O2 amp CRC, diode, cap, and heatsink mods
I received my O2 headphone amp PC boards today! I was able to try out a couple of mods I had in mind. Here are some pictures in case anybody else is thinking of trying something similar. It all works just great - I'm listening to the amp right now.
I should say upfront that none of these are likely to be audible improvements of any sort. RocketScientist has done a great job on the design. These mods are just for fun.
1. rectifiers changed to 2A 100V ultrafast silicon rectifiers powered by the 16VAC 1000mA Triad unit in the BOM (WAU16-1000). The mouser link for the diodes is
MUR210G ON Semiconductor Rectifiers
The first LT Spice plot shows the peak charging current hits about 1.8A in the original circuit due to the half wave rectifier. As RocketScientist noted early in the O2 thread the IN4002s are just fine regardless. They are 1A DC, but surge to 30A non-repetitive and at least 6A repetitive from the data sheet graph. But for fun I'm subbed in 2A diodes so the peak is under the DC level.
Ultrafast diodes usually don't add anything at 60Hz, but they are being used here to help limit reverse recovery current that might ping the resonant tank formed by the diode capacitance and the secondary leakage inductance. Highly unlikely that would ever be an issue, but the diodes are cheap so might as well. :)
These particular diodes are in the same DO-41 case as the 1Nxxxx so they fit in the holes just fine.
2. Input capacitor circuit converted to a CRC filter and 0.1uF MLCC ceramics placed across the electrolytic caps.
This one took a bit more work. I first cut the rail traces between C2 and C4, and between C3 and C5, as in the first two photos below. One of the traces was on both the top and bottom of the PCB and had to be cut in both places.
Then on the underside of the PC board I soldered in 0.1uF 50VDC MLCC ceramics across the 4 electros, then jumpered the electros with 10R 1W 1% metal film resistors to form a CRC filter on each rail.
Using the filter calculators on this site for the original RC filter (with the 1.3R transformer secondary impedance and the two 470uF in parallel) and the new RCRC filter (1.3R - 470uF - 10R - 470uF)
RC Low-pass Filter Design Tool
moves Fc back to about 11Hz and doubles the slope since it is second order. The photos below show the capacitors and resistors mounted on the back of the PCB. It all slides into the case just fine. Plenty of room below the PCB.
From the LT Spice plot below for the CRC filter the output has significantly less ripple variation, and a lower voltage, going into the voltage regulators - the red and aqua plots. The peak charging current through the rectifiers is also reduced.
3. Finally, I was able to get Avid DIP8 heatsinks on both NJM4556 chips, as the photo shows. I had to bend one fin on each slightly to avoid the electrolytics, and had to trim off about 1/8 of the bottom clip with wire cutters to clear the film caps. One heatsink touches the mosfet, but I chose the 30V gate units for both and that particular one is fully insulated. They work well! The chip heat transfers to the sinks quite well.
Congratulations to RocketScientist on a great sounding headphone amp. :)
I forgot to say that I also changed all 4 470uF 35V electrolytic caps to these with twice the ripple current rating
EEU-FR1V471B Panasonic Electronic Components Aluminum Electrolytic Capacitors - Leaded
The original caps are rated at 650mA ripple with a 0.8 de-rating factor for 60Hz = 580mA, which is fine for the case with two in parallel to share ripple duty. But for the CRC filter the input cap views a much higher ripple current than the output cap, as the plots below show. The new caps are rated at 1790mA ripple with a 0.65 de-rating factor for 60Hz = 1160mA. The caps are the same diameter, length, and lead spacing as the originals so they fit right in, plus 8000 hour life vs 2000. Just $0.71 vs. $0.08. :) Only the input caps (C2, C3) need to be changed - the originals are fine for C4 and C5. Without the change C2 and C3 would probably get warm under load.
I also found a typo. The load resistors are set to pull about 200mA to simulate max amp load plus max quiescent on each channel, but I labeled it 240mA and unfortunately calculated the CRC resistor wattages from that, coming up with 1W 10R which is what's in the photo of the PCB back. At 200mA max just 1/2W 10R resistors will actually work fine (0.4W max) and be even easier to solder on the back of the PCB.
