Back in the day we used to have problems with the same type of oscillator circuit, i.e. some brands/type of 04 inverter would be temperamental in starting up, some even double oscillating!. I looked at my SR650 repair video and it has the 04 socketed also.......so I believe it's just a precaution by the factory that in the event there's a startup issue they can easily change it out.
CMOS gate oscillators tend to have problems because of low transconductance, especially over their operating temperature range. This problem sometimes shows up with microcontrollers where they have two external pins for connecting the crystal. Single transistor bipolar oscillators are much more reliable, but nobody has time for that anymore.
Regarding the RC marked capacitors, the schematic shows a LT1037 decompensated operational amplifier which is only stable at gains of 5 or greater, so adding a feedback capacitor between the output and inverting input is going to cause a lot of peaking if not outright oscillation. 130 picofarads is pretty low, but is way larger than would be necessary to compensate for stray capacitance at the inverting input. I thought maybe the board used LT1007s instead which are unity gain stable, but your video shows LT1037s on the board.
Thanks for the training. I spent my carreer working on HP test and measurement gear, mostly RF and microwave. From 1970 to 2015. Happily retired, but miss the challenges of component level repair. Ex-USN ET.
In EEVblog #621 Dave repaired an SR650 with a very similar problem. In his unit the cover over the mains filter was missing, so the 5V regulator had no heat sink at all.
An amazing instrument with that rate of cutoff! Kronhite makes a lower cost version that is entirely analog using ganged pots for the filter adjustment. Of course nowhere near that cutoff rate or dynamic range. (or price!) What era is this instrument? Looks sort of 1980's? I had the impression the 7805's were kill proof with the thermal shutdown.
@12:45, Are you sure you wanted to replace that bridge? It's a linear supply, so the voltage has to be dropped somewhere. If it isn't dropped in the bridge, won't it be dropped in the regulators, which are clearly also pretty stressed? I was thinking maybe you would want to throw some extra dissipating components in to add a little drop, at least before the 5v reg.
It looks like it is a two stage supply, which is typical for this sort of equipment. The first stage (typically an SMPS these days) drops from the mains voltage to the 9V he talked about, then the linear regulator drops that to 5V as a second stage. The reason for this is that the first stage will create significant ripple and rather than attempt to filter that away with inductors and capacitors, the second stage just chops off that ripple. That is the theory and it almost works like that in practice, but the higher frequency noise elements still require some filtering and the current oscillations can induce noise wherever they can couple. The V[F] on that bridge sounded pretty good. I wonder which one he selected. I often over spec the V[RRM] and end up with more voltage drop, which adds heat, but otherwise is allegedly a preferred tradeoff since the leakage current does more damage over time than thermal wear out. I have never seen the math on this, though.
You may be right. Maybe I missed it, but Shariar never resolved the "it's supposed to be 9v, but it's actually 13v" conundrum. I wonder if the root problem here is that the pre-regulator's pass transistor is shorted, causing the 5v final regulator to do all the work? Regardless, my point stands. Everything's in series, so one way or another, there are I amps flowing from the transformer winding, through the high side of the bridge, through both regulators' pass transistors, to the 5V rail, through the instrument, through the low side of the bridge, and back to the transformer. There's 5V across the instrument, and 13 volts on the unregulated rail. Which means 8I watts are being dissipated in the 5V power supply, whether that's being dropped in the rectifier or the pass transistor(s). So if you drop less in the rectifier, you will just drop more in the pass transistor(s). That analysis does overlook one detail. An A/C current flows through the rectifier and the tank capacitor, which doesn't flow through the rest of the circuit, so reducing the rectifier drop will reduce the dissipation in the rectifier more than it increases the dissipation in the pass transistors. I don't feel like doing the math to figure it out now, but intuitively, it seems to me that that effect will be pretty small, owing to the fact that the rectifier forward drop is more or less constant regardless of current, so the rectifier dissipation will be proportional to Iavg, not Irms (approximately).
agreed, the lower Vf of the new bridge is just adding more voltage drop that the 5V reg has to dissipate. I'm sure he's well aware of that; hopefully the "V_in = 13 instead of 9V" situation was also just resolved off-camera
13:25 These silicone pads are supposed to be dry. I use them all the time in my projects. They work somewhat worse than the paste but last effectively forever and don't smudge everything around.
