Nada Posted July 24, 2013 Posted July 24, 2013 (edited) Does this board look any good? Edited August 14, 2013 by Nada
hochopeper Posted July 24, 2013 Posted July 24, 2013 (edited) Honestly, not great IMO. The chip at the heart of it is one of my favorites though. Those heatsinks won't do much other than look cool. The thermal conductivity of the plastic case is not great, I did the maths for heatsink through chip case like that for a power amp that uses smt diamond buffer chips ... Thermal limit is given by the PCB design, that design won't get 1A, not even close. The noise performance is dictated by capacitor choice and they've just plastered numbers from the data sheet onto the PCB silk screen, I wouldn't expect it to measure as well as they suggest with that layout. Hopefully something similar and locally designed will become available later in the year. Edited July 24, 2013 by hochopeper
Nada Posted July 24, 2013 Author Posted July 24, 2013 So if proper heatsinks were placed on the TPS7A4700 and better caps used it could be OK?
hochopeper Posted July 24, 2013 Posted July 24, 2013 The heat is all dissipated by the copper in the PCB, so it is probably OK for 100 - 200mA per regulator. Changing the heatsinks won't help at all. Depends on the application ... What're you planning using them for? 1
Jventer Posted July 24, 2013 Posted July 24, 2013 Hopefully something similar and locally designed will become available later in the year. Now you have me very interested.
Nada Posted July 25, 2013 Author Posted July 25, 2013 (edited) The heat is all dissipated by the copper in the PCB, so it is probably OK for 100 - 200mA per regulator. Changing the heatsinks won't help at all. Depends on the application ... What're you planning using them for? Thanks for your expert help yet again mate. Blimy. I would never have guessed the heat is dissipated internally into the middle of the board. That's a bit odd isnt it? Its just asking for thermal overload. I want it for powering a mico-computer like a Beagleboard, an XMOS USB-12S board and two clocks. So this board looked really nice with two 5v and two 3.3v supplies but if it cant pump out the power its not going to cope. Bummer. Please tip us off when a well designed option is available. PS - how good are Hyne's regulators etc compared to the TPS7A4700 ? Edited July 25, 2013 by Nada
fetischizm Posted July 25, 2013 Posted July 25, 2013 (edited) PS - how good are Hyne's regulators etc compared to the TPS7A4700 its hard to tell, as i've never seen a single set of measurements, or specs; ever and he paints over or scratches off component numbers. people just seem to trust that they are the best. I dont think they are incompetent, but its hard to form an informed view of technical merits when they are intentionally left blank. heatsinking to the PCB is how the vast majority of heatsinking is done these days with SMD, even for high power devices. its actually pretty efficient if you have designed for it, which this guy hasnt. its not sunk to the middle of the board, its sunk to the bottom of the board to the copper through thermal vias and the component has a powerpad on the underside that is soldered to the copper on the top side. if you think of it that way, doesnt seem that strange right? the problem here is, he doesnt seem to understand the thermal transfer characteristics and hasnt read, or hasnt understood the datasheet for the part. perhaps he knows its not capable of meeting his claims and has simply pulled the numbers from the datasheet headline specs; since the chances of it meeting the datasheet noise performance with this layout are not possible, its hard to tell which. thats what he did with his dac too though, just copied the datasheet performance numbers and printed them on the PCB. with some parts you can then have heatsinks on both sides, the board and the top of the part and in high end industrial power supply designs now, they are using heat-pipes, or liquid cooling fabricated right into the PCB and highly conductive ceramic PCB substrates $$$ Hopefully something similar and locally designed will become available later in the year. :ph34r: Edited July 25, 2013 by fetischizm 1
Nada Posted July 25, 2013 Author Posted July 25, 2013 (edited) Very interesting. Thanks for the heads up, I look forward to a quality local design being available soon then..... Hynes gives these cool specs for the SR1 s : slew rate 5kV/us , wideband noise 2 nanovolts root HZ. Regulator output impedance < 1mohm and the power supply line rejection ratio is 110 dB to 200 KHz Edited July 25, 2013 by Nada
hochopeper Posted July 25, 2013 Posted July 25, 2013 Thanks for your expert help yet again mate. Blimy. I would never have guessed the heat is dissipated internally into the middle of the board. That's a bit odd isnt it? Its just asking for thermal overload. I want it for powering a mico-computer like a Beagleboard, an XMOS USB-12S board and two clocks. So this board looked really nice with two 5v and two 3.3v supplies but if it cant pump out the power its not going to cope. Bummer. Please tip us off when a well designed option is available. PS - how good are Hyne's regulators etc compared to the TPS7A4700 ? Which beagleboard are you looking at? I think the beaglebone needs 700mA or so to hit max clock speed, they can run on less but won't reach peak performance and you'll need to configure it to not try to switch to the faster modes. I haven't read the data sheet for the beaglebone-xm since ~12months ago so can't remember it's power req. I would expect it won't be any less than that though. The xmos chips, depending on which chipset it is, I'd expect it will need ~350mA to run for 192k+ lower sample rate will use less. I have some tps7a4700 based regulators (different PCB) as local supply to clock in my DIY headphone amp + DAC but that's only using maybe 60mA from memory. To supply the two loads you're looking at I would need to do some calcs on thermal performance when I get home again next week to come up with the best options. I already have a beaglebone and a spare waveio so if you're just wanting to see how a setup like that goes then I can hook something up and loan it to you after I get my beaglebone black perhaps? 1
fetischizm Posted July 25, 2013 Posted July 25, 2013 (edited) Chris, Ians PCB wont get to that load current, the 'locally available one' probably wont either well actually there is another one that might I would suggest running 2 regulators regardless Very interesting. I look forward to a quality local design being available soon then..... Hynes gives these specs for the SR1 - cool numbers : slew rate 5,000 volts per microsecond, gain bandwidth product 1 GHz, rise time < 1 nanosecond, settling time < 10 nanoseconds and wideband noise 2 nanovolts root HZ. Regulator output impedance is less than 0.001 ohm and the power supply line rejection ratio is 110 dB from DC to 200 KHz yeah cool enough, but worth the hubbub? the slewrate is a red herring for an external supply and unattainable in reality how it will be used, connected to a load over a cable. the noise is completely beatable, the output impedance is fairly typical of a feedback amplifier due to the error amp and sensing, not that much better than the 4700 actually … but i'm going to stop the critique there, Paul does know what hes doing, I just dont understand the mythical properties they are often claimed to have and $$$ Edited July 25, 2013 by fetischizm
Nada Posted July 25, 2013 Author Posted July 25, 2013 ....... I already have a beaglebone and a spare waveio so if you're just wanting to see how a setup like that goes then I can hook something up and loan it to you after I get my beaglebone black perhaps? That would be great. If you ever get the spare time to put it together please let me know. I just dont understand the mythical properties they are often claimed to have and $$$ Most of us dont know enough to critique electronic designs or interrogate numbers with discrimination so we follow what's recommended. I reckon it must be really cool to understand this stuff and then build your own.
Gordon Macfarlane Posted July 25, 2013 Posted July 25, 2013 (edited) Hi Nada ,Kingwa now has a new psu that might suit : http://www.audio-gd.com/Pro/diy/PSU2013/PSU2013EN.htm I plan on using one for my Anamero usb and dac. He will also sell you his nice R-core trafos for $25 or so. As pointed out by Fet below you need something with higher current output for the Beagle. Analogmetric have a range of linear psu's that might suit the Beagle. Edited July 25, 2013 by Ozcall 1
fetischizm Posted July 25, 2013 Posted July 25, 2013 (edited) thats not suitable for either the beaglebone, or the xmos (let alone both) and Ozcall, I doubt its suitable for your dac either and its not suitable for your pi. see below. I expect the tiny 4700 IC will better it, its a pretty standard design of the type you can find on ebay for less. Note: This is a class A PSU, its current output had limit at 80 to 240 MA Nada: sure, I get that, they are just numbers though, not measurements, they may well come from actual measurements, or they may be predicted, which is more what I meant. I dont think there is any doubt they are good regulators, but as shunt regs go, I do wonder if they are any better than a Salas regulator, which you can unfortunately only build yourself, he hasnt officially licensed the schematic (even though there have been some less than successful attempts to use it without his permission). when Paul started, there werent really any competitors, but even IC regulators are catching up in many aspects these days and are infinitely more practical and affordable. at the currents you are talking about though and purpose, I wouldnt use a shunt reg, way too much heat and the resulting size, would mean it couldnt be close to the board. will you install in the same chassis? I wont comment on the actual performance we're expecting, but I expect it to be rather difficult to measure. Edited July 25, 2013 by fetischizm
A J Posted July 25, 2013 Posted July 25, 2013 They look interesting - and well priced - not for dummies though..... "Connect by wrong will shatter the PSU or the gear."
Gordon Macfarlane Posted July 25, 2013 Posted July 25, 2013 thats not suitable for either the beaglebone, or the xmos (let alone both) and Ozcall, I doubt its suitable for your dac either and its not suitable for your pi. see below. I expect the tiny 4700 IC will better it, its a pretty standard design of the type you can find on ebay for less. Nada: sure, I get that, they are just numbers though, not measurements, they may well come from actual measurements, or they may be predicted, which is more what I meant. I dont think there is any doubt they are good regulators, but as shunt regs go, I do wonder if they are any better than a Salas regulator, which you can unfortunately only build yourself, he hasnt officially licensed the schematic (even though there have been some less than successful attempts to use it without his permission). when Paul started, there werent really any competitors, but even IC regulators are catching up in many aspects these days and are infinitely more practical and affordable. at the currents you are talking about though and purpose, I wouldnt use a shunt reg, way too much heat and the resulting size, would mean it couldnt be close to the board. will you install in the same chassis? I wont comment on the actual performance we're expecting, but I expect it to be rather difficult to measure. Hi Fet here is the dac : http://www.ebay.com/itm/DAC-24192-ULN-24Bit-192KHz-DAC-I2S-Input-Ultra-Low-Noise-Regulator-Circuit-/330886206404?pt=LH_DefaultDomain_0&hash=item4d0a5ad3c4. Current draw is stated as 31ma. I planned to use the A-gd psu to power the dac and Amanero usb board. I already have a linear psu form my Raspberry.
henry218 Posted July 25, 2013 Posted July 25, 2013 audio-gd sells discrete PSU also which you can consider.
fetischizm Posted July 25, 2013 Posted July 25, 2013 Hi Fet here is the dac : http://www.ebay.com/itm/DAC-24192-ULN-24Bit-192KHz-DAC-I2S-Input-Ultra-Low-Noise-Regulator-Circuit-/330886206404?pt=LH_DefaultDomain_0&hash=item4d0a5ad3c4. Current draw is stated as 31ma. I planned to use the A-gd psu to power the dac and Amanero usb board. I already have a linear psu form my Raspberry. ahh, IC, it doesnt have a clock, or an output stage, thats cheating I hope your amp has nice high input impedance? so looks like with the amanero and dac you should allow about 200mA to have a little bit leeway, so make sure you have decent ventilation. one thing, I wouldnt supply the dac with 20V, bet they just pulled that from the datasheet, looks like a standard 'low noise' LDO and the spec of up to 20V usually has restrictions ie, not 3.3v out. i'm sure 5-9vdc will do fine.
hochopeper Posted December 27, 2013 Posted December 27, 2013 I'm in need of a swag of overkill PSU for some measurement and DSP gear. I was expecting another guy to come up with some PSU designs but family health issues have taken priority for him the last few months. I'm now working on a design that will fit a lot of purposes from low noise wide bandwidth ADC power supplies, DSP, FIFO and general digital/mixed signal supplies Plan for now is to have two isolated independent positive regulators on each PCB. Rectifiers, snubber, lm317 pre-regulator, tps7a4700 for main regulator. Should have very low noise output but I am building some measurement pre-amps at the same time to find out exactly how it looks. Hopefully will make good progress over the next month or so though I am concerned Chinese new year will slow down the PCB manufacturing stage. Was hoping I'd have finished PCB before Christmas but as always work got busy and then time away from home has slowed all of that down. When I'm home next week I'll post up some details/thoughts etc and document some of the process here if there is some interest. Chris
Recommended Posts