In the sim plot of the original circuit I(C1) and I(C3) are the currents through the 470uF caps in parallel, about 0.8A peak each (cap ripple rating is RMS, not peak). In the plot of the CRC circuit the red is the input capacitor at a peak current of about 1.2A, while the aqua is the output capacitor at about 270mA peak.
Re: rectifier diodes, commonly one does not even want them to be ultra-fast. They'll just be mixing whatever RF is caught by the mains wiring with supply frequency and its harmonics, generating intermodulation that shows up as hum on an AM receiver. This is particularly noticeable if said receiver is powered from this very supply - I have one such case here, Sony ICF-7600A with the original AC-456C AC adapter, absolutely terrible on MW, fine on batteries or with another adapter...
Luckily, adding 22..100 nF caps across every rectifier diode seems to eliminate the problem. As noted, they don't really do anything about secondary resonance save for moving it to a lower frequency, but they make a good RF bypass for our wannabe mixer.
The O2 design unfortunately omits such diodes, but the length of cable from an external wall-wart type transformer should make for a reasonably useful inductance.
A toroidal xmfr + super-fast rectifier diodes could become a serious EMI problem...
sgrossklass - thanks for that! I was actually stewing over that very issue but assumed the following CRC filter would effectively snuff out any RF, especially with the 0.1uF across the electrolytics.
I was thinking about adding the snubber capacitors across the diodes. Like you say, if the issue is diode capacitance forming a resonant tank with the transformer then parallel caps will just lower the resonant frequency, which wouldn't be a bad thing if the filter caps then nail it. Unless it is snubber caps in series with 10R or so to lower the Q and provide some loss to damp the tank, which might be even better. Lol - maybe another O2 mod is born. There should be room above the diodes for a series cap + resistor in parallel with each.
I'm having some fun with the mods using the "z" dimension above and below the PCB. :) I know the PC board as sits is out of space so there would be no way RocketScientist could add any of this, at least without tombstoning some parts or using some SMD on the back, or making the whole PCB larger.
Too late to add to my post above, lol. :) More info. I was working on a couple of old GE "Super Radios" a couple of weeks ago and noticed on the PS portion of the schematic the PS rectifiers were bypassed with 0.01uFs
Those old radios big claim to fame was AM sensitivity. I have to admit it never occurred to me they may be bypassing the diode RF mixer effect there rather than just snubbers. That is very interesting. Maybe going back to the good old 1N4002s with the snubber/bypass caps added is the best after all. :)
Note there is a stealth CRC filter in that schematic too. C66 the 1000uF is the input cap, R73 the 10R is the "R" and the 15V cap multiplier made up from Q7 is effectively the output "C", then off it goes to feed the audio circuits.
neg rail 2.2uf, 9.6V nimh batteries, and snubbers
Some more O2 mods in case anyone is pondering something similar. :D Again, none of this is likely to have any audible, or even likely measurable, difference. Just for fun. :)
1. Increase the negative rail transient capacitor C7 to 2.2uF.
The LM7812 and LM7912 regulators are almost complements, but not exactly. The datasheets show a recommended 0.1uf stability capacitor C6 after the regulator for the LM7812, but one nearly 10 times bigger at 1uF for the negative LM7912 (C7). Figure 17 in the LM7812 data sheet shows it for dual rail supplies
The datasheets for the mc7812 / mc7912 regulators on the BOM also mention the issue for capacitor "C0" in the writeup in the lower right corner on the first page
RocketScientist likely made the positive rail capacitor C6 0.22uF to take care of tolerance variations, which is good practice, over the minimum 0.1uF. So keeping in line with that I'm using a 2.2uF cap instead of the minimum 1.0uF for C7
MMK5225K50J04L16.5TR18 Kemet Polyester Film Capacitors
The trouble is that even the flattest form factor here with 5mm leads still is not thin enough for the PCB space available. The capacitor goes right in the holes just fine and sits flat on the board, but then is about 1mm too wide to allow the board to slide into the box without hitting the side of the box.
So the solution is... second and third photos. Mount the capacitor sideways above the schottky rectifiers. :) Photos are before and after insulating tape on the leads. The longer leads will add some inductance which should be insignificant here.
2. 9.6V NiMH batteries.
