What is the level of success for soldering the cooling pad to the PCB with these ICs?
Some one recently did a dead bug style where the IC was placed upside down with respect to all the other components and then a separate cooling heatsink added to the exposed tap. Of course the IC pads need to be laid out for this upside down placement.
Some one recently did a dead bug style where the IC was placed upside down with respect to all the other components and then a separate cooling heatsink added to the exposed tap. Of course the IC pads need to be laid out for this upside down placement.
Hi Andrew,
That totally sounds feasible for something like the power pad SOIC parts, but I'm just using the standard (non-power pad) op-amps on this board.
For the TPS regs, I don't think this would work for two reasons:
1. The QFN pack means the power pad is on the same plane as the pads, so soldering leadwires to the pads would mean you'd have to elevate the heatsink and bridge the gap between the pad and heatsink with a metal shim or something. This would be tricky.
2. It would probably require more skill to dead-bug a QFN then it would to just solder it correctly. I have had success with ChipQuick low meting temp solder paste on the pad itself, and then applying heat with an iron to the vias on the other side of the board.
Begin rant
-----------------------
On the topic of part packages, I'm getting to a stage where I just don't understand the package choices by manufacturers anymore. Both the power pad QFN and the MSOP parts (used for the TPS series regulators and LT3042 regulator) are incredibly hard to solder correctly. Even with a stencil and an oven, I regularly have to touch up nearly every MSOP package I do thanks to solder bridges, and maybe 1 in 5 QFN packs fail and are very difficult to fix. Coming from a mobility background, we wouldn't touch anything in a package this big, so who's driving the need to such small and impractical packages on high power parts? For a truly size constrained application it would have to be a 1.6 x 1.6 4-bump WCSP pack. Even an MSOP would be out of the question. So what I don't understand is if it's a relatively high voltage and high power regulator, why not ship it in an appropriate size package? I know parts are not designed for hand soldering anymore, and that's totally understandable considering how small that market is, but they are regularly shipping packages that are too small to be practical and end up limiting the functionality of the part itself.
------------------------
End rant
Regards,
Owen
That totally sounds feasible for something like the power pad SOIC parts, but I'm just using the standard (non-power pad) op-amps on this board.
For the TPS regs, I don't think this would work for two reasons:
1. The QFN pack means the power pad is on the same plane as the pads, so soldering leadwires to the pads would mean you'd have to elevate the heatsink and bridge the gap between the pad and heatsink with a metal shim or something. This would be tricky.
2. It would probably require more skill to dead-bug a QFN then it would to just solder it correctly. I have had success with ChipQuick low meting temp solder paste on the pad itself, and then applying heat with an iron to the vias on the other side of the board.
Begin rant
-----------------------
On the topic of part packages, I'm getting to a stage where I just don't understand the package choices by manufacturers anymore. Both the power pad QFN and the MSOP parts (used for the TPS series regulators and LT3042 regulator) are incredibly hard to solder correctly. Even with a stencil and an oven, I regularly have to touch up nearly every MSOP package I do thanks to solder bridges, and maybe 1 in 5 QFN packs fail and are very difficult to fix. Coming from a mobility background, we wouldn't touch anything in a package this big, so who's driving the need to such small and impractical packages on high power parts? For a truly size constrained application it would have to be a 1.6 x 1.6 4-bump WCSP pack. Even an MSOP would be out of the question. So what I don't understand is if it's a relatively high voltage and high power regulator, why not ship it in an appropriate size package? I know parts are not designed for hand soldering anymore, and that's totally understandable considering how small that market is, but they are regularly shipping packages that are too small to be practical and end up limiting the functionality of the part itself.
------------------------
End rant
Regards,
Owen
Last edited:
Yes, if they are in similar packages. Pad layout details on page 17 of the opa1632 and page 28 for the lme part. Both parts are interchangeable.
I still believe the LME49724 part is technically superior to the OPA1632 in several ways.
1. The LME part has similar distortion performance in whether hooked up in single or differential input mode. The 1632 performs worse in SE mode than in DE mode as most amps do.
2. The 1632 distortion rises fast from 3kHz and above. The distortion of the LME part is flat and constant from 20Hz to 20kHz, a desirable quality for a musical amplifier based off Lamm Industries founder and designer, Vladimir Lamm.
3. The distortion performance of the LME part is load invariant, meaning it gives similar distortion specs loaded at 600 ohms or 2kOhms. The TI part shows significant performance spec changes based on its output load.