They are "dry", but they become less flexible and shrink with time. They won't make good contact with the two surfaces anymore. Thermal pads like this may be convenient, but really should only be used when electrical isolation is needed. They perform much worse than any paste.
At 18:03 you demo the BP. Ignoring the 60Hz as best as one can, why does there seem to be a difference in the stop bands of the low compared to the high?
@@Thesignalpath did you sweep noise or a fixed amplitude sine? In either case I would think the signal coming in is the same amplitude across the sweep range. No?
@@Thesignalpath thanks. If the sweep is constant amplitude as a sweep should be, what then accounts for the noise floor difference of the low frequencies as compared to the high? Again, it’s a little hard to tell because the 60 Hz dominates quite a bit. Or is it because of the 60Hz interacting with the windowing in the spectral plot that’s causing the low end noise floor to look higher than the high noise floor?
I think I misunderstood your original question. Remember the scale is logarithmic and the noise goes up at lower frequencies because the front-end circuits have higher 1/f noise near DC.
It is no surprise you are getting noise with 40dB gain in the filter. A much fairer test would be to run 0dBm in and look at S/N in that setting. Other than that, crazy good response curves. Truly a problem solver filter!
Also: how are the DACs designed to affect Fc in the filter loops to get smooth Fc changes instead of the discrete Fc poles/zeros caused by the resistors and capacitors switched in or out of the filter loops?
This thing might be the wet dream of any CW operator with mild hearing impairments. I'm always in awe with these high class instruments. I am just a hobbyist and all my devices have spec tolerances +/-1000% 😀 Well at least it teaches you to never blindly trust any readout.
Hello, rolloff steepness and passband flatness are great, but I am suprised to see that the 60 Hz powerline has such a big influence. Is this normal for this device (or similar devices from other vendors) or is it a malfunction of this particular unit?
You know, this video, with the failing inverter, vintage chips, clock generator, and mystery capacitors floating in the air, reminds me a lot of Adria's Digital Basement, and his "whodunnit" style Amiga/vintage PC repairs ;)
I have another instrument, that an 74hc04 is dead in the PFD path of a PLL, replaced & everything worked. Wondering why those 74XX04 failed so frequent :(
My guess for the "floating" capacitor: They were definitely put there by the factory. The solder-joints and everything else looks really clean. Obviously that was an afterthought after the PCB layout was already done. So what I think that happend was an "after-after-thought". So for some reason, someone decided, this capacitor was not helpful after all. But since this batch of boards was already fully assembled with the capacitor in place, someone had to come in again an "remove" them all. Lifting just one leg takes only half the time, so if they had to go through a good number of boards, they just left it at that. They were not gonna re-use the caps anyway. It's just my guess. If someone had worked on these parts after they left the factory, the joints would not look this clean still.
If this were done at factory they could just "snip" the part of which would be much faster and less messy for them. This looks like someone was "experimenting" and wanted to go back and forth between the two mode (capacitor included/not included). The capacitor can also easily bend and touch another pin. I would be surprised if factory would leave it like that.
If the schematic says you are supposed to see 9v at the 7805 input, and you are actually seeing 13v, could there be a pre-regulator whose pass transistor is shorted? The circuit would still function, but the dissipation in the 7805 would be double the design, as you are dropping 8v instead of 4v. And if the failed pre-regulator presents a low-impedance path to ground, that could also explain excessive dissipation in the diode bridge.
The cover would not fix the issue with the 5V regulator. The cover was on when it originally overheated and the display dimmed away. The thermal interface was faulty. The rectifier is also not in the path of any air-flow. The only components which are in the air-flow where the +/- 15V regulators.
Back in the day we used to have problems with the same type of oscillator circuit, i.e. some brands/type of 04 inverter would be temperamental in starting up, some even double oscillating!.
I looked at my SR650 repair video and it has the 04 socketed also.......so I believe it's just a precaution by the factory that in the event there's a startup issue they can easily change it out.
Thanks Ian for checking!
CMOS gate oscillators tend to have problems because of low transconductance, especially over their operating temperature range. This problem sometimes shows up with microcontrollers where they have two external pins for connecting the crystal. Single transistor bipolar oscillators are much more reliable, but nobody has time for that anymore.
@@AlamagosaYou mean they don't have enough current drive strength?
Regarding the RC marked capacitors, the schematic shows a LT1037 decompensated operational amplifier which is only stable at gains of 5 or greater, so adding a feedback capacitor between the output and inverting input is going to cause a lot of peaking if not outright oscillation. 130 picofarads is pretty low, but is way larger than would be necessary to compensate for stray capacitance at the inverting input. I thought maybe the board used LT1007s instead which are unity gain stable, but your video shows LT1037s on the board.
Nice observation!
Thanks for the training. I spent my carreer working on HP test and measurement gear, mostly RF and microwave. From 1970 to 2015. Happily retired, but miss the challenges of component level repair. Ex-USN ET.
Tech at Santa Rosa?
In EEVblog #621 Dave repaired an SR650 with a very similar problem. In his unit the cover over the mains filter was missing, so the 5V regulator had no heat sink at all.
Repairs are everyone's favorite!
Greetings from Romania
Great analysis and repair, love how you approach the problem solving. Cheers.
An amazing instrument with that rate of cutoff! Kronhite makes a lower cost version that is entirely analog using ganged pots for the filter adjustment. Of course nowhere near that cutoff rate or dynamic range. (or price!) What era is this instrument? Looks sort of 1980's? I had the impression the 7805's were kill proof with the thermal shutdown.
@12:45, Are you sure you wanted to replace that bridge? It's a linear supply, so the voltage has to be dropped somewhere. If it isn't dropped in the bridge, won't it be dropped in the regulators, which are clearly also pretty stressed? I was thinking maybe you would want to throw some extra dissipating components in to add a little drop, at least before the 5v reg.
if he just puts the cover on it, it will be just fine.
It looks like it is a two stage supply, which is typical for this sort of equipment. The first stage (typically an SMPS these days) drops from the mains voltage to the 9V he talked about, then the linear regulator drops that to 5V as a second stage. The reason for this is that the first stage will create significant ripple and rather than attempt to filter that away with inductors and capacitors, the second stage just chops off that ripple.
That is the theory and it almost works like that in practice, but the higher frequency noise elements still require some filtering and the current oscillations can induce noise wherever they can couple.
The V[F] on that bridge sounded pretty good. I wonder which one he selected. I often over spec the V[RRM] and end up with more voltage drop, which adds heat, but otherwise is allegedly a preferred tradeoff since the leakage current does more damage over time than thermal wear out. I have never seen the math on this, though.
You may be right. Maybe I missed it, but Shariar never resolved the "it's supposed to be 9v, but it's actually 13v" conundrum. I wonder if the root problem here is that the pre-regulator's pass transistor is shorted, causing the 5v final regulator to do all the work?
Regardless, my point stands. Everything's in series, so one way or another, there are I amps flowing from the transformer winding, through the high side of the bridge, through both regulators' pass transistors, to the 5V rail, through the instrument, through the low side of the bridge, and back to the transformer. There's 5V across the instrument, and 13 volts on the unregulated rail. Which means 8I watts are being dissipated in the 5V power supply, whether that's being dropped in the rectifier or the pass transistor(s). So if you drop less in the rectifier, you will just drop more in the pass transistor(s).
That analysis does overlook one detail. An A/C current flows through the rectifier and the tank capacitor, which doesn't flow through the rest of the circuit, so reducing the rectifier drop will reduce the dissipation in the rectifier more than it increases the dissipation in the pass transistors. I don't feel like doing the math to figure it out now, but intuitively, it seems to me that that effect will be pretty small, owing to the fact that the rectifier forward drop is more or less constant regardless of current, so the rectifier dissipation will be proportional to Iavg, not Irms (approximately).
agreed, the lower Vf of the new bridge is just adding more voltage drop that the 5V reg has to dissipate. I'm sure he's well aware of that; hopefully the "V_in = 13 instead of 9V" situation was also just resolved off-camera
13:25 These silicone pads are supposed to be dry. I use them all the time in my projects. They work somewhat worse than the paste but last effectively forever and don't smudge everything around.
They are "dry", but they become less flexible and shrink with time. They won't make good contact with the two surfaces anymore. Thermal pads like this may be convenient, but really should only be used when electrical isolation is needed. They perform much worse than any paste.
At 18:03 you demo the BP. Ignoring the 60Hz as best as one can, why does there seem to be a difference in the stop bands of the low compared to the high?
Because in the low-pass mode the signal present at low-frequencies by the sweep generator dominates and is much larger than the 60Hz noise.
@@Thesignalpath did you sweep noise or a fixed amplitude sine? In either case I would think the signal coming in is the same amplitude across the sweep range. No?
The sweep is at a fixed swing across frequency, it is a sine wave not broadband noise.
@@Thesignalpath thanks. If the sweep is constant amplitude as a sweep should be, what then accounts for the noise floor difference of the low frequencies as compared to the high? Again, it’s a little hard to tell because the 60 Hz dominates quite a bit. Or is it because of the 60Hz interacting with the windowing in the spectral plot that’s causing the low end noise floor to look higher than the high noise floor?
I think I misunderstood your original question. Remember the scale is logarithmic and the noise goes up at lower frequencies because the front-end circuits have higher 1/f noise near DC.
It is no surprise you are getting noise with 40dB gain in the filter. A much fairer test would be to run 0dBm in and look at S/N in that setting. Other than that, crazy good response curves. Truly a problem solver filter!
Also: how are the DACs designed to affect Fc in the filter loops to get smooth Fc changes instead of the discrete Fc poles/zeros caused by the resistors and capacitors switched in or out of the filter loops?
Hello Shahriar!
Hi!
This thing might be the wet dream of any CW operator with mild hearing impairments.
I'm always in awe with these high class instruments. I am just a hobbyist and all my devices have spec tolerances +/-1000% 😀
Well at least it teaches you to never blindly trust any readout.
Perfect diagnosis and repair of course.
For me it's a pleasure see these kind of video.
Hello, rolloff steepness and passband flatness are great, but I am suprised to see that the 60 Hz powerline has such a big influence. Is this normal for this device (or similar devices from other vendors) or is it a malfunction of this particular unit?
You know, this video, with the failing inverter, vintage chips, clock generator, and mystery capacitors floating in the air, reminds me a lot of Adria's Digital Basement, and his "whodunnit" style Amiga/vintage PC repairs ;)
I have another instrument, that an 74hc04 is dead in the PFD path of a PLL, replaced & everything worked. Wondering why those 74XX04 failed so frequent :(
Nice , dry capacitor may cause diode bridge failed, high ripple current,
I use mosfet rectifier in my all own design , no more drope too. Cold as ice
the signal phat hello may from Argentina broooooo 🇦🇷
Christmas has come early❤
as someone more familiar with synthesizers (the musical kind) than high-end test gear, that thumbnail is crazy lol
wow 60db of gain too? this thing is basically just a really fancy microphone preamp too
"lights are up, but nobody's home" situation
My guess for the "floating" capacitor:
They were definitely put there by the factory. The solder-joints and everything else looks really clean. Obviously that was an afterthought after the PCB layout was already done.
So what I think that happend was an "after-after-thought". So for some reason, someone decided, this capacitor was not helpful after all. But since this batch of boards was already fully assembled with the capacitor in place, someone had to come in again an "remove" them all. Lifting just one leg takes only half the time, so if they had to go through a good number of boards, they just left it at that. They were not gonna re-use the caps anyway.
It's just my guess. If someone had worked on these parts after they left the factory, the joints would not look this clean still.
If this were done at factory they could just "snip" the part of which would be much faster and less messy for them. This looks like someone was "experimenting" and wanted to go back and forth between the two mode (capacitor included/not included). The capacitor can also easily bend and touch another pin. I would be surprised if factory would leave it like that.
If the schematic says you are supposed to see 9v at the 7805 input, and you are actually seeing 13v, could there be a pre-regulator whose pass transistor is shorted? The circuit would still function, but the dissipation in the 7805 would be double the design, as you are dropping 8v instead of 4v. And if the failed pre-regulator presents a low-impedance path to ground, that could also explain excessive dissipation in the diode bridge.
Come to think of it, this might also explain excessive power line harmonics bleeding into the signal path.
Nice thumbnail :)
the signal phat hello plis laser module😊
Amazing what a cover will do for thermals huh.
The cover would not fix the issue with the 5V regulator. The cover was on when it originally overheated and the display dimmed away. The thermal interface was faulty. The rectifier is also not in the path of any air-flow. The only components which are in the air-flow where the +/- 15V regulators.
@@Thesignalpath 5 volt reg bad, yes. where are the other holes let air in?
Nowhere near the bridge rectifier, which is why it would get so hot.
Это же не цифровой, не математический фильтр сигналов...