Wayne,
It may be within specs (barely). But I would not want it to run that hot, and would probably assume that there was some problem that was causing it, which should be fixed. Andrew beat me to it, but I like that same 'rule of thumb' (or 'rule of finger'?).
If it's not just the lack of a heatsink with a low-enough thermal resistance, then it might be oscillating at a higher-than-audio frequency. If you can borrow the use of an oscilloscope (or maybe even just a frequency counter), then you will know. Have you tried a better heatsink, yet?
I wouldn't take much comfort from the fact that the original 47 Labs Gaincard also got hot. That design had major flaws.
It may be within specs (barely). But I would not want it to run that hot, and would probably assume that there was some problem that was causing it, which should be fixed. Andrew beat me to it, but I like that same 'rule of thumb' (or 'rule of finger'?).
If it's not just the lack of a heatsink with a low-enough thermal resistance, then it might be oscillating at a higher-than-audio frequency. If you can borrow the use of an oscilloscope (or maybe even just a frequency counter), then you will know. Have you tried a better heatsink, yet?
I wouldn't take much comfort from the fact that the original 47 Labs Gaincard also got hot. That design had major flaws.
Wow! That's even hotter than mine. Ah, but I usually have (cause) a reason why there's heat, and mine are cool at idle.
If you used Radio Shack parts, the Nichicon and film caps are okay, but the Xicon (baby blue electrolytics) are undersize models--asking for trouble.
Say, there's 0.1uf polyester and 0.22uf polyester (built-in resistance--high esr="gentle" effects for zobel that's very pleasant).
There's two really good low-esr film caps there. Little blue is 0.1uf almost as good as polypro on power cap bypass. Big blue is 1uf that would give Auricap a run for it if you need input filter caps (DC blockers), although some people use them in pairs for 2uf.
What is the ground loop in the photo? I see three ground wires touching the chassis. That's fairly unusual.
Have you measured the DC output at the speaker terminals?
One cause for heat is DC going in ground loops, and the other cause that I can think of. . . is DC going into speakers.
If you used Radio Shack parts, the Nichicon and film caps are okay, but the Xicon (baby blue electrolytics) are undersize models--asking for trouble.
Say, there's 0.1uf polyester and 0.22uf polyester (built-in resistance--high esr="gentle" effects for zobel that's very pleasant).
There's two really good low-esr film caps there. Little blue is 0.1uf almost as good as polypro on power cap bypass. Big blue is 1uf that would give Auricap a run for it if you need input filter caps (DC blockers), although some people use them in pairs for 2uf.
What is the ground loop in the photo? I see three ground wires touching the chassis. That's fairly unusual.
Have you measured the DC output at the speaker terminals?
One cause for heat is DC going in ground loops, and the other cause that I can think of. . . is DC going into speakers.
Ok, tonight I pulled the heatsinks and added some Arctic Silver heatsink compound....and ran it for about 4 hours at low/mid level volume....so far so good....it is warm to the touch. I took a reading and it only hit about 70f.
I ran it at idle for about 15-20 minutes and it is still dead silent (well if you stick your ear into the woofer you might be able to hear something).
I am going increase the volume and take another reading in about and hour. I will report back in the morning.
I should have the use of a scope this weekend.
Wayne
I ran it at idle for about 15-20 minutes and it is still dead silent (well if you stick your ear into the woofer you might be able to hear something).
I am going increase the volume and take another reading in about and hour. I will report back in the morning.
I should have the use of a scope this weekend.
Wayne
wboyd said:
That sounds like very good news!
(But...) What are you using to measure the temperature? Because, at 70 F, and 98.6F or so body temp, it seems like it should not feel warm to the touch.
So, if you are using some type of contact thermometer, is it possible that it is cooling the point of contact, that much? Or, if you are using a remote infrared sensor type of thermometer, is it possible that it's not focusing narrowly-enough, or was improperly boresighted, and is 'averaging' the chip's temp with some of the surroundings, or reading some part of the surrounding area instead?
I'm glad to hear that you will have the use of a scope. Of course, if your chip is now really at 70 F, or even any reasonable temperature, there might not be any reason to suspect significant oscillation, after all.
[You might already know all of the following. And, (disclaimer: ) some of it might be wrong, and some of it I've never even tried. But I thought it might be worth mentioning some of it, just in case:]
Although it might not matter too much, for this type of measurement, remember that for accurately seeing high frequencies with a scope probe, you need to use the shortest-possible probe ground connector (and a 10X probe, for high impedance). Ideally, it would be a clip-on 'prong' type probe ground, instead of the lead-and-alligator-clip type. But, probably (hopefully), if there is an oscillation and it is large-enough, none of that will matter, all that much.
While you're at it, it might also be 'interesting' to probe the power rails, et al, while a signal is being amplified (e.g. across the bypass caps near the chips). You might even be able to use the scope's probe-calibration squarewave output as the amplifier's input. (Oh: Also remember to calibrate the probe, with the probe-cal squarewave, adjusting the probe's compensation trimmer for the squarest square wave, before starting any measurements. I was well-bitten by not doing that, once, to the point of never forgetting.)
Also, for any conductors with very high frequencies on them, you can usually use the probe tip as an antenna, if you suspect that the probe, itself, kills some oscillation that you are interested in seeing. You just connect the probe's ground lead like normal, and then hover the probe tip near suspected oscillation carriers. Although that obviously wouldn't work too well for any accurate amplitude-type measurements, you can often see the waveform's shape, and measure its frequency, pretty well, that way, without altering the circuit's behavior by touching the probe to it.
That reminds me: If using a clip-lead probe ground, don't assume that just any ground point will do. Use the closest ground that is a ground reference point for whatever you're trying to display.
It might also be interesting to see if you can display any 'ground bounce' voltage, while amplifying a signal. A good one to check, just to see whether or not it's even possible, might be between a speaker output's ground terminal and the main star ground point (with a speaker in use, the louder the better). You could also try looking between the chip's bypass caps' grounds and the star ground point. If you can see anything for those, you could try to look at the actual critical ones, such as between a chipamp input's ground reference (e.g. input resistor's ground) and the star ground point, although the amplitudes there should be MUCH smaller, and might (hopefully!) be extremely difficult to detect, especially without something like a good differential probe, or an active FET-input probe.
With a really-good high-bandwidth scope, you might even be able to check to see if you can detect any RF across the chipamp input resistors (or, better, from chip input pin to nearby input ground reference), by, for example, setting the scope's sweep rate to try to catch the strongest local FM or AM station's carrier frequency. It would be small and probably indistinct; might be difficult to see, really. But you might be able to see it _diminish_ if you parallel the input R with a few hundred pF cap, or hold the cap between the input pin and input ground (if it doesn't already have an RF filter cap). [Don't try that if it's an inverting input, unless you just want to possibly see what oscillation might look like, at the output.]
As always, be very CAREFUL, when playing with metal objects (such as probes etc), not to mention your hands, in and near a working circuit. Besides the very real possibility of killing yourself, I have also found that it's a VERY easy way to destroy things in the live circuit that you're working on. One small slip with a probe tip or ground lead and... Zzzzt!! For your own health and safety, though, one good rule of thumb, when lethal power is present, in addition to removing watches and jewelry etc and having someone else nearby, is to always keep one hand in your pocket. If necessary, set up 'hands-free' test connections with no power even connected (not just switched off), and only then connect power to the circuit under test.
I'm sorry this message got so long.
Have fun.
Ah, Silver. That's old-school stuff and very electrically conductive.
What its supposed to do:
Its supposed to get the heatsink as hot as possible so that the chip as as cool as possible.
Since that's not what happened. . .
What it may do:
Well, Ceramique is thermal paste. Silver is thermo-electric paste, so it can actually fetch signals out from behind plastic and drive them to ground with a capacitance effect.
I have no problem with this being potentially beneficial.
