them... I seem to recall using higher power output devices, maybe even MJ15003s, etc... ? Not sure now.
Someone mentioned the SOA of the drivers being suspect too...
Current limiting might be an option for testing... but kinda voids the point of the amp in application...
Hmmm...
_-_-bear
> The MJ802/4502s are nice parts especially in their day
Darn nice parts, and still nothing to sneeze at. But SOA was just beginning to be understood. I think Dan touches on this in his writings, but it seems to have taken another 20 years before real-world worst-case effects were really understood.
Even a decade ago, 40V/1A was a good SOA point, and big reliable amps used bunches of devices to get enough total SOA to hang together in the field. I was startled to look at what ONsemi sells now: SOA per die is finally up significantly from the 1970s. Self may have bought buckets of 802/4502 pairs and hasn't caught up.
My old rule of thumb is that a pair of fat TO3s is good for around 70 Watts audio, and no more. I have very crude old PA amps running 75W on two 2N3055s and 210 Watts on a six-pack of slightly better devices, fault-free. And more adventurous amps like the Tigers and the orginal Phase Linear that go "fwitzzzz" and die sometimes. So at the 70W level, 802/4502 are still valid parts.
> the Tiger did over 100W into 4 ohms IIRC.
IIRC, the rated power was 90W in 4 ohms, and that was contaminated with power ripple. The transformer was large but apparently not that good, and the filter-caps now look teeny (they sure were big for their day). I don't doubt you could show 100W in 4 ohms before obvious clipping, and they would certainly peak higher on transients. And maybe Dan changed transformers along the way. And of course many people tried bigger caps, and even other power transformers.
Today we can buy five channel amps for the home, 90W in 5 ohms, the whole package not a lot bigger than a pair of Tigers, and probably cooler-running, and generally much more reliable. Dan's heatsinking was a bit artless, and I never understood how that much iron didn't work better than it did. OTOH I remember what amps were before Dan and his generation, and those pioneers paved the way for everything that followed. Both their good examples, and the things they got 99% right that later workers improved on.
As for Tiger Stability/Destruction: I expect to die without proving it, but I bet on layout and parasitic subtleties at RF frequencies that can't easily be modeled on slide-rule or in SPICE.
> When you say "stick" regarding the output stage - you mean latch up?
IIRC (it sure has been some years) if you saturated an output (or driver?) it would not turn off for a long time (many microseconds) after base drive went away. Classic BJT storage effects. So if it clipped, it stayed stuck to the rail well into the next quarter-cycle. I don't know why Tiger topology would be worse than others; I'm not really going to fret over it.
Darn nice parts, and still nothing to sneeze at. But SOA was just beginning to be understood. I think Dan touches on this in his writings, but it seems to have taken another 20 years before real-world worst-case effects were really understood.
Even a decade ago, 40V/1A was a good SOA point, and big reliable amps used bunches of devices to get enough total SOA to hang together in the field. I was startled to look at what ONsemi sells now: SOA per die is finally up significantly from the 1970s. Self may have bought buckets of 802/4502 pairs and hasn't caught up.
My old rule of thumb is that a pair of fat TO3s is good for around 70 Watts audio, and no more. I have very crude old PA amps running 75W on two 2N3055s and 210 Watts on a six-pack of slightly better devices, fault-free. And more adventurous amps like the Tigers and the orginal Phase Linear that go "fwitzzzz" and die sometimes. So at the 70W level, 802/4502 are still valid parts.
> the Tiger did over 100W into 4 ohms IIRC.
IIRC, the rated power was 90W in 4 ohms, and that was contaminated with power ripple. The transformer was large but apparently not that good, and the filter-caps now look teeny (they sure were big for their day). I don't doubt you could show 100W in 4 ohms before obvious clipping, and they would certainly peak higher on transients. And maybe Dan changed transformers along the way. And of course many people tried bigger caps, and even other power transformers.
Today we can buy five channel amps for the home, 90W in 5 ohms, the whole package not a lot bigger than a pair of Tigers, and probably cooler-running, and generally much more reliable. Dan's heatsinking was a bit artless, and I never understood how that much iron didn't work better than it did. OTOH I remember what amps were before Dan and his generation, and those pioneers paved the way for everything that followed. Both their good examples, and the things they got 99% right that later workers improved on.
As for Tiger Stability/Destruction: I expect to die without proving it, but I bet on layout and parasitic subtleties at RF frequencies that can't easily be modeled on slide-rule or in SPICE.
> When you say "stick" regarding the output stage - you mean latch up?