I've already posted about this weeks ago in the O2 thread, but here is the final result in the 4th and 5th photos. These are the 8-cell 9.6V "9V" NiMH Maha powerex 230mAh batteries rather than the more standard common 8.4V 7-cell "9V" NiMH batteries in the BOM.
MAHA / POWEREX 9.6V 230mAh Rechargeable NiMH Battery
R1 and R2 are changed to 150R 1W. RocketScientist probably picked 1W for the original resistors to handle the case where the batteries might dead short, which would leave the full 12V across the 220R charging resistors. The same applies here, if the batteries dead short the 1W 150R resistors can handle it (0.9W in that case). As I mentioned in the O2 thread, if anyone makes the resistor mod it is very important to put a sticker under the batteries saying "9.6V 8-cell NiMH only". If a 8.4V NiMH battery was installed with 150R resistors the charnging current would initially be 30mA, which is double what the original design intended and would result in twice the trickle charge rate at the end of charge.
3. 1/2W 10R CRC resistors
As I noted above I mis-calculated the wattages of the "R" resistors. 1/2W works just fine and I've made the change in the 6th photo.
Also as posted above I've added 0.1uF 100V snubber capacitors across the rectifier diodes in the 7th photo. It is kind of hard to see, but from the lettering on the caps the higher ripple current 470uF 35V caps can also be seen in the photo with the original 470uF 35V caps behind those.
Updates on C7 and 9.6V battery mods, plus 4.7uF mod
A couple of updates on the O2 mods I posted above...
1. C7 neg rail transient cap now a 2.2uF 25V MLCC ceramic
I found a 2.2uF capacitor that fits in the existing PCB board space/holes perfectly and allows the B2-080 case to slip right on with room to spare. Doesn't look as cheesy as the metal film installed sideways, above. :)
RDEC71E225K1K1C03B Murata Multilayer Ceramic Capacitors (MLCC) - Leaded
In the photo below the cap is that little blue one in the C7 position between the negative regulator and the board edge.
2. 9.6V NiMH charging resistors now 75R 2W
I received the data sheet from Maha on their 9.6V 8-cell NiMH and had a chat with tech support. Their suggested long-term trickle charge rate at end-of-charge (around 1.43V per cell) is 5mA, actually a bit higher than the 4mA I had designed for. So never passing up a chance to charge faster (!) I've changed the charging resistors to 75R 2W units that fit in just fine. RocketScientist did a great job of leaving enough space for several resistor sizes. This one is a mini 2W model
MOS2CT52R750J KOA Speer Metal Oxide Resistors
As can be seen in the photo below I have them soldered in just slightly above the board. The 2W end of things would only happen if a battery dead shorted and all 11.7Vdc went across the resistor. Normal charging doesn't come anywhere near 2W.
As for the resistor value, from some measurements the regulator voltages are both on the low end of tolerances at +/-11.94Vdc. Minus the Schottky drop leaves +/-11.72Vdc going into the charging resistors. With a 1.43V per cell charging endpoint that gives [11.72V - (8 * 1.43V)] / 0.005A = 72R. 75R is what Mouser has in stock.
3. 2.2uF coupling caps C13 and C14 changed to 4.7uF
This is another one of those mods that will have just zero audible effect and would probably even be very hard to measure any sort of difference. So why do it? Because I can! :D This is the largest stacked metal film I can find at Mouser that will fit in the existing holes and PCB space. You can see them at the top of the photo below, those two red Wima caps
MKS2B044701K00KSSD WIMA Polyester Film Capacitors
The electrical effect is plotted below. The first sim plot is with the original 2.2uF coupling caps. On the left side the response is at 9.3dB at 10Hz. In the second plot with the 4.7uF caps the drop is "only" about 9.5dB. :)
Mainly what this shows is why bigger coupling caps don't improve anything here. My headphones don't reproduce frequencies that low, my ears can't hear frequencies that low, and even then the improvement difference is miniscule.