However, I'm very concerned if a small capacitance (with one heck of a lot of resistance) between V- and ground is cooling off the chip. . . Maybe you should review the wiring and layout a bit, to compare your power layouts to examples here at diyaudio. . . so maybe you can use capacitance in a more predictable location.
Of course, I could be entirely mistaken about this scenerio. . .
Because if you also put Silver between your heat spreaders and chassis, then you just turned the whole thing into a nicely grounded, very large heatsink.
What its supposed to do:
Its supposed to get the heatsink as hot as possible so that the chip as as cool as possible.
Since that's not what happened. . .
What it may do:
Well, Ceramique is thermal paste. Silver is thermo-electric paste, so it can actually fetch signals out from behind plastic and drive them to ground with a capacitance effect.
I have no problem with this being potentially beneficial.
However, I'm very concerned if a small capacitance (with one heck of a lot of resistance) between V- and ground is cooling off the chip. . . Maybe you should review the wiring and layout a bit, to compare your power layouts to examples here at diyaudio. . . so maybe you can use capacitance in a more predictable location.
Of course, I could be entirely mistaken about this scenerio. . .
Because if you also put Silver between your heat spreaders and chassis, then you just turned the whole thing into a nicely grounded, very large heatsink.
Daniel,
I should have added that I am using Ceramique by Arctic Silver...or something like that - the exact name on the lable escapes me write now. I aquired several tubes from an electronics engineer when I was working on a project for Raytheon Systems a couple of years ago. I have both types of compound....they are in small-ish syringes. They had boxes of samples.
Tom;
And thank you so much for the scope guidance...my scoping skills are limited but I have used one before. I think that is something that I should eventually invest in...got any recommendations? I would be looking for something used, modestly priced, but able to analyze basic curcuits....probably $100 - $500 range. I will do some research around the forums here as well...I am sure that there is wealth of info floating around.
Also a big thanks to all the others that have chimed in!
Here is an update on the clone;
Had a few friends over last night and played the amp at about 3/4 volume....it was warm to the touch and after taking it's temperature it was about 130f. And by the way tom I am using a contact theremometer and testing several places on the heatsinks (close to the chip) and reporting the highest temp. It sustained this temp from about an hour after turn on untill my last test approx 4-5 hours later...so I ran it for maybe 5 to 6 hours last night. Sounds great...as a matter of fact I got alot of compliments on it.
By the way....both sides are pretty much the same temp....does this mean that it could be a power supply related issue, or just my stupidity in adding the larger caps on the rails?
Thanks,
Wayne-o
I should have added that I am using Ceramique by Arctic Silver...or something like that - the exact name on the lable escapes me write now. I aquired several tubes from an electronics engineer when I was working on a project for Raytheon Systems a couple of years ago. I have both types of compound....they are in small-ish syringes. They had boxes of samples.
Tom;
And thank you so much for the scope guidance...my scoping skills are limited but I have used one before. I think that is something that I should eventually invest in...got any recommendations? I would be looking for something used, modestly priced, but able to analyze basic curcuits....probably $100 - $500 range. I will do some research around the forums here as well...I am sure that there is wealth of info floating around.
Also a big thanks to all the others that have chimed in!
Here is an update on the clone;
Had a few friends over last night and played the amp at about 3/4 volume....it was warm to the touch and after taking it's temperature it was about 130f. And by the way tom I am using a contact theremometer and testing several places on the heatsinks (close to the chip) and reporting the highest temp. It sustained this temp from about an hour after turn on untill my last test approx 4-5 hours later...so I ran it for maybe 5 to 6 hours last night. Sounds great...as a matter of fact I got alot of compliments on it.
By the way....both sides are pretty much the same temp....does this mean that it could be a power supply related issue, or just my stupidity in adding the larger caps on the rails?
Thanks,
Wayne-o
Yep Daniel...."Rock'n da house"....a little Jazz, some Rock, etc.
Maybe it is ok and I am just crazy. Most of my reasoning for originally posting was the fact that most poeple have stated that thier gainclones run quite cool....and yes in the beginning my was way to freaking hot....but it seems much better now after some light rewiring and and some thermal paste. I will most likely still go with Tom's suggestion of scoping it, also replacing the 0.1uf poly caps with some ceramics (makes sense).
By the way...I hooked it up to my modified Philips SACD 1000 last night...singing like sweet siren! Still driving the BR-1's, although I want to try it with my Fostex back loaded horns.....
Wayne-o
Maybe it is ok and I am just crazy. Most of my reasoning for originally posting was the fact that most poeple have stated that thier gainclones run quite cool....and yes in the beginning my was way to freaking hot....but it seems much better now after some light rewiring and and some thermal paste. I will most likely still go with Tom's suggestion of scoping it, also replacing the 0.1uf poly caps with some ceramics (makes sense).
By the way...I hooked it up to my modified Philips SACD 1000 last night...singing like sweet siren! Still driving the BR-1's, although I want to try it with my Fostex back loaded horns.....
Wayne-o
Arctic silver thermal paste is a silver colloid in silicone oil - it's only slightly conductive at low voltages. I don't understand what you were trying to say about capacitance and thermal paste. If your package is insulated it doesn't matter, and if it isn't you can't rely on the paste for insulation anyways.danielwritesbac said:
Boutique thermal compounds are not worth the money, especially for low performance applications like chip amps (heat per unit area is small). Cheap silicone paste from your electronics supplier (I pay ZAR5, about $0.70 for a 10g tube) works just as well and is much cheaper and easier to get. When using thermal paste, "less is better" is always a good motto.
wboyd - have you had access to a high speed oscilliscope yet? This still sounds like an HF oscillation issue. My 20MHz scope couldn't see the oscillation on the last chipamp I built - it was up at 28MHz. Also, use a 10x probe - my HF oscillator wouldn't oscillate when plugged into a 1x probe.
Your description of the temperature for 31v on those chips, is quite a decent result.
I didn't quite catch the exact location of the ceramic caps, but they can be surprisingly good. Hey Tom!! Where do those go?
The inexpensive paste, GC Waldham Type 44 works like Artic Ceramique. Both are long lasting (doesn't turn to gravel), nonconductive, white color, work well applied properly or a messy glob, and leaves no doubt.
While it was necessary to leave no doubt, I also believe that it was the rewiring that cooled off this particular project.
I'm so glad that you're enjoying your new amp.
I didn't quite catch the exact location of the ceramic caps, but they can be surprisingly good. Hey Tom!! Where do those go?
The inexpensive paste, GC Waldham Type 44 works like Artic Ceramique. Both are long lasting (doesn't turn to gravel), nonconductive, white color, work well applied properly or a messy glob, and leaves no doubt.
While it was necessary to leave no doubt, I also believe that it was the rewiring that cooled off this particular project.
I'm so glad that you're enjoying your new amp.
Hi Daniel,
Aw shucks. Me just simple, unfrozen caveman engineer, not familiar with your modern ways.
But you're just yanking my chain, aren't you(?), because you already know that I think that a millimeter away from the power pin is too far away.
;-)
Seriously, though, at one or more centimeters from the pin, you almost might as well not even bother connecting a small bypass cap.
And for ceramic bypass caps, you should use the X7R type, rather than the usually-much-better NPO or C0G type, since the X7R dielectric is much more lossy (think of it as almost a built-in snubber) and therefore much less prone to excite any unwanted CLC 'tuned circuit' resonances (with the ever-present parasitic inductances), which might otherwise result in high-frequency ringing or even outright oscillation. [Film caps are risky, there, for that reason. When they're good, they're very good. But they're actually 'too good', for bypassing, sometimes, and it's very difficult for most people (me included) to know in advance if they will be a problem or not. If film is all I could get, I might want to place at least a very small resistance in series with it (i.e. 'snubberize' it, sort-of), to try to pre-empt any high-frequency misbehavior.]