IIRC (it sure has been some years) if you saturated an output (or driver?) it would not turn off for a long time (many microseconds) after base drive went away. Classic BJT storage effects. So if it clipped, it stayed stuck to the rail well into the next quarter-cycle. I don't know why Tiger topology would be worse than others; I'm not really going to fret over it.
CFP is subject to power supply 'shoot through' when driven into clipping.
A Baker Clamp fixes this.
Come to think of it, this may be what ails the Threshold S150/300/500 designs too. The amplifier sounds great 'till it clips, and then gets real nasty, right now.
The original 175 was 125W4R, newer versions had much smaller iron and were only rated at 60W8R/60W4R.
The caps were the biggest in the Mallory FP series, but needed to be about 2.5X bigger (IMO).
A Baker Clamp fixes this.
Come to think of it, this may be what ails the Threshold S150/300/500 designs too. The amplifier sounds great 'till it clips, and then gets real nasty, right now.
The original 175 was 125W4R, newer versions had much smaller iron and were only rated at 60W8R/60W4R.
The caps were the biggest in the Mallory FP series, but needed to be about 2.5X bigger (IMO).
"I've updated one tiger to use 10,000uF PS caps."
That's what I had in mind too.
" I've read other posts (like the Hafler mod/updates) to bypass these PS caps with non-polarlized 47uF (ala Leach). Does this still apply? "
Yes.
"22uF better? It seems like it should still apply."
Subjectively , the value of the main filter cap has less to do with the sound, and the bypass cap more to do with the sound.
I use 22µF per rail on amps with dual ±V supplies, and 47µF per rail with amps with just one ±V supply.
As the value is decreased towards 10µF the bass becomes 'thick' sounding again, as the value in increased towards 100µF it becomes a bit 'hollow' (perhaps needing another smaller value).
I used some 200µF Solen in parallel with 38,000µF electrolytics for an amp driving a one ohm load, and was very happy with the sound (if you go backwards, this would be a ratio like 100µF/20,000µF driving 2 ohms, or 50µF/10,000µF driving 4 ohms, or 25µF/5,000µF driving 8 ohms) and right in line with a typical 47µF/10,000µF mono supply driving two 8 ohm loads.
Most of the time I am too cheap to use the film caps, they are too big to fit, and they must be checked for oscillations.
One of the best sounding amps I have heard is a PA amplifier with a shared pair of 10,000µF for the main rails, a pair of 10µF bypass per channel (four total), and a pair of 0.47µF film per channel (four total) with 10 ohms in series with each cap to stop oscillation (these are mounted right on the output boards).
That's what I had in mind too.
" I've read other posts (like the Hafler mod/updates) to bypass these PS caps with non-polarlized 47uF (ala Leach). Does this still apply? "
Yes.
"22uF better? It seems like it should still apply."
Subjectively , the value of the main filter cap has less to do with the sound, and the bypass cap more to do with the sound.
I use 22µF per rail on amps with dual ±V supplies, and 47µF per rail with amps with just one ±V supply.
As the value is decreased towards 10µF the bass becomes 'thick' sounding again, as the value in increased towards 100µF it becomes a bit 'hollow' (perhaps needing another smaller value).
I used some 200µF Solen in parallel with 38,000µF electrolytics for an amp driving a one ohm load, and was very happy with the sound (if you go backwards, this would be a ratio like 100µF/20,000µF driving 2 ohms, or 50µF/10,000µF driving 4 ohms, or 25µF/5,000µF driving 8 ohms) and right in line with a typical 47µF/10,000µF mono supply driving two 8 ohm loads.
Most of the time I am too cheap to use the film caps, they are too big to fit, and they must be checked for oscillations.
One of the best sounding amps I have heard is a PA amplifier with a shared pair of 10,000µF for the main rails, a pair of 10µF bypass per channel (four total), and a pair of 0.47µF film per channel (four total) with 10 ohms in series with each cap to stop oscillation (these are mounted right on the output boards).
bear said:
Stick seems to be the term used in audio to describe the slower turn off of BJTs especially when they saturate. The more technical term is mutual conduction which is often used in digital TTL electronics as is seen in high power totem pole outputs - I have a lot of experience with this. A clamp to keep the output stage out of saturation was the basis for the original Schottky-S high speed logic family and also the advanced families Fairchild-F and TI-AS.
Someone also mentioned a baker clamp, I've rarely if ever seen one in the output stage of a power amp but have thought about using one. The JE990 uses a SI diode clamp on the VAS and I think it's a good idea.
I used the term latch-up loosely, there's no obvious way that the circuit can "latch" but I used it because as you say once this amp takes off it seems to go into self distruct mode and up in smoke fast.
I didn't spend a lot of time on the simulation, but as I remember, the "sticking" was more than just slow turn off. The output stage seemed to overshoot and ring on turn off, then the feedback would notice that the output was not following the input and turn the devices back on to try to stop the ringing. I saw the feedback driving the output devices back on in an attempt to stop the ringing with peak mutual conduction currents of 1.5 amps during this ringing period. I don't fully trust the transistor models and don't know if I'll have time to investigate it. I do remember the output stage running hotter than I thought it should in the real amp.
The crossover distortion was nearly covered by the feedback, but when I reduced the feedback there was even ringing as it went through the crossover region, just looked strange. Again I think it's partly the device models.
I mentioned driver SOA.
Current limiting in the power supply feed would be just to find the problem without destroying output devices, then correct it, then remove the current limiting.
Pete B.
PRR said:
Interesting points PRR, I like your theory about Self's use of 802/4502s, but I would use a more modern example given the differences if I was writing a book.
I agree my rule of thumb is 60 to 70W for an 802/4502 pair, and I would use them today if their SOA was better.
I substituted heat sinks with nearly twice the area and I agree he should have used more.
I agree that there's probably a layout/parasitic issue with the design. I like to take a first cut in SPICE, then check for agreement in the lab to determine if the model needs fine tuning but I don't have an original Tiger. SPICE is very accurate with good device models and just a bit more detail in the circuit model. It's only as good as the device and circuit models.
I'm in no rush to solve the mystery but it would be nice to find out some day.
Pete B.
rickl said:
I've always found bigger caps to improve/clean up the sound of an amp with unregulated supplies when driven hard. Amps with unregulated supplies have more dynamic headroom, and bigger caps reduce the amount of sag and ripple in the rails. More peak output is available before clipping.
Bypassing is complicated, large caps are series resonant circuits and usually become inductive in the audio range. A smaller cap with lower series inductance is simply another resonant circuit in parallel with the larger cap. It's complicated.
http://www.diyaudio.com/forums/attachment.php?postid=243000&stamp=1064845527
A clamp someone has added to a CFP stage.
A clamp someone has added to a CFP stage.
SWTPC Tiger
I need to find out why my Tigers have Q1 through Q11 instead of the 9 transistors shown in most schematics I've seen posted.
Q1, Q2, Q9 are TO-92
Q3, Q4, Q5, Q6, Q10, Q11 are TO-39
Q7 and Q8 are TO-3 output transistors on the heat sinks.
Does anyone know if there might have been a different version of the Universal Tiger using the two extra transistors?
I need to find out why my Tigers have Q1 through Q11 instead of the 9 transistors shown in most schematics I've seen posted.
Q1, Q2, Q9 are TO-92
Q3, Q4, Q5, Q6, Q10, Q11 are TO-39
Q7 and Q8 are TO-3 output transistors on the heat sinks.
Does anyone know if there might have been a different version of the Universal Tiger using the two extra transistors?
I unfortunately cleaned out my basement a couple of years ago and through away a complete set of building instructions and old Audio Amateur and Popular Electronics articles on the amps from between 70' and 73'.
I am sure it was called "Universal Tiger 100" on the instruction set. What did the ".01" represent on the Tiger 01?
If you happen to come across anything else that might shed some light on it, let me know.
Allen
I am sure it was called "Universal Tiger 100" on the instruction set. What did the ".01" represent on the Tiger 01?
If you happen to come across anything else that might shed some light on it, let me know.
Allen
What you have there is the Universal Tiger MKII (this was also called the 175-A which I think was the number used when ordering the kit) which added two transistors for output stage protection. They decided this was needed and started shipping the new design and discontinued the old design. I think it was some time later when it was written about in the Audio Amateur where a few other points were covered. Upsizing those resistors to 2W is an important upgrade.
I wrote:
"Optimizing the Tiger 175-A" Audio Amateur 3 - 1972
This article covered minor circuit changes to add output protection, and I believe minor bias changes, better diff amp transistors, and increased R12 and R14 to 2W, this was the Universal Tiger MKII version."
I wrote:
"Optimizing the Tiger 175-A" Audio Amateur 3 - 1972
This article covered minor circuit changes to add output protection, and I believe minor bias changes, better diff amp transistors, and increased R12 and R14 to 2W, this was the Universal Tiger MKII version."
Allen Bartee said: