|
|
Post by Johnkenn on Feb 25, 2015 21:00:01 GMT -6
|
|
|
|
Post by svart on Feb 26, 2015 8:21:30 GMT -6
Personally I'm good. The project is going slowly due to day job commitments and the continuous winter storms that seem to hamper my parts orders. I've been already waiting 6 days for an order that should have been 2 days shipping, tops. Anyway, I suppose I'll let the cat out of the bag.. I've built a DAC board and it works! We were doing listening tests over the weekend and all seems well with the sound quality. Now, I've found a problem with the physical board itself and I'm working with the PCB house to figure out if it was my fault somehow or if the boards were made wrong. They seem to think it's my fault, and I'm positive that it wasn't. In any case, I need a new set of DAC boards before I can start populating them. However, I can use what I have to work on the software for the DAC boards now and should have that done in the next couple weeks. I'm also starting on the cutting jig for the front panel tonight so I should have that sorted out in the next few days. I still need to do something with the front bezel. I have it designed in general but it's rather plain. I need to figure out a logo or something to make it look nicer. |
|
kcatthedog
Temp
Super Helpful Dude
Posts: 16,042 
Member is Online
|
Post by kcatthedog on Feb 26, 2015 8:46:51 GMT -6
devils always in the deets !
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 26, 2015 20:15:47 GMT -6
Is it the chinese online-ordering service, PCBCART?
Heard only good of them, generally, good quality and very reasonably priced on small series production, but sometimes communication needs some ... improvement, but they have at least one service guy that is capable of good english communication with minimized misunderstandings...don't remember his name...ordered once years ago...
Just in case...it's been a long time favorite for many small series designers, therefore i ask...
Nowadays there are some more competitors with similar quality service, also china-based, that even offer live online tech support in english....
|
|
|
|
Post by formatcyes on Feb 27, 2015 14:22:37 GMT -6
Hi svart Some of us only want the AD If you have that ready to go cash will flow  Stop reading this crap and get back to work. |
|
|
|
Post by svart on Feb 27, 2015 14:36:53 GMT -6
Hi svart Some of us only want the AD If you have that ready to go cash will flow Stop reading this crap and get back to work. Understood. However, since this will be a system, I need to make sure it all works together flawlessly.. I don't want to have a recall for any reason if you know what I mean! Anyway, it's getting close. Just have to get some more time! |
|
|
|
Post by svart on Mar 2, 2015 9:47:44 GMT -6
Not feeling that great today, head is really stuffy and I'm tired.
Plugged the DAC board in backwards while working on it on the bench.
All I heard was PFFFFT as at least one of the chips died.
Setback.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Mar 2, 2015 15:55:44 GMT -6
Ouch.
Shit happens. More luck another day....
|
|
|
|
Post by Johnkenn on Mar 2, 2015 19:46:22 GMT -6
|
|
|
|
Post by svart on Mar 2, 2015 19:54:11 GMT -6
Some good came of it though, i was doing testing with the bench power supply, troubleshooting the board, and found a better filter setup for the regulators.
|
|
|
|
Post by svart on Mar 3, 2015 9:25:27 GMT -6
It also seems that the PCM1794 draws considerable short-duration currents on it's analog VCC inputs. The datasheet shows, but does not mention, bulk capacitors at 10uF.
I used 10uF tantalum at these points since the local regulators have the bandwidth to oscillate if the ESR is too low, or too high. Tantalum is neither low nor high ESR for average part. Testing showed that there was more current/voltage surging than I'd like, so I attempted to use 10uF ceramic to see if lower ESR would help.
Oscillation city. If left like this, the IC and/or the regulator would eventually overheat.
For the given current draw, and it's jerky nature of it, the PCM1794 sent the regulator into oscillation.
I finally settled on a 47uF tantalum for the analog supply's primary regulator output cap. It has the right ESR and plenty of capacitance to keep the part happy.
Because of this, I measured a 20-30F drop in temps on the output opamps. They were running extremely hot, now they are just hot. There was no oscillation in the I/V circuit, nor the last buffer, but the virtual earth setup causes currents to be drawn in/out of the opamps as the PCM1794 attempts to slew them. The excess current spikes were obviously causing this.
In other words, if someone were to use POL (point of load) regulation and only the datasheet's "typical application circuit" then they'd possibly see some extra heat and distortion from the part.
|
|
kcatthedog
Temp
Super Helpful Dude
Posts: 16,042
Member is Online
|
Post by kcatthedog on Mar 3, 2015 9:28:03 GMT -6
I got that last sentence perfectly |
|
|
|
Post by jimwilliams on Mar 3, 2015 9:59:05 GMT -6
I typically hang a 1000uf very low ESR cap (Panasonic FR series) on the output of linear regulators, 317, 337, 1085, etc. The ref pins also get a large cap, 100 uf. All of those regulator designs are stable.
They convert 60 hz to DC. Why are wideband regulators necessary for a 60 hz to DC rectification? The fact that they are finiky with the loads makes them not very reliable compared to more robust designs. If the 4898 opamps are getting hot that makes me suspect there is an instability someplace. Those opamps have a pretty good phase margin so they tend to be stable if the circuit is.
Can you tell us the part numbers of those regulators? I want to make sure I never select them.
|
|
|
|
Post by svart on Mar 3, 2015 10:42:00 GMT -6
I typically hang a 1000uf very low ESR cap (Panasonic FR series) on the output of linear regulators, 317, 337, 1085, etc. The ref pins also get a large cap, 100 uf. All of those regulator designs are stable. They convert 60 hz to DC. Why are wideband regulators necessary for a 60 hz to DC rectification? The fact that they are finiky with the loads makes them not very reliable compared to more robust designs. If the 4898 opamps are getting hot that makes me suspect there is an instability someplace. Those opamps have a pretty good phase margin so they tend to be stable if the circuit is. Can you tell us the part numbers of those regulators? I want to make sure I never select them. Hmm. Obviously you don't work with highspeed devices like FPGAs and DDS's.. Which need high bandwidth, very fast stepping regulation. The old style of large-caps-slow-regulators doesn't work, you'd end up with a heaving mess of noise. I'm not sure if you are attempting to insinuate that what I'm doing is somehow wrong. It's not, in fact it's perfectly normal in my line of work. Just because you aren't familiar with it, doesn't mean it's wrong. Digital stepping on VCC can happen in the MHz range as I/O toggles and clocks are being buffered, etc. Once you get into this realm, the "regular" way of doing things no longer applies. Besides, I'm not rectifying anything. This is POL design. Rectification happens elsewhere. I bus higher voltage rails to the point of load and regulate at the device itself, and only for that device. This isolates noise and current glitches. This offers a multitude of benefits that transcend application. 1. Higher voltage rails for less current noise across planes. 2. Smaller bulk capacitors needed due to close proximity of regulation. 3. Decrease in parasitic trace inductance issues due to proximity of regulation. 4. Fewer power planes, and/or power traces needed resulting in reduced layer count. 5. Localized ground bouncing due to shorter current return paths. 6. Tighter layout abilities. 7. Localized EMI. 8. Lower noise regulators available due to lower current requirements. 9. Spreads heat around the board rather than concentrating in one power supply. 10. Isolates noise and current glitches to specific components keeping noisy devices from polluting the system. So I can do with a 47uF tant, a 10uF ceramic, a couple 1uF ceramics and a SOT23-5 regulator inside of a 1cm^2 area that takes you a mucho expensive 1000uF cap, a large and wasteful regulator and 100uF cap, not to mention the couple voltage set resistors and 2 diodes you need for reverse voltage protection... AKA my 1.25$ worth of parts to your 3$, and my 1cm^2 to your 2"^2 area. The 4898-2 gets hot even with 10mV of noise on it's I/O. There are no oscillations present. However, if I remove the PCM1794 and leave the 4898-2 I/V stage inputs hanging wide open, they do not get hot. I mean, the 4898-2 does have a thermal paddle on it's bottom for a reason.. |
|
|
|
Post by jimwilliams on Mar 3, 2015 11:23:10 GMT -6
The Mytek/Levenson DAC I designed with Michal at Mytek also used a POL design, with through hole LT1085 regs, those are easily replaced if there is a service need. Reliability and ability to service/fix was important. Saving a couple of bucks wasn't as important as manufacturing a solid product that wouldn't cause service problems as all things break. Swapping a chip was prefered to throwing out the entire pcb.
If the AD opamps were running hot with your regulators and something changed to allow them to run cooler, I'd look for the cause/fix.
I've replaced 5 pin SOIC regs on Antelope converters, a PITA as the pcb traces can also get damaged as the pcb is the mechanical connection and heat sink, a less than ideal combination. I understand that costs are always issues but I usually draw the line at potential service issues making up the difference downstream.
Burrbrown doesn't specify wideband regulation to power their converter chips, 317 types will do fine.
|
|
|
|
Post by svart on Mar 3, 2015 12:19:04 GMT -6
The Mytek/Levenson DAC I designed with Michal at Mytek also used a POL design, with through hole LT1085 regs, those are easily replaced if there is a service need. Reliability and ability to service/fix was important. Saving a couple of bucks wasn't as important as manufacturing a solid product that wouldn't cause service problems as all things break. Swapping a chip was prefered to throwing out the entire pcb. If the AD opamps were running hot with your regulators and something changed to allow them to run cooler, I'd look for the cause/fix. I've replaced 5 pin SOIC regs on Antelope converters, a PITA as the pcb traces can also get damaged as the pcb is the mechanical connection and heat sink, a less than ideal combination. I understand that costs are always issues but I usually draw the line at potential service issues making up the difference downstream. Burrbrown doesn't specify wideband regulation to power their converter chips, 317 types will do fine. I hear what you are saying about serviceability. 10 seconds with a hot air pencil and tweezers and the SOT23-5 package comes off, and another can go on in it's place. No solder sucking of through-holes, etc. Maybe it's just me, but 15 years of working in SMD, I don't have the patience to work on through-hole parts ever again. I did 317 regulators for the ADC board. They work fine, but the design was simple. I still wish I had done a POL design for the ADC for improved heat spreading but it's now a done deal. With the dual PCM1794 design and the large number of supplies that were needed for the parts and to appease those asking for isolated supplies, having a bank of large regulators was not possible without starting to have a cramped and noisy design, thus my switch to POL on this design. The DAC board draws considerably more current than the ADC board overall and spreading heat around the board was necessary as there would be no way that I could deal with concentrated heat from a bank of regulators without resorting to forced cooling. Maybe it's a case of 6/half-dozen, but where you see finicky regulators that are unstable, I see high performance regulators that need to be tuned to work best. Besides, I was working with a seasoned veteran engineer once and he kept having issues with modulation of an oscillator system. It turns out the venerable LM7805 "unconditionally stable" regulator was slightly oscillating due to the extremely low loading of the oscillator and lack of proper capacitance. I suggested either moving to an LDO regulator or using a larger value cap with higher ESR. He elected for the cap change and the modulation stopped. The moral is that even the parts taken for granted as being "stable" are susceptible to the odd operational fluke in the right conditions. The datasheet can't account for all situations. |
|
|
|
Post by tonycamphd on Mar 3, 2015 13:06:03 GMT -6
The Mytek/Levenson DAC I designed with Michal at Mytek also used a POL design, with through hole LT1085 regs, those are easily replaced if there is a service need. Reliability and ability to service/fix was important. Saving a couple of bucks wasn't as important as manufacturing a solid product that wouldn't cause service problems as all things break. Swapping a chip was prefered to throwing out the entire pcb. If the AD opamps were running hot with your regulators and something changed to allow them to run cooler, I'd look for the cause/fix. I've replaced 5 pin SOIC regs on Antelope converters, a PITA as the pcb traces can also get damaged as the pcb is the mechanical connection and heat sink, a less than ideal combination. I understand that costs are always issues but I usually draw the line at potential service issues making up the difference downstream. Burrbrown doesn't specify wideband regulation to power their converter chips, 317 types will do fine. I hear what you are saying about serviceability. 10 seconds with a hot air pencil and tweezers and the SOT23-5 package comes off, and another can go on in it's place. No solder sucking of through-holes, etc. Maybe it's just me, but 15 years of working in SMD, I don't have the patience to work on through-hole parts ever again. I did 317 regulators for the ADC board. They work fine, but the design was simple. I still wish I had done a POL design for the ADC for improved heat spreading but it's now a done deal. With the dual PCM1794 design and the large number of supplies that were needed for the parts and to appease those asking for isolated supplies, having a bank of large regulators was not possible without starting to have a cramped and noisy design, thus my switch to POL on this design. The DAC board draws considerably more current than the ADC board overall and spreading heat around the board was necessary as there would be no way that I could deal with concentrated heat from a bank of regulators without resorting to forced cooling. Maybe it's a case of 6/half-dozen, but where you see finicky regulators that are unstable, I see high performance regulators that need to be tuned to work best. Besides, I was working with a seasoned veteran engineer once and he kept having issues with modulation of an oscillator system. It turns out the venerable LM7805 "unconditionally stable" regulator was slightly oscillating due to the extremely low loading of the oscillator and lack of proper capacitance. I suggested either moving to an LDO regulator or using a larger value cap with higher ESR. He elected for the cap change and the modulation stopped. The moral is that even the parts taken for granted as being "stable" are susceptible to the odd operational fluke in the right conditions. The datasheet can't account for all situations.you got that right, thats why listening tests are SOOOOOOOOooooooooooo important 8) |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Mar 3, 2015 14:37:57 GMT -6
svart and jimwilliams, i so much enjoy your discussions here for it's wealth of information and experience in design questions. There is so much to learn from this. Thank you for that! Also thanks for sharing your knowledge in such a patient and reasonable manner. Although you obviously don't agree in all aspects of the topics or at least seemingly have different viewpoints and approaches to a specific technical topic/problem, you both tend to stay very rational and focused and continue to back up your statements with a load of worthy knowledge. I highly appreciate this kind of discussion. It's like a greek antique dialogue type teaching book for electronics sometimes. Sorry for this non-technical intermission, but i felt i had to write some words about this...
|
|
|
|
Post by Johnkenn on Mar 3, 2015 18:30:41 GMT -6
|
|
|
|
Post by Johnkenn on Mar 3, 2015 21:45:29 GMT -6
|
|
|
|
Post by svart on Mar 4, 2015 10:58:03 GMT -6
Yeah I can see that. I've revised a number of things about the unit after doing thermal testing. The ADC is all done and ready, but the DAC is pretty damn hot overall. Besides the uninformed banter about things not getting hot, these things do indeed get hot without any of the obvious issues you'd expect, like oscillation, etc. The DAC chip itself puts out over a quarter of a watt when operating. However it's still puzzling that the opamps do not get hot unless the DAC is installed. I'm assuming that the 8mA being drawn into each of the DAC outputs has something to do with it. A number of forums also mention the I/V stage opamps getting hot. I have more parts coming to do more debugging, along with a second batch of boards that fix all the issues I had with the first. I just need to figure out if this heat is "normal" or if it's something I need to fix and we'll be headed down the road to the finish line. |
|
|
|
Post by svart on Mar 4, 2015 13:24:05 GMT -6
So actual measurements of the DAC chips:
3.3V = 44mA = 145mW
5.0V = 40mA = 200mW
x2 = 690mW
DAC Regulator power drops and waste heat:
15V-5V = 10V*40mA = 400mW
15V-3.3V = 11.7V*44mA = 514mW
x2 = 1828mW
1828mW+690mW = 2518mW
And then add another 1000mW for the receiver regulator dissipation, and then another 200mW (60mA*3.3V) for the receiver IC itself..
2518mW + 1000mW + 200mW = 3718mW
Plus the drop across the main LDO regs, 24V-15V=9V*500mA(est)= 4500mW
4500mW + 3718mW = 8218mW
Plus the 6 opamps, estimated around 300mA draw at full power for 4500mW
8218mW + 4500mW = 12718mW
So we have the DAC board possibly putting out 13 watts of heat.
That's a lot of heat, and I believe it. The sucker gets warm.
|
|
|
|
Post by tonycamphd on Mar 4, 2015 13:49:53 GMT -6
^why would anyone respond to this technical post? So you can call it "uniformed banter"? Jim Williams is uninformed? lmao
you do realize you're killing yourself here? after reading the last couple pot shots at Jim in as many posts, i just can't take it anymore, good luck.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Mar 4, 2015 13:53:40 GMT -6
Wow, that's alot of heat.
Do you expect any problems having both boards in your case because of this?
|
|
|
|
Post by svart on Mar 4, 2015 14:48:05 GMT -6
Wow, that's alot of heat. Do you expect any problems having both boards in your case because of this? The ADC will be fine. I'm working on the DAC. Most of the IC's will be fine at higher than ambient operation, but the opamps are a bit hot for my liking. |
|