4. The LME part was tested and spec'd over an 80kHz(!) bandwidth as indicated on the distortion plots. The test bandwidth is not spec'd for the 1632 part on the datasheet but i asked their tech support and they "believe" it was 20Hz - 20kHz to give best performance.
I'm not sure why people prefer to use the OPA1632 over the LME49724. Both amps provide superior audio performance compared to anything else out there, but the LME part trumps the OPA part in several measures imho.
Obviously some may defer with my anaysis above and that's fine, really! Like i said before, both amps are excellent and the cream of the crop of their category.
1. The LME part has similar distortion performance in whether hooked up in single or differential input mode. The 1632 performs worse in SE mode than in DE mode as most amps do.
2. The 1632 distortion rises fast from 3kHz and above. The distortion of the LME part is flat and constant from 20Hz to 20kHz, a desirable quality for a musical amplifier based off Lamm Industries founder and designer, Vladimir Lamm.
3. The distortion performance of the LME part is load invariant, meaning it gives similar distortion specs loaded at 600 ohms or 2kOhms. The TI part shows significant performance spec changes based on its output load.
4. The LME part was tested and spec'd over an 80kHz(!) bandwidth as indicated on the distortion plots. The test bandwidth is not spec'd for the 1632 part on the datasheet but i asked their tech support and they "believe" it was 20Hz - 20kHz to give best performance.
I'm not sure why people prefer to use the OPA1632 over the LME49724. Both amps provide superior audio performance compared to anything else out there, but the LME part trumps the OPA part in several measures imho.
Obviously some may defer with my anaysis above and that's fine, really! Like i said before, both amps are excellent and the cream of the crop of their category.
Andrew brings up a critical point... I assumed the LME part came in power pad or non-power pad variants like the OPA, but it turns out it only comes in power pad packages.
You can still use the LME power pad version, but you will need to cut little pieces of mylar tape (or some other thin electrical insulator) to electrically isolate the power pad on the bottom of the IC from the exposed vias on the PCB beneath the op-amp. If this isn't done, things will go badly.
Otherwise, both are pin-compatible SOIC-8 packages, so they will work on the same footprint save for the power pad.
Regards,
Owen
You can still use the LME power pad version, but you will need to cut little pieces of mylar tape (or some other thin electrical insulator) to electrically isolate the power pad on the bottom of the IC from the exposed vias on the PCB beneath the op-amp. If this isn't done, things will go badly.
Otherwise, both are pin-compatible SOIC-8 packages, so they will work on the same footprint save for the power pad.
Regards,
Owen
Thank you for the warning. I have used both with buffalo IVY, but they have no vias under the IC. There are some thin isolating tape for power ICs in my drawer, so i must rember to tape some under the LME.
May I ask why there are no high frequency decupling of the signal as the max signal frequency are 96 khz?
Regards Torgeir
May I ask why there are no high frequency decupling of the signal as the max signal frequency are 96 khz?
Regards Torgeir
Ehm, 48kHz max signal frequency. Sorry about that.
From http://www.pimmlabs.com/web/behringer.htm we can see that the original board has quite som HF protection on the input.
My reason for worrying is the differential opamps high bandwith and aliased HF noise in the ADC if unlucky.
TI recomends some filtering (C1,2,3) themselves in the figure 11 of http://www.ti.com/lit/ds/symlink/opa1632.pdf
It looks like, at least there, are a 6dB LP filter on the ADC board?. (C31 or C32) This will not be removed by the upgrade?
I know the measurements looks good, but the HF noise the device is exposed to can variate a lot from from one enviroment to another.
Regards Torgeir
From http://www.pimmlabs.com/web/behringer.htm we can see that the original board has quite som HF protection on the input.
My reason for worrying is the differential opamps high bandwith and aliased HF noise in the ADC if unlucky.
TI recomends some filtering (C1,2,3) themselves in the figure 11 of http://www.ti.com/lit/ds/symlink/opa1632.pdf
It looks like, at least there, are a 6dB LP filter on the ADC board?. (C31 or C32) This will not be removed by the upgrade?
I know the measurements looks good, but the HF noise the device is exposed to can variate a lot from from one enviroment to another.
Regards Torgeir
Hi Torgier,
The original Behringer ADC input schematic yields an LPF that is about 1dB down at 120kHz, 9dB down at 500kHz and 17dB down at 1MHz.
If you look at the datasheet for the ADC used (the AKM5393) you'll see that they simply suggest using a filter similar to the one already on the Behringer digital board at the input of the ADC. In the case of the Behringer, the filter at the input of the ADC is down about 3dB at 500kHz and 7dB at 1MHz. I don't feel that there's much to gain adding additional filtering.
If you do want that though, it's very easy to tack two NPO caps on the feedback resistors, and another cap across the inputs or outputs to form a filter similar to what is in the OPA1632 datasheet.
I agree that there are noisier environments out there, but for your average hifi setup, with proper shielded XLR cables and relatively short runs, this generally won't be an issue.
All the above also applies to the outputs. It's easy to add additional filtering there if you feel you need it. I did provide a measurement of the output as an FFT up to 1MHz and as you can see, there isn't anything out of the ordinary there. The peak at the sampling frequency (48kHz) is at about -124dB and none of the filters mentioned above would impact it anyhow.
Regards,
Owen
The original Behringer ADC input schematic yields an LPF that is about 1dB down at 120kHz, 9dB down at 500kHz and 17dB down at 1MHz.
If you look at the datasheet for the ADC used (the AKM5393) you'll see that they simply suggest using a filter similar to the one already on the Behringer digital board at the input of the ADC. In the case of the Behringer, the filter at the input of the ADC is down about 3dB at 500kHz and 7dB at 1MHz. I don't feel that there's much to gain adding additional filtering.
If you do want that though, it's very easy to tack two NPO caps on the feedback resistors, and another cap across the inputs or outputs to form a filter similar to what is in the OPA1632 datasheet.
I agree that there are noisier environments out there, but for your average hifi setup, with proper shielded XLR cables and relatively short runs, this generally won't be an issue.
All the above also applies to the outputs. It's easy to add additional filtering there if you feel you need it. I did provide a measurement of the output as an FFT up to 1MHz and as you can see, there isn't anything out of the ordinary there. The peak at the sampling frequency (48kHz) is at about -124dB and none of the filters mentioned above would impact it anyhow.
Regards,
Owen
Thank you, Owen, for your comments.
I think i will solder up the board, put it inside, plug in input cables and measure with my 100MHz oscilloscope on the output of the inputbuffer. If I see HF noise, I will add the filtereing caps you mention. (If unlucky I have EMI at multiples of 128fs).
As a sidenote I found one mesurement "FFT 20Hz to 1MHz digital input" but that does not include the analog input I guess.
Maybe I choose to dedicate some outputs to bass and filter, under 300 Hz at 6dB/oct, the way you suggest.
I have a high sensitivity 15 inch for the upper bass and it has really high sensitivity at 1.5k (106dB/w) so I hear noise in that area from that element very easily. http://www.cadaudio.dk/pr380m2.pdf
I first must check if I will only hear the noise from the poweramp and must use a passive LP filter at the speaker to get rid of any noise around 1.5 kHz.
I really look forward to test your design. I think it will work really well and that I will not hear noise from the filter at all.
Regards Torgeir
https://sites.google.com/site/passivefilter/
I think i will solder up the board, put it inside, plug in input cables and measure with my 100MHz oscilloscope on the output of the inputbuffer. If I see HF noise, I will add the filtereing caps you mention. (If unlucky I have EMI at multiples of 128fs).
As a sidenote I found one mesurement "FFT 20Hz to 1MHz digital input" but that does not include the analog input I guess.
Maybe I choose to dedicate some outputs to bass and filter, under 300 Hz at 6dB/oct, the way you suggest.
I have a high sensitivity 15 inch for the upper bass and it has really high sensitivity at 1.5k (106dB/w) so I hear noise in that area from that element very easily. http://www.cadaudio.dk/pr380m2.pdf
I first must check if I will only hear the noise from the poweramp and must use a passive LP filter at the speaker to get rid of any noise around 1.5 kHz.
I really look forward to test your design. I think it will work really well and that I will not hear noise from the filter at all.
Regards Torgeir
https://sites.google.com/site/passivefilter/
This is the way electronics has been going for quite a while, smaller and smaller parts, driven by the demand for more functionality in smaller cases etc. You may think its hard for DIYer (which I know it is) but it also makes life bloody hard as a PCB designer, we are faced with more and more functionality, move devices crammed into a smaller area, no room for power tracks, more heat to dissipate and less area to get the heat out. A lot is driven by handheld devices, some of the Ti headphone amps are in 0.5mm pitch BGA packages, with few pins, a nightmare to use a you have to use via in pad to get the traces out, and the soldered pad to device ratio is very small so they can drop off at the slightest bang.
Also getting these parts evaluated and approved for high reliability designs is a nightmare if not impossible in some cases. The one advantage these devices do have is getting the heat directly from the actual chip itself, providing you can get the heat out, the lack of lead frame and the chip being mounted directly on the thermal pad does mean that the chip temperature can be kept down, it just takes a lot of careful PCB layout, thermal vias, multiple layers, bespoke heat5 sink etc. as the power/thermal density of assemblies increases. Some designs would bring tears to your eyes (they do mine) when you have such high density's including the number of connections. The problem is for cost based designs they can and are often compromised, but as a limited life expectancy if common with throw away devices it doesn't seem to be a major concern, often the bean counters rule....... To often sometimes.
Now if you want a challenge do some high density LED layouts that are done to a restricted cost, ie normal PCB material instead of metal cored and only two layers, with SMPSs and have to pass EWMC and thermal requirements plus have at least a % year life span (tin whiskers will be a concern after that so often guarantees are only 5 years these days).
They make what sells, and handheld devices and similar are the big market these days.
I received two boards today, thank you!
I have three questions:
A: The C6-7 C9-10 C44 Nichicon CAP ALUM 330UF 16V 20% RADIAL caps: Are they critical or can I use my normal 220 uF audio caps laying in my drawer. I am not using the regulators.
B: Looks like voltage rating is a little small? 15V rails and 16V caps?
C: How low would you go on OPA1632 resitors? We are only talking 2.5 volts here so 1K seems a little high for lowest noise?
And the Panasonic RES 1.18K OHM 1/8W .1% 0805 SMD is a little pricy with 50 pieces minimum for sensible price in Europe.
(I know I must recalculate the amplification
Regards Torgeirs
I have three questions:
A: The C6-7 C9-10 C44 Nichicon CAP ALUM 330UF 16V 20% RADIAL caps: Are they critical or can I use my normal 220 uF audio caps laying in my drawer. I am not using the regulators.
B: Looks like voltage rating is a little small? 15V rails and 16V caps?
C: How low would you go on OPA1632 resitors? We are only talking 2.5 volts here so 1K seems a little high for lowest noise?
And the Panasonic RES 1.18K OHM 1/8W .1% 0805 SMD is a little pricy with 50 pieces minimum for sensible price in Europe.
(I know I must recalculate the amplification
Regards Torgeirs
Then you don't have 15Vdc supply rails.
They vary with loading and with mains supply voltage.
Hi Torgeirs,
Here are the answers to your questions:
A: These are not critical at all and you can use whatever you'd like in these locations. The polymer parts were chosen for low ESR and long life, but they won't impact audio performance.
B: The 15V rails are regulated, so 16V caps are fine. You can use 25V parts if you prefer though.
C: Since the feedback and input resistors set the gain as well as the input impedance, they need to be considered together so as to not load the output of the DAC, or whatever happens to be driving the analog input. You can reduce these down to whatever you'd like, but you will start loading the DAC and presenting an excessively low input impedance for the analog inputs. With ~1K resistors the noise levels of the op-amps will not be the dominant contributor to the noise floor, so it's best to leave them as they are.
If you can't find the Panasonic 1.18K parts, then any other thin film part will work. If you can't find anything in that value, then try for 1.15K or 1.20K. This will change the gain somewhat, but that's probably not critical as long as they all match.
Regards,
Owen
Here are the answers to your questions:
A: These are not critical at all and you can use whatever you'd like in these locations. The polymer parts were chosen for low ESR and long life, but they won't impact audio performance.
B: The 15V rails are regulated, so 16V caps are fine. You can use 25V parts if you prefer though.
C: Since the feedback and input resistors set the gain as well as the input impedance, they need to be considered together so as to not load the output of the DAC, or whatever happens to be driving the analog input. You can reduce these down to whatever you'd like, but you will start loading the DAC and presenting an excessively low input impedance for the analog inputs. With ~1K resistors the noise levels of the op-amps will not be the dominant contributor to the noise floor, so it's best to leave them as they are.
If you can't find the Panasonic 1.18K parts, then any other thin film part will work. If you can't find anything in that value, then try for 1.15K or 1.20K. This will change the gain somewhat, but that's probably not critical as long as they all match.
Regards,
Owen
Hi Andrew,
The existing +/-15VDC rails in the Behringer are regulated with a 7815 and a 7915 mounted inside the SMPS enclosure. The output voltage tolerance on these is usually 2% but can be as high as 5% depending on the manufacturer. Either way though, you're guaranteed to be below 16VDC.
The purpose of the secondary regulators (which drop the +/-15VDC down to +/-12VDC) is to provide better PSRR at the frequencies the SMPS operates at, and to lower the dissipation on the OPA1632 op-amps since they tend to run a little warm at +/-15V.
Even if the on-board TPS regs are bypassed, 16V caps can safely be used in all locations on the output board.
Regards,
Owen
Forgot about the DAC drive capacity....
I see Behringer uses 2k//5k and AKM has 620 to ground. Is the input impedance 1k or is the input impedance 1k//1k18?
Tried to read the
http://www.ti.com/lit/an/sloa054d/sloa054d.pdf
to get some more insigth in differential amps.
Think I can see that johnson noise is function of the sum of the resistors when gain is 1. That should be about -125dBV for each resistor at 1K Ohm and 20kHz BW. About -119dBV total and back to about -125dB referred to max output.
I also see that "Figure 26. First-Order Active Low-Pass Filter" shows the ekstra 1 order LP filter you suggest can be used in post#150.
About the caps, just wondered if there was some new "Supercaps" around that don't sag with temperature and age etc. I think i will stay on the safe side and use my old 25V parts;-)
Thank you again for a great design.
Regards Torgeir
I see Behringer uses 2k//5k and AKM has 620 to ground. Is the input impedance 1k or is the input impedance 1k//1k18?
Tried to read the
http://www.ti.com/lit/an/sloa054d/sloa054d.pdf
to get some more insigth in differential amps.
Think I can see that johnson noise is function of the sum of the resistors when gain is 1. That should be about -125dBV for each resistor at 1K Ohm and 20kHz BW. About -119dBV total and back to about -125dB referred to max output.
I also see that "Figure 26. First-Order Active Low-Pass Filter" shows the ekstra 1 order LP filter you suggest can be used in post#150.
About the caps, just wondered if there was some new "Supercaps" around that don't sag with temperature and age etc. I think i will stay on the safe side and use my old 25V parts;-)
Thank you again for a great design.
Regards Torgeir
Last edited:
I also see I can use the pads of the voltage regulator Cs to add low value (100pF - 1 nF ) NP0 caps and midvalue (10uF X7R) to remove HF noise from the Behringer powersupply.
I would also be tempted to add piggyback NP0 caps to the 1uF caps near the OPA1632s
(Just to make it more immune to HF noise, but it might be overkill with my level of HF noise)
At least it is objectively a better way of using the money on caps than four 330uF caps when not using the regulator
http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf
http://www.avx.com/docs/techinfo/mlcbypas.pdf
I would also be tempted to add piggyback NP0 caps to the 1uF caps near the OPA1632s
(Just to make it more immune to HF noise, but it might be overkill with my level of HF noise)
At least it is objectively a better way of using the money on caps than four 330uF caps when not using the regulator
http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf
http://www.avx.com/docs/techinfo/mlcbypas.pdf
I think i found the answer to the input imp question in:
http://www.ti.com/lit/ds/symlink/lmh6554.pdf section 9.2.2. RIN = 2RG = 2 * 1K = 2 K. So it is maybe some room for lower resistors if noise is still an issue.
I am trying to find the resistance that the output of the OPA1632 sees with no load. Is it Rg + Rfb? Then the OPA just have to deliver about 2 mA. And the 6 A ripplecurrent in one 330uA cap should be more than enough for the 8 devices. (And even my 50V 220uF panasonic caps can deliver 1.5A)
http://www.ti.com/lit/ds/symlink/lmh6554.pdf section 9.2.2. RIN = 2RG = 2 * 1K = 2 K. So it is maybe some room for lower resistors if noise is still an issue.
I am trying to find the resistance that the output of the OPA1632 sees with no load. Is it Rg + Rfb? Then the OPA just have to deliver about 2 mA. And the 6 A ripplecurrent in one 330uA cap should be more than enough for the 8 devices. (And even my 50V 220uF panasonic caps can deliver 1.5A)
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.