I tend to also want to solder a 0.1 uF or 0.22 uF X7R ceramic directly from one power pin to the other (soldered right onto the two pins, with leads as short as possible), since, in many amplifier topologies, that is an important current-flow/return path, and its high-frequency bypassing should be by the shortest-possible route, which some (most) layouts might not facilitate very well, since the bypassed frequencies would need to go through one power pin's separate bypass cap and all the way to ground and then back to the other power pin's bypass cap and then to the other power pin (being tortured and disfigured by parasitic inductance and resistance, all the way). Note, too, that in that case, it's even worse than that, since the two small bypass caps would appear to be in series, which halves their effective/total value, as seen for that bypass path.
For anyone interested in actually trying to determine what is better and what is worse, in bypassing and connection strategies for example, LTspice can give you at least a relative picture, IF you ALSO include estimates of the parasitic impedances of both the components and the conductors (e.g. the resistances and inductances of the PCB traces and wires). You can download everything you need for a good start, here: http://www.fullnet.com/~tomg/gooteesp.htm .
Aw shucks. Me just simple, unfrozen caveman engineer, not familiar with your modern ways.
But you're just yanking my chain, aren't you(?), because you already know that I think that a millimeter away from the power pin is too far away.
;-)
Seriously, though, at one or more centimeters from the pin, you almost might as well not even bother connecting a small bypass cap.
And for ceramic bypass caps, you should use the X7R type, rather than the usually-much-better NPO or C0G type, since the X7R dielectric is much more lossy (think of it as almost a built-in snubber) and therefore much less prone to excite any unwanted CLC 'tuned circuit' resonances (with the ever-present parasitic inductances), which might otherwise result in high-frequency ringing or even outright oscillation. [Film caps are risky, there, for that reason. When they're good, they're very good. But they're actually 'too good', for bypassing, sometimes, and it's very difficult for most people (me included) to know in advance if they will be a problem or not. If film is all I could get, I might want to place at least a very small resistance in series with it (i.e. 'snubberize' it, sort-of), to try to pre-empt any high-frequency misbehavior.]
I tend to also want to solder a 0.1 uF or 0.22 uF X7R ceramic directly from one power pin to the other (soldered right onto the two pins, with leads as short as possible), since, in many amplifier topologies, that is an important current-flow/return path, and its high-frequency bypassing should be by the shortest-possible route, which some (most) layouts might not facilitate very well, since the bypassed frequencies would need to go through one power pin's separate bypass cap and all the way to ground and then back to the other power pin's bypass cap and then to the other power pin (being tortured and disfigured by parasitic inductance and resistance, all the way). Note, too, that in that case, it's even worse than that, since the two small bypass caps would appear to be in series, which halves their effective/total value, as seen for that bypass path.
For anyone interested in actually trying to determine what is better and what is worse, in bypassing and connection strategies for example, LTspice can give you at least a relative picture, IF you ALSO include estimates of the parasitic impedances of both the components and the conductors (e.g. the resistances and inductances of the PCB traces and wires). You can download everything you need for a good start, here: http://www.fullnet.com/~tomg/gooteesp.htm .
The only application in which I know artic silver is not recommended is on ram chips in some gfx cards... and even there it is not a set rule...
But the reason is the capacitive properties...
In my experience it survives application much longer than any of the white goops I tried so far... It tends to form a very tight capillary bond... it actualy has to cycle a few times from hot to cold after application to get the best effect... the info used to be on their site... probably still is.
PS... I use white silicone oxide all the time. Gt a huge tube...for prototypes etc... but final builds = artctic silver for me.
But the reason is the capacitive properties...
In my experience it survives application much longer than any of the white goops I tried so far... It tends to form a very tight capillary bond... it actualy has to cycle a few times from hot to cold after application to get the best effect... the info used to be on their site... probably still is.
PS... I use white silicone oxide all the time. Gt a huge tube...for prototypes etc... but final builds = artctic silver for me.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.