A better, second, use for these bigger 4.7uF caps would be offset reduction. With the coupling cap values doubled to 4.7uF the resistors R12 and R13 could be cut in half to 20k and still keep Fc of the filter the same (leaves the 10Hz gain roll-off at 9.3dB). But what you gain is cutting the output offset voltage in half, since the input bias current of the op amp now goes through half the resistance. This has the potential to lower the already very low output offset voltage (the offset voltage the headphones see) even lower. But that also is not "necessary" in any way - the level is already too low to be of any concern - just something to mess with for fun.
4. The LME49720 in for U1
I currently have a LME49720 in for U1 and can report that despite my very best efforts to hear something, anything, "better" about the $3 chip vs. the NJM2068, I have to admit that so far I can't. But I'm going to keep trying. :) I have a second O2 built up the standard way with all of RocketScientist's specified values - no mods. I'm doing the A/B thing between them. If RocketScientist were to hook his dScope up there would probably be some very tiny and subtle improvements in some parameters, but whatever they might be are just way below my threshold of hearing it seems.
Voltage regulators mounted on backpanel modification
Here is another O2 mod that is not "necessary" in any fashion, but again just done for fun. I saw a big old metal box sitting there and just couldn't resist messing with it. :D
This mod moves the two O2 voltage regulators, the MC7812 and MC7912, out to the back panel of the B2-080 chasis where they are mounted. This heatsinks both regulators to the back panel, and hence to the rest of the metal chasis once the back panel is screwed onto the box.
The mod makes use of the typical TO-220 insulated mounting kit with a nylon sleeve washer that goes through the hole and a mica washer for the back.
Zinc oxide heat sink compound is put on both sides of the mica washer. The mounting hardware is 4-40. I had intended to use stainless steel button head (allen wrench) M3 metric hardware since it looks cool, but it was too big for the hole. I'll probably get some 4-40 socket head screws as time permits. Here is a great place to get bolts - they even have chrome hardware!
Bolt Depot - Nuts and Bolts, Screws and Fasteners online
I have no connection with them other than having bought a bunch of their stuff.
First off, two holes have to be drilled in the back panel to mount the regulator chips, The two voltage regulators will be mounted sideways in the space above R1 and R2, between those two resistors and the top of the box. The regulators are just barely too tall to be mounted vertically.
The first photo shows the holes. Anyone who has done metal work knows that despite careful measurements things rarely work right the first time. This one was no exception. The top panel in the photo was the first try. The one below is the final result after discovering the holes were 2mm too far to the right and hit the corner extrusion where the screw goes in. The holes are M4 metric which turned out to be just a perfect fit for the nylon insulating bushing.
And.. I was planning on posting measurements for the drill holes at the point, but in looking at the pictures it just occurred to me I have the box upside down! :p I've just checked and the mod still works fine the right way, but I will have to made slight measurement changes since they are referenced to the bottom and right side of the chassis. I will post the measurements another day. For now this is works as a proof-of concept. :)
Next the voltage regulators have to be prepared. These are new regulator chips since after removing the existing regulators from the PCB the leads were too short to use here. I have bent the leads out 90 degrees from the body so parts and wire leads can be attached. Since the regulators will be extended from the PCB on some 20ga wires, it is important to solder bypass/decoupling capacitors right onto the regulators. I've followed the same schematic I mentioned in a post above, figure 17 in this data sheet for the LM78xx series
I've soldered all the capacitors and diodes shown, with the values shown, with the exception of the 0.33uF being replaced by a 0.47uF since that is what I had on hand. The diodes are the typical "polarity reversal protection" diodes found on many dual rail power supplies. I've been intending to add those as a separate modification, but decided to just do it here. All capacitors are MLCC ceramic 50V or 100V.
The 2nd and 3rd photos show the result for the positive MC7812 regulator. You can see the two capacitors and diode soldered on. All soldering was done on an anti-static mat, with a grounded wrist strap, and grounded soldering iron. Very important to play nice with the semiconductors when handling them a lot like this!
The 4th photo shows the wire leads attached. This is 20ga silver wire with teflon insulation. No particular reason for that other than I like the fact the insulation doesn't pull back with soldering iron heat the way pvc does. At this stage the wire lengths are just roughed-in to be long enough to go from the back panel to the voltage regulator holes in the PCB.
The 5th photo shows the insulator kit, while the 6th photo shows the chip subsequently mounted on the panel with the (white) zinc oxide heat sink compound.
The negative regulator is prepared the same way so I'm leaving out the prep steps for that one, but do note the order of the leads on the chips are different. The 7th photo shows both regulators mounted on the back panel. The 8th photo shows the wires bent to fit and soldered into the holes on the PCB. The 9th photo shows the whole back panel and PCB assembly slipped into the box which is great - except as noted above I got it upside down. :p
Using a surface temperature IR meter I measured around 142 F (61 C) at the voltage regulators when soldered into the PCB board. That was with my headphones being powered (44R) and charged batteries attached. RocketScientist has posted on his blog that 60C or so is perfectly OK since the parts are specified up to 125C. So there is absolutely no "problem" of any kind here with the original PCB mounted regulators. This mod is just for fun and has one minor benefit of moving some of the dissipated heat from the regulators out of the inside of the box to the metal case.
I'll post some temperature readings in a few days of just the back panel, plus the back panel mounted onto the box.
Here are the measurements for the O2 voltage regulator mod above - with the metal case turned the right way around. :p Turns out my original measurements worked just fine. They are even fit a bit better now. The case is symmetrical about the center line from top to bottom. The PC board just mounts 1.5mm higher in the right direction and the channel for the screws is higher. That channel is what caused my first set of measurements to be off and hit one regulator body, requiring drilling the second set of holes, so now it fits even better.
Top of the case is the side with the two rows of groves. Top of the metal back panel is the side with the 2 mounting holes closest to the edge.
Both regulator holes are 19.7mm up from the bottom of the rear panel. One is 43.2mm over from the right edge, the other is 24.2mm. I used a center punch to mark the holes, a 1/16" bit to pilot the hole, then a 4mm metric bit for the final hole.
I've also taken some temperature readings with the amp on and powering a headphone for about 2 hours. The rear panel directly opposite the regulators is 102.5 F. The short piece of rear panel to the right of the regulators was 103.0 F, while the long panel to the left of the regulators tapered down from 102.5 F to 97F at the far edge. The regulators themselves measure 103.0 F now, down about 40 F from my measurements of the regulators on the PC board.
This is just the panel in free air too, not mounted onto the box yet! Once mounted I'll bet the temperatures are just 10-20 F or so over room temperature. All in all I'm pretty happy with the outcome of this mod. :)
O2 Headphone output delay-on relay mod
Here is yet another O2 mod that is completely unnecessary but fun to build regardless. :) This one adds relays to the output of the O2, between the 1R resistors and the headphone jack, that an associated control circuit uses to delay turn-on by about 2 seconds. The control circuit also will block turn-on if the rail voltages are less than about 14Vdc, the amount of two completely dead NiMH batteries down to 7V each, or more importantly one battery in and one battery out.
The power management circuit RocketScientist created in the O2 is excellent! I couldn't have come up with something that slick. It works very well and this relay project is not "adding" any protection whatsoever to what is already there in the O2. This project is just for fun. In fact, the control circuit here is triggered by the existing power management circuit in the O2. When that circuit trips it powers up the op-amps in the usual way. This circuit just adds an additional 2 seconds before the relays kick in to connect the headphones.
In big amps there is a reason for this kind of thing, to let the amp circuits stabilize before connecting the load. But in a small headphone amp like this, especially given how very well RocketScientist's power management circuit is working, there just is no need for relays. But here they are anyway. :D I think RocketScientist may have posted somewhere he was thinking about a relay circuit of some kind for the desktop version, too, where there would be more board space.
The relays are from RadioShack, of all places. Who said there isn't any good stuff at RadioShack anymore? Mouser and Digikey both have extensive selection of reed relays, including DPST types that would only require one relay, but this was just sort of a challenge to see if I could get the rat shack relay to work.
The relay is a 12Vdc SPST relay, so two will be needed, one for each channel. The relay is the first photo. The coils are wired in series to give a 24Vdc total since this will be going across both rails. The relay coil pulls 11mA, about as much as two njm2068 chips, so it will reduce battery life slightly. Some measurements came up with about 6Vdc for the pulldown voltage (minimum voltage to engage the relay) and about 4V for the dropout voltage (lowest voltage before the relay lets go). That is good news since in addition to AC power, 24V rails, the relay still needs to engage and hold with two fairly dead batteries - 14Vdc rails or about 7V per battery. Note that 6Vdc per coil would be a minimum of 12Vdc pulldown for two in series. Since they are in series the 11mA coil current goes through both, of course, so that is just one single 11mA current draw.
The next photo shows two of the relays super glued together side by side. They will be mounted above the headphone "out" jack using some super-sticky urethane sealant. Turns out there is just enough room there to clear the metal box with the B2-080 chassis.
The control circuit diagram is next, along with a plot of the function with 24V rails (AC power) and 14V rails (2 dead batteries). When the O2 power management circuit switches on C2 here is initially a short, of course, and begins charging at a time constant set by R4 and R5. The R4/R5 voltage divider here is used to keep the maximum voltage to the mosfet gate under its 20V maximum. There is a second reason, though. If just one resistor were used the mosfet gate capacitance charging current and mosfet leakage current would substantially reduce the time delay. With the divider just one fairly small capacitor, a 2.2uF MLCC 50V XLR, gives a 2 second delay. That is the point where the M2 mosfet gate voltage drops below 2.5V and turns off.
The mosfet is across the timing resistor rather than the capacitor due to the shape of the charging curve. The capacitor has a fairly fast intial charge accumulation followed by a long tail. By inverting things here we get advantage of the tail. But that means the output is the mosfet "off" rather than "on" to turn on the relay. That means an inverter is needed, which is the purpose of M3. It performs a logical "NOT" along with the pullup resistor for M2. This pullup though provides an opportunity to logical "OR" another control signal in. Pondering it a bit I added the 10V zener which has the effect of keeping M3, and hence the relay, off for any rail voltage less than 10V + mosfet threshold (about 2.5V) = 12.5. Therefor one single battery, no matter how charged, won't be enough to turn on the circuit. Another function of the zener is to keep the gate voltage of M3 just barely on, at about 3Vdc, when the applied voltage is about 14V. That arrangement actually provides accelerated "off" timing for the relay, as can be seen on the plots over just the natural rail voltage drop timing.
The only place on the O2 PCB big enough to hold the control circuit parts was above the AC rectifier diodes. Luckily the control circuit proved to have enough noise immunity. :) The third photo shows the complete control board next to the relay. Note that the position of the 27V reverse spike protector across the relay changed from this photo. I had forgotten about an effect when two reed relays are placed next to each other, as described in figure 6 here
The external magnetic field lines will cut through the second relay and either buck or add to that relay's field, depending upon the direction of current through the coils. My photo here was matching polarity, which would slighly buck. I later changed it to anti-polarity to make the magnetic fields additive. This is great since the two relays are to engage together anyway, one for each channel, so if they assist each other the more the merrier.
Also as most folks here probably know, a zener across the relay coil provides faster turn-off than just a suppression diode like a 1N400x. Here is a good writeup on that - see the chart at the very bottom for turn-off times vs. suppression methods
To make the connections to the O2, the end of resistors R10 and R11 closest to the front of the PCB have to be unsoldered, pulled out, and resoldered together above the PCB. This reconnects these two 1R resistors but disconnects them from the P2 holes. One end of one relay's contacts is then connected to these two soldered resistors, and the other end of the contact is connected to the P2 hole furthest from the front board edge. That in turn connects to the headphone jack.
Similarly a trace going from the second hole up in P2 to that via under C9 that RocketScientist mentions soldering must be cut. The next photo shows that cut on the back of the PCB. This breaks the connection between the R15 and R18 1R resistors and the headphone jack. One contact of the other relay is soldered to the lead of R15 closest to the PCB front edge. The other contact is soldered the second hole up in P2. That connects the second relay contact to the headphone jack.
Next the wire leads need to be prepared on the relay. I'm using 26 gauge silicon covered multi-strand test lead wire from here
Test Lead Wire - Silicone Test Lead Wire - Pomona Silicone Test Lead Wire - Cal Test Silicone Test Lead Wire
since it is uber-flexible to get around small places. The red and black are the coil connections while the green are the relay contact connections.
Lol - I'm overly verbose and used up my photo quota here. :) Continued below...
|All times are GMT. The time now is 10:51